Step hyperpolarizations evoked slowly activating, noninactivating, and slowly deactivating inward currents from membrane patches recorded in the cell-attached patch configuration from the soma and apical dendrites of hippocampal CA1 pyramidal neurons. The density of these hyperpolarization-activated currents (I h) increased over sixfold from soma to distal dendrites. Activation curves demonstrate that a significant fraction of I h channels is active near rest and that the range is hyperpolarized relatively more in the distal dendrites. I h activation and deactivation kinetics are voltage-and temperature-dependent, with time constants of activation and deactivation decreasing with hyperpolarization and depolarization, respectively.I h demonstrated a mixed Na+–K+ conductance and was sensitive to low concentrations of external CsCl. Dual whole-cell recordings revealed regional differences in input resistance (R in) and membrane polarization rates (τmem) across the somatodendritic axis that are attributable to the spatial gradient of I hchannels. As a result of these membrane effects the propagation of subthreshold voltage transients is directionally specific. The elevated dendritic I h density decreases EPSP amplitude and duration and reduces the time window over which temporal summation takes place. The backpropagation of action potentials into the dendritic arborization was impacted only slightly by dendriticI h, with the most consistent effect being a decrease in dendritic action potential duration and an increase in afterhyperpolarization. Overall, I h acts to dampen dendritic excitability, but its largest impact is on the subthreshold range of membrane potentials where the integration of inhibitory and excitatory synaptic inputs takes place.
- hyperpolarization-activated current
- synaptic integration
- CA1 pyramidal neuron
- action potential backpropagation
The primary site of excitatory synaptic input into hippocampal pyramidal neurons is the vast dendritic arborization that comprises >95% of the membrane surface area of the cell. Here in the dendrites, information received from tens of thousands of synaptic inputs is coordinated and stored via the highly complex processes of dendritic integration and synaptic plasticity. It is now well known that a rich variety of voltage-gated ion channels bestows on dendrites unique active properties that impact all aspects of dendritic function (Yuste and Tank, 1996; Stuart et al., 1997; Magee et al., 1998).
Although the relative densities and key biophysical characteristics of several Na+, K+, and Ca2+ channel types have been reported for CA1 pyramidal neurons (Magee and Johnston, 1995; Hoffman et al., 1997), very little is known about hyperpolarization-activated (I h) channels in CA1 dendrites.I h plays a variety of important roles in many neuronal and non-neuronal cell types (for review, see Pape, 1996) and has been reported to be present in CA1 pyramidal neurons (Halliwell and Adams, 1982; Maccaferri et al., 1993; Gasparini and DiFrancesco, 1997). Various membrane parameters (input resistance, membrane time constant, and resting potential) as well as general membrane phenomena (rectification, oscillatory activity, and action potential firing rates) are all modulated by I h (Mayer and Westbrook, 1983; Spain et al., 1987; Pape and McCormick, 1989;Maccaferri et al., 1993; Maccaferri and McBain, 1996; Gasparini and DiFrancesco, 1997). Thus, the presence of I h in the dendrites of CA1 pyramidal neurons potentially could have a significant impact on the dendritic integration of subthreshold synaptic activity and the propagation of action potentials.
The basic biophysical properties and the subcellular distribution ofI h were investigated in hippocampal CA1 pyramidal neurons, using cell-attached and outside-out patch-clamp techniques. The impact of these channels on the shape and propagation of subthreshold voltage signals and action potentials also was determined by using simultaneous whole-cell voltage recordings from both the soma and dendrites.
MATERIALS AND METHODS
Hippocampal slices (400 μm) were prepared from 5- to 16-week-old Sprague Dawley rats, using standard procedures that have been described previously (Magee et al., 1996). Pyruvic acid (3 mm) and ascorbic acid (1 mm) were added to both the perfusion and incubation solutions. Individual neurons were visualized with a Zeiss Axioskop microscope (Oberkochen, Germany) fit with differential interference contrast (DIC) optics, using infrared illumination. All neurons exhibited resting membrane potentials between −62 and −75 mV.
For channel recordings the bath solutions contained (in mm): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 2.0 CaCl2, 1.0 MgCl2, and 25 dextrose. The external solution was bubbled with 95% O2/5% CO2 at 23 or 33°C, pH 7.4, for all recordings. The standard cell-attached recording pipette solution consisted of (in mm): 120 KCl, 10 HEPES, 2.0 CaCl2, 1.0 MgCl2, 20 tetraethylammonium-Cl (TEA-Cl), 5.0 4-aminopyridine (4-AP), and 1 BaCl2, pH 7.4 with KOH. The “physiological” cell-attached recording pipette solution consisted of (in mm): 110 NaCl, 10 HEPES, 2.5 KCl2, 2.0 CaCl2, 1.0 MgCl2, 30 TEA-Cl, 5.0 4-AP, and 1 BaCl2, pH 7.4 with NaOH. For outside-out patch recordings the pipette solution consisted of (in mm): 120 KMeSO4, 20 KCl, 10 HEPES, 10 EGTA, 4.0 Mg2-ATP, 0.3 Tris2-GTP, 14 phosphocreatine, and 4 NaCl, pH 7.25 with KOH. Pipettes were pulled from borosilicate glass (9–12 MΩ) and coated with Sylgard. The tips were inspected visually before use and had uniform tip diameters of ∼1 μm for both dendritic and somatic recordings. Channel recordings, using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA), were analog-filtered at 1 or 2 kHz and digitally filtered at 1 or 0.5 kHz off-line. Leakage and capacitive currents were subtracted digitally by scaling traces evoked by steps from −20 to −50 mV.
All curve fitting (I h activation and deactivation time constants, activation curves, voltage decays, and various x–y plots) was performed with a least-squares program (IgorPro, WaveMetrics, Lake Oswego, OR). Activation curves are least-square fits of the data to a Boltzmann function. CsCl (Sigma, St. Louis, MO) was added to a solution identical to the standard cell-attached solution and was applied to the excised patch via a small-bore perfusion pipette. Error bars represent SEM, and the number of patches (n) is given in parentheses.
Whole-cell patch-clamp recordings were made by using two Dagan (Minneapolis, MN) BVC-700 amplifiers in active “bridge” mode. The external recording solution contained (in mm): 124 NaCl, 2.5 KCl, 1.2 NaH2PO4, 25 NaHCO3, 2.0 CaCl2, 1.5 MgCl2, 10 dextrose, and 0.005 DNQX, bubbled as above at ∼35°C, pH 7.4. Whole-cell recording pipettes (somatic, 2–4 MΩ; dendritic, 5–7 MΩ) were pulled from borosilicate glass. The internal solution was the outside-out patch recording solution with the addition of 0.5 mm EGTA. Series resistance for somatic recordings was 6–20 MΩ, whereas that for dendritic recordings was 15–40 MΩ. Dendritic pipettes were coated with Sylgard. Voltages have not been corrected for the theoretical liquid junction potentials (6–7 mV).
Hyperpolarization-activated inward current
Long-duration (1–3 sec) step hyperpolarizations evoked inward currents from cell-attached patches obtained from both the somatic and apical dendritic regions of hippocampal CA1 pyramidal neurons (Fig.1 A,B). These inward currents were slowly activating, noninactivating, and slowly deactivating. In general, patches were held 20 mV depolarized to the resting potential (soma: −67 ± 2 mV, n = 15; >100 μm dendrite: −70 ± 2 mV, n = 18), which was determined by patch rupture after the recording period. Currents began activating near −60 mV, and steady-state current amplitude increased in an approximately linear manner for the membrane potentials that were tested (up to −140 mV). On the other hand, tail current amplitude (recorded at approximately −50 mV) peaked after steps to approximately −100 mV (Fig. 1 C,D). These general features of hyperpolarization-activated currents were found in both somatic and dendritic recordings. For the majority of I hrecordings a high external K+ recording solution was used because the greatly increased current amplitude provided by such a solution (DiFrancesco, 1981; Mayer and Westbrook, 1983; Spain et al., 1987; Maruoke et al., 1994) allowed for an accurate comparison ofI h properties and distribution over a large part of the somatodendritic axis of the neurons.
I h was observed in all regions of CA1 pyramidal neurons from which recordings were obtained, including regions of the apical dendrite located in the most distal stratum radiatum and proximal stratum moleculare (Fig.2 A). Steady-state current amplitude at approximately −130 mV progressively increased with distance away from the soma (soma: 8.9 ± 1.6 pA,n = 21; 300–350 μm dendrite: 62.3 ± 8.5 pA,n = 14). The mean current could be converted to mean current density (per μm2) by normalizing to an ∼5 μm2 patch area (Magee and Johnston, 1995) and was 1.8 ± 0.3 pA/μm2 (n = 21) at the soma as compared with a density of 12.5 ± 1.7 pA/μm2 (n = 14) recorded from dendrites located 300–350 μm away from the soma (Fig.2 B, Tables 1,2). Maximum I h conductance (from the fully activated tail current recorded at approximately −50 mV, using a reversal potential of 0 mV) increased from 63.6 ± 12.2 pS (n = 6) at the soma to 429.7 ± 87.2 pS (n = 7) 300–350 μm out in the dendrites. For all of these measures this is an approximately sevenfold increase across the somatodendritic axis. Single I h channel activity was too small to be measured accurately with the noise levels (no less than 300 fA RMS, 5 kHz) present in these experiments.
I h amplitude was reduced significantly when external recording solutions contained 60 mmNa+/60 mm K+. In this solution the mean current amplitude was 22.6 ± 6.2 pA at −130 mV (all dendritic recordings were from ∼300 μm; n = 9), which is reduced to nearly one-third of the mean current amplitude recorded at the same dendritic location in 0 Na+external (54.6 ± 7.6 pA, 130 mV; n = 10). Current amplitude was reduced even further when the external recording solution contained more physiological K+ (2.5 mm) and Na+ (110 mm) concentrations (4.5 ± 1.1 pA, −130 mV; n = 3; ∼300 μm). Using a range of reversal potentials from −20 to −40 mV (Pape, 1996), we can calculate a conductance density estimate of 8–10 pS/μm2 for the distal dendritic regions. Using the density decrease reported in Figure 2 B, we can infer a 1–2 pS/μm2 for the somatic compartment. The dendritic I h conductance density would be ∼10-fold lower than the conductance density previously reported for Na+ channels located in the distal dendritic regions (Magee and Johnston, 1995). Na+ channel recordings made from the same patches as above (with 110 mmNa+) would support this estimate (Fig.2 C).
Voltage ranges of activation
Activation curves were generated by using the slowly decaying tail currents present on membrane repolarization (to approximately −50 mV) from the command potential (Fig.3 A,B). Tail current amplitude was measured at ∼5 msec after step repolarization, allowing for the complete decay of any residual capacitive current. Recordings obtained from dendritic membrane (>100 μm away from soma) presented a voltage range of activation (V 1/2 = −89.5 ± 1.0 mV; k = 8.5 ± 0.5; n = 13) that was shifted in a hyperpolarized direction as compared with that obtained from the more proximal regions of the neuron (V 1/2 = −81.9 ± 1.2 mV; k= 8.8 ± 0.5; n = 10) (<100 μm) (Fig.3 C). These voltage ranges determine that ∼10% of the maximal current is activated at rest in the soma, whereas somewhat less (∼5%) relative channel activation occurs at rest in the more distal dendrites for the 0 mm Na+ solutions. The hyperpolarized shift in the dendritic activation curve thus reduces the impact of the extremely elevated distal dendritic channel density on resting membrane properties.
The presence of Na+ in the external solution (10–60 mm) shifted the activation curve of dendritic channels nearly 10 mV in a depolarized direction (V 1/2 = −80.5 ± 1.3 mV; k = 6.7 ± 0.7;n = 6) (see Fig. 6 C). Proximal currents were too small in amplitude to generate accurate activation curves, but they too appeared to be shifted similarly. This indicates that under more physiological conditions ∼25% of hyperpolarization-activated channels would be open at rest in the soma, whereas nearly one-half this much would be activated in the distal regions of the neuron.
Activation and deactivation kinetics
For both dendritic and somatic recordings the rate of current activation was voltage-dependent such that the activation time constant decreased with increasing hyperpolarization (from ∼50 msec at −75 mV to ∼16 msec at −125 mV; Fig.4 A,C). These time constants were recorded with a bath temperature of 33°C. In five recordings the temperature of the bathing solution was reduced from 33 to 23°C, and activation time constants slowed from 17.8 ± 1.3 to 80.2 ± 10.7 msec. Thus, activation has aQ 10 of 4.5. There was very little differences observed between the activation kinetics of somatic versus dendritic currents, although somatic activation may have been slightly faster at the most negative potentials (Fig. 4 C).
Current deactivation was also voltage- and temperature-dependent. Tail current decay time constants decreased with increasing depolarization (30 msec at −70 mV to ∼7 msec at −30 mV; 33°C; Fig.4 B,C). Deactivation kinetics were also temperature-dependent, going from 14.4 ± 2.0 to 68.2 ± 9.8 msec when the bathing solution was reduced from 33 to 23°C (Q 10 = 4.7; n = 5). As with other channel types the temperature dependence of current amplitude was much lower (Q 10 = 1.95; n = 5).
Although single I h channel activity has been reported in sinoatrial (SA) node cells (DiFrancesco, 1986), there still remains some possibility that I h is not a classically voltage-gated channel and that I hperhaps may be mediated by some type of ion carrier. Although the noise levels in the present recordings did not allow for the observation of single channel activity (the reported single channel conductance is <1 pS), the temperature dependence exhibited by the currents is very similar to that observed for other voltage-gated ionic channels (Byerly et al., 1984; van Lunteren et al., 1993; Haverkampf et al., 1995;McAllister-Williams and Kelly, 1995; Milburn et al., 1995). The gating kinetics of voltage-gated ion channels generally showQ 10 at ∼5, whereas theQ 10 of current amplitude is generally <2. The difference in these values results from differences in the thermodynamics of channel gating, which is more similar to that of enzyme reactions, whereas channel permeation shows a lower temperature sensitivity as a result of being a more diffusion-like process (van Lunteren et al., 1993). I h, therefore, exhibits thermodynamic properties that are more similar to those of voltage-gated ion channels instead of some ion carrier for which both the kinetics and current amplitude would present similarQ 10.
The presence of Na+ in the external solution (10–60 mm) did not affect the activation kinetics but did slow channel deactivation ∼83% (τ = 26.3 ± 2.2 msec at −50 mV; 33°C; n = 5) (see Fig.5 A,B). TheQ 10 values given above were not affected by the change in external [Na+]. The ability of external Na+ to slow channel deactivation has been observed in SA node cells and was attributed to the presence of an external Na+ binding site capable of modulating channel gating (Maruoke et al., 1994). Perhaps a similar mechanism is present in I h here and also may be responsible for the depolarizing shift in the voltage range of activation observed in Na+ containing external solutions.
Reversal potentials determined from fully activated instantaneousI–V relationships were used to estimate the ionic selectivity of hyperpolarization-activated currents in CA1 pyramidal neurons (Fig. 6). With external solutions containing 120 mm K+, the instantaneousI–V relationship reversed at −1 ± 2.1 mV (n = 9; no difference was observed between somatic and dendritic recordings, so the data were pooled). With external solutions containing 60 mm Na+/60 mmK+, the reversal potential shifted 12 mV hyperpolarized (−13 ± 4; n = 8). From the mean reversal potentials a Na+/K+permeability ratio of ∼0.35 could be determined with the Goldman–Hodgkin–Katz equation. These data indicate thatI h has a mixed ionic conductance with significant permeabilities to both Na+ and K+. It should be noted that, becauseI h conductance has been reported to vary significantly from the independence principle (Wollmuth, 1995), the permeability ratios given here are only approximations.
The effect of external Na+ on current amplitude fits well with the idea that the I h channel is a multi-ion pore possessing high-affinity binding sites for both Na+ and K+. That the addition of a few millimolars concentration of Na+ to the external recording solution was capable of decreasing the current amplitude to a much greater extent than would be expected from changes in reversal potential can be explained by the anomalous mole fraction mechanism described for many other channel types (includingI h) (Hille, 1992; Wollmuth, 1995).
External Cs+ blockade
Inclusion of Cs+ in the high K+ external recording solution reduced the peak hyperpolarization-activated current amplitude in a concentration-dependent manner. Current kinetics were not affected by external Cs+. Cs+(0.3 mm) reduced the hyperpolarization-activated current amplitude to ∼0.5 that of control, whereas 5 mmCs+ produced a nearly complete channel blockade (Fig. 7).
Dual whole-cell recordings from the soma and distal dendrites of the same neuron were used to determine the effect of the differential distribution of I h channels on basic membrane properties and the subcellular propagation of potentials. As would be expected, hyperpolarizing current injections (300 msec duration) caused substantially reduced voltage deflections in the dendritic compartments (11 ± 1 mV peak; 8 ± 1 mV steady-state; n = 13; −200 pA) when compared with the somatic compartment (15 ± 2 mV peak; 13 ± 1 mV steady-state; n = 13; −200 pA) (Fig. 8 A,B). Input resistances (R in) were calculated as the slope of the current–voltage relationship for current injections ranging from −50 to +50 pA; a R in of 39 ± 2 MΩ was calculated for dendrites as compared with 66 ± 5 MΩ for somata (Fig. 8 E,F). Peak depolarizing voltage deflections in response to 300 msec, +100 pA current injections likewise were reduced in the dendritic compartments (4 ± 1 mV;n = 13) as compared with the somatic compartments (6 ± 2 mV; n = 13).
The blockade of ∼80% of I h by bath application of 3 mm Cs+ increasedR in in all regions of the neuron and greatly reduced the regional differences in R in(dendrite: 100 ± 3 mΩ, n = 13; soma: 112 ± 6 mΩ, n = 13) (Fig. 8 C,D). The increase in R in observed under I hblockade significantly increased the excitability of the neuron such that previously subthreshold current injections (100–200 pA, either dendritic or somatic) could now evoke repetitive action potential firing (Fig. 8 C). Application of external Cs+ also caused a slight hyperpolarization in resting membrane potential that was similar in both the proximal and distal regions of the neuron (dendrite: −4 ± 0.4 mV,n = 11; soma: −5 ± 0.7 mV, n = 10). In all cases a depolarizing current was injected to compensate for this change in membrane potential.
Subthreshold potential propagation
As a result of the regional R in differences the propagation of subthreshold potentials is directionally specific. In general, there is a greater amount of decrement in voltage signals as they propagate from the soma to the dendrites (decrease to 49 ± 5%, n = 6, 250–300 μm) (Fig.9 A–C, open diamonds) than from the dendrites to the soma (decrease to 79% ± 4%, n = 6, 250–300 μm). Also, the progressively lower R in of the dendritic compartment increasingly dampens voltage changes such that the amplitude of the dendritic voltage change in response to a dendritic current injection (termed the local dendritic signal) decreases with distance (see Figs. 8, 9 C, open squares). The rate of this decrease is very similar to the decrement of the voltage signal generated by somatic current injection as it propagates into the dendrites (termed the propagated dendritic signal; Fig.9 C, open diamonds). The result is that the ratio of the two signals (propagated-to-local) is maintained fairly constantly across the dendritic axis (92 ± 2%, n = 10; this value does not change with distance, so recordings ranging from 175 to 325 μm were averaged) (see Fig. 9 D, open circles). Thus, the final amplitude of a voltage signal generated by a current injection (300 msec) into the dendrite is almost identical to the final amplitude of a voltage signal generated by the same current injection into the soma. The opposite is true when recording from the soma, where the ratio of the somatic voltage change in response to dendritic versus somatic current injection (propagated-to-local signal) decreases as the dendritic injection site is moved distally (from 77 ± 6% at 200 μm, n = 4, to 46 ± 6%, n = 3, at 300 μm) (Fig.9 D, open triangles). External Cs+ (3 mm) reduces all regional differences in signal propagation, making both the somatic and dendritic propagated-to-local signal ratios decrease as the dendritic recording site is moved farther away from the soma (Fig. 9 B,D, filled circles andtriangles).
The effect is that for any given slow (>100 msec) current injection the voltage change occurring at a dendritic location will be the same regardless of whether the current was injected at the dendrite or at a more proximal region. Under conditions of I hblockade the amplitude of the local dendritic signal decreases very little (Fig. 9 C, filled squares) with distance so that now the filtering experienced by the propagated signal is greater than the decrease in the local signal (Fig. 9 C, filled diamonds). The result is a distance-dependent propagated-to-local signal ratio that eliminates this unique form of amplitude normalization (in that it’s proximal to distal) provided by the spatial gradient of dendriticI h (Fig. 9 D, filled circles).
Kinetics of voltage changes
The membrane charging and discharging rates for voltage deflections spanning a range from +15 to −20 mV were faster in the dendritic compartment when compared with the somatic. The decay of ∼10 mV, 50 msec membrane depolarizations could be well fit by single exponential functions (at least an initial period of the decay), and the mean time constant (τmem) recorded for the dendritic compartments was 15 ± 2 msec as compared with 28 ± 4 msec for somata of the same neurons (n = 6; Fig.10).
In the dendritic compartment a small (1–2 mV) hyperpolarization was observed after the 50 msec depolarizations, whereas no such potential was observed after 50 msec, 10 mV depolarizations in the soma (Fig.10). These hyperpolarizations were voltage- as well as time-dependent such that the amplitude increased with increasing depolarization and step duration. Significant hyperpolarizations (as well as the depolarizing counterpart; see Fig. 8) could be observed in the somatic compartment only after longer duration (300 msec) steps. The hyperpolarizations result from I h channel deactivation during depolarization and subsequent slow channel activation, whereas the depolarizations after negative current injection are, conversely, the result of the slowI h deactivation rate (Mayer and Westbrook, 1983;Spain et al., 1987; Schwindt et al., 1988; McCormick and Pape, 1990).
Application of external Cs+ (3 mm) increased τmem in both the somatic and dendritic compartments and eliminated the hyperpolarization normally observed after the 50 msec depolarizations. However, regional differences in τmem were still present under I hblockade, with the dendritic compartment remaining faster (dendrite, 27 ± 5 msec; soma, 42 ± 10 msec; n = 6) (Fig. 10). These data suggest that, although I hparticipates in setting both R in and τmem (at least in terms of 5–10 mV transients), other factors also are involved in setting a faster τmem in the dendritic compartment even in the absence of substantialI h. Such factors could be nonuniformities in membrane resistivity (R m) or in other ionic channel populations (Drake et al., 1997; Hoffman et al., 1997;Stuart and Spruston, 1998) Also, the incompleteI h blockade by 3 mmCs+ could account for part of the remaining regional differences (see also Fig. 8).
Because of the prominent role that I h plays in setting both R in and τmem, the impact of I h on the summation of synaptic activity was tested directly by using dendritic current injections. Under control conditions, exponential current injections (τrise = 2 msec; τdecay= 20 msec) into the dendritic compartment resulted in EPSP-shaped voltage transients, the amplitude and kinetics of which were filtered significantly as they propagated from the dendritic injection site to the somatic recording site (dendritic amplitude, 8 ± 0.5 mV; dendritic duration, 15 ± 0.8 msec; somatic amplitude, 3 ± 0.2 mV; somatic duration, 39 ± 2 msec) (Fig.11 A,D,E). Repetitive current injections also were given to mimic repetitive synaptic input, and these events were filtered similarly by the dendritic arborizations (dendritic amplitude, 24 ± 3 mV; dendritic duration, 136 ± 1 msec; somatic amplitude, 8 ± 1 mV; somatic duration, 154 ± 2 msec) (Fig. 11 B,D,E). As with step depolarizations, a slight hyperpolarization was generated (particularly in the dendrites) after EPSP depolarization. The amplitude of this hyperpolarization was larger after trains of input (<5 mV) than for single EPSPs (<2 mV).
I h channel blockade increased both single EPSP amplitude (soma, 10 ± 2%; dendrite, 7 ± 4%) and duration (soma, 38 ± 9%; dendrite, 10 ± 2%) as well as repetitive EPSPs amplitude (soma, 42 ± 8%; dendrite, 22 ± 5%) and duration (soma, 7 ± 2%; dendrite, 3 ± 1%) (Fig. 11). Although the peak amplitude of single EPSPs increased only slightly in Cs+, the increase was fairly uniform in both the soma and the dendrites such that the amount of amplitude attenuation occurring between the dendrites and soma was unaffected byI h blockade (control was 63 ± 3% vs 64 ± 4% in Cs+). On the other hand, the peak amplitude reached during EPSP trains increased twice as much in the soma as in the dendrite; thus, for repetitive activity the amount of attenuation occurring between the dendrites and soma was reduced by I h blockade (control was 66 ± 5% vs 58 ± 6% in Cs+).
Under conditions of I h blockade the increased amplitude of single somatic EPSPs is mostly the result of an elevation in effective R in, whereas the more pronounced effect on EPSP duration results from both the increasedR in and τmem. Because the amount of summation occurring during repetitive activity is dependent on both effective R in and τmem, the effect of Cs+ on the peak amplitude reached during an EPSP train is more pronounced than for single EPSPs. Presumably, the slight improvement in the propagation of EPSP trains and not single EPSPs during the blockade of I h reflects the involvement of both R in and τmem.
Action potential backpropagation
The backpropagation of action potentials was affected only slightly by I h blockade, with the most consistent effect being on the more hyperpolarized regions of the repolarization phase (Fig. 12). The amplitudes of single somatic and dendritic action potentials were not changed by I h blockade (soma: 101 ± 4 to 103 ± 5 mV, n = 6; dendrite: 42 ± 9 to 41 ± 9 mV, n = 6), whereas the duration (at half-maximum amplitude) of dendritic action potentials was prolonged considerably by external Cs+ (dendrite: 2.2 ± 0.5 to 4.4 ± 1.3 msec, n = 6). In the dendrites, single action potentials and particularly trains of multiple action potentials generally were followed by an afterhyperpolarization (AHP) (1–2 mV), a component of which was sensitive to external Cs+ (Fig. 12). This suggests thatI h deactivation occurring during action potentials significantly contributes to dendritic spike repolarization, particularly in the −40 to −65 mV range (Spain et al., 1987; Schwindt et al., 1988; Maccaferri et al., 1993). In a minority of cases (two of six neurons), I h blockade reduced the amount of amplitude attenuation occurring during a train of five dendritic action potentials, and in these two neurons the accompanying decrease in the action potential maximum rate of rise was reduced also. Thus the dampening effect of dendritic I h does appear to increase the susceptibility of the dendrites to frequency-dependent action potential backpropagation, perhaps by loweringR in and by providing a mechanism for dendritic AHP generation.
Hyperpolarization-activated currents were found throughout the entire extent of CA1 pyramidal neurons presently studied (soma to 350 μm of apical dendrite). These currents displayed a biophysical and pharmacological profile that is most characteristic of the current termed I h. Generally speaking, the ionic selectivity, voltage ranges of activation, and kinetics of activation and deactivation as well as the sensitivity to external Cs+ all fall within the ranges reported for a wide variety of central and peripheral neurons as well as cardiac cell types (DiFrancesco, 1981; Halliwell and Adams, 1982; Mayer and Westbrook, 1983; Spain et al., 1987; McCormick and Pape, 1990; Maccaferri et al., 1993). The presence of a significant dendriticI h density acts to lower theR in of the dendritic compartment as well as to speed the kinetics of membrane polarization. These alterations in basic dendritic membrane properties impact both the shape and propagation of sub- and suprathreshold electrical signals.
An increase in I h density was found across the somatodendritic axis such that the density of the most distal regions was nearly sevenfold larger than that in the somatic region. A similar nonuniform channel density has been reported for a transient K+ current in CA1 pyramidal neurons, whereas Na+ and composite Ca2+ current density remains relatively constant from soma to distal apical dendrite (Magee and Johnston, 1995; Hoffman et al., 1997). Very recently, the products of several cloned genes have been shown to produce hyperpolarization-activated currents that are characteristically similar to I h, and corresponding RNA has been found in abundance in the hippocampus (Santoro et al., 1997; Gauss et al., 1998; Ludwig et al., 1998; Santoro et al., 1998). Unfortunately, extensive channel labeling or binding studies are not yet available to corroborate the presence of these channels in the dendritic arbors of CA1 pyramidal neurons. However, the presence of a greatly elevated membrane sag during hyperpolarizations (a classic hallmark of I h) has been reported in several intradendritic recording studies from at least two different pyramidal cell types (Andreasen and Lambert, 1995; Tsubokawa and Ross, 1996; Stuart and Spruston, 1998). In fact, an elevated dendriticI h density was required for the realistic neuronal model of Stuart and Spruston (1998) to match whole-cell data recorded from the soma and dendrites of neocortical pyramidal neurons. The data presented here support the hypothesis that a nonuniform density of I h exists in cortical pyramidal neurons.
Even with the elevated dendritic I h densities, absolute I h density is quite small when compared with other dendritic channel densities, Na+ and K+ channels in particular. Despite being a relatively small conductance, the persistent nature of this current, as well as the voltage dependence of its activation range, determines thatI h current amplitudes are sufficient to impact profoundly most aspects of subthreshold dendritic membrane activity.
Activation voltage ranges
The observed hyperpolarized shift (∼10 mV) in the voltage range of activation for I h channels located in more distal regions is possibly the result of differential levels of channel modulation between the distal and proximal regions of the CA1 pyramidal neurons. The I h activation curve is reportedly very sensitive to intracellular adenylate cyclase and cAMP activities. Increases in cAMP (such as observed during β-receptor activation) lead to depolarizing shifts, whereas decreases (such as during muscarinic receptor activation) shift the curve in a more hyperpolarized direction (Ingram and Williams, 1996; Pape, 1996; Jafri and Weinreich, 1998). A similar hyperpolarized shift in the voltage range of activation also has been reported for a transient K+ current located in the distal dendrites of CA1 pyramidal neurons (Hoffman et al., 1997). These channels likewise are sensitive to cAMP levels (Hoffman and Johnston, 1998). Taken together, it seems probable that the shift in the I hactivation curve could be the result of a lower basal level of cAMP activity in the more distal regions of CA1 pyramidal dendrites. Because the shift in the voltage range of activation of transient K+ current ultimately appears to be regulated by phosphorylation by PKA, an increased phosphatase activity also could be involved in this modulation. However, such regulation by phosphorylation does not appear to be involved in the shifting ofI h activation ranges (Pedarzani and Storm, 1995); therefore, the simplest explanation is that a lower resting cAMP activity results in a hyperpolarized shift in theI h activation curve for channels located in the more distal regions of the dendritic arbor.
Functional impact of dendritic Ih
The major impact of the elevated dendritic density ofI h channels is a decreased apparentR in and a faster τmem in the dendritic compartment. As a result of these effects, membrane depolarizations and, to an even greater extent, hyperpolarizations are increasingly blunted the more distal one proceeds across the somatodendritic axis. Therefore, the absolute effectiveness of more distal synaptic input (i.e., the total charge transferred from synapse to soma) will be reduced by the increasingly largeI h conductance. A similar situation appears to be present in neocortical pyramidal neurons, whereI h has been shown to decrease the current transmitted from the dendrites to the soma for both EPSPs and more prolonged input (Schwindt and Crill, 1997; Stuart and Spruston, 1998). The lowered dendritic R in and τmemalso act to reduce the amount of summation that occurs during repetitive synaptic activation. Such an effect will reduce the time window over which temporal summation and coincidence detection can occur, allowing higher frequency input to be discriminated (Banks et al., 1993; Shadlen and Newsome, 1994; Softky, 1995; Hausser and Clark, 1997).
The spatial gradient of I h determines that the amplitude decrement exhibited by slower voltage signals as they propagate from the soma into the dendrites is strikingly similar to the distance-dependent decrease in the amplitude of local voltage signals. Because of this, a form of amplitude normalization occurs where, for example, slow inhibitory synaptic input into the proximal regions would have almost exactly the same impact on the distal dendritic compartments as would inhibitory input that was received locally in the distal dendrites themselves. The opposite effect will occur for the impact of slow input into the dendrites on the somatic compartment. In this case the spatial gradient of I h causes inhibition to the dendrites to have a much reduced impact on the soma as compared with that received by somatic inhibition. Thus, proximal slow polarization events (such as repetitive ∼5 Hz inhibitory input observed during theta activity) will impact the dendritic integration of synaptic input with almost exactly the same efficacy as distally arriving events.
Although the impact of I h on subthreshold voltage events is more obvious than on action potential propagation, the general membrane-shunting effect of I h and the generation of AHPs determine that I h does act, along with the transient K+ current, to reduce the excitability of dendritic membrane. In fact, withoutI h the more distal dendrites do not appear to contain any mechanism for generating AHPs. Although the primary regulator of dendritic excitability remains the large transient K+ current present in the distal dendrites, the elevated I h density does have an additional blunting effect, particularly where trains of action potentials are concerned. In fact, the muscarinic modulation of spike backpropagation observed in hippocampal CA1 neurons may be attributable in part to the inhibitory effects of muscarinic receptor activation onI h (Tsubokawa and Ross, 1997).
As discussed above, the voltage range of I hactivation is extremely modulatable (Pape, 1996). The hyperpolarized voltage range of activation observed in the more distal dendrites under basal slice conditions lowers the impact of these channels, whereas the elevated density provides the potential for an even greater impact on all of the properties discussed above. Thus, the influence of the elevated dendritic I h density on the resting membrane potential, the integration of synaptic input, and action potential propagation are all highly dependent on the modulatory state of the I h channels.
Overall, I h acts to dampen dendritic excitability in much the same way as the transient K+ current found in high density in CA1 dendrites. However, in contrast to dendritic K+ currentsI h has its largest impact on the subthreshold range of membrane potentials where the integration of inhibitory and excitatory synaptic inputs takes place. Because of the prominent role that I h plays in the determination of basic membrane properties and the susceptibility of I hto modulation, the manner in which synaptic potentials are integrated within the dendritic arborization can be altered easily and effectively during different physiological conditions.
This work was supported by National Institutes of Health Grant NS35865. I thank Michael Carruth for technical assistance.
Correspondence should be addressed to Dr. Jeffrey C. Magee, Neuroscience Center, Louisiana State University Medical Center, 2020 Gravier Street, New Orleans, LA 70112. E-mail: