Cell, Vol. 119, 719–732, November 24, 2004, Copyright 2004 by Cell Press
A Behavioral Role for Dendritic Integration: HCN1
Channels Constrain Spatial Memory and Plasticity at
Inputs to Distal Dendrites of CA1 Pyramidal Neurons
to regulate the learning of new information by gating
the initiation of somatic or dendritic spikes that serve
as associative signals for the induction of synaptic plas-
ticity (Bi and Poo, 1998; Golding et al., 2002; Magee
and Johnston, 1997; Markram et al., 1997). At present,
however, little is known about how these active integ-
rative properties may control learning and memory.
We have focused our attention on the integrative role
of the HCN1 channel subunit, one of four HCN (hyper-
polarization-activated, cyclic-nucleotide gated, cation
nonselective) subunits that generate hyperpolarization-
activated inward currents (Ih) (Robinson and Siegel-
baum, 2003). HCN1 is highly expressed in cerebellar
cortical neurons and in pyramidal cells of the neocortex
and hippocampus (Lorincz et al., 2002; Notomi and Shi-
that knockout of the HCN1 gene causes profound defi-
cits in motor learning that may be accounted for by
actions of HCN1 in cerebellar Purkinje cells (Nolan et al.,
2003). In these spontaneously spiking neurons, HCN1 is
activated by inputs that hyperpolarize the membrane
below the threshold for spontaneous spiking. By stabi-
lizing the integrative properties of Purkinje cells, HCN1
enables reliable encoding of information independently
of their previous history of activity (Nolan et al., 2003).
Here we ask, is HCN1 generally required in neuronal
circuits important for learning and memory? Does it al-
Or does its function vary, depending on its cellular
To approach these questions, we have studied the
roleof HCN1inforebrain pyramidalneurons. UnlikePur-
kinje cells, these neurons are generally not spontane-
ously active in the absence of synaptic input. Pyramidal
neurons also have resting potentials substantially more
icant fraction of Ihin pyramidal cells is activated at rest
and therefore can contribute to the integration of post-
integration in pyramidal neurons is underscored by its
distinctive dendritic gradient of expression, with levels
in apical dendrites increasing steeply as a function of
distance from the soma (Lorincz et al., 2002). The re-
sulting gradient of Ihdensity is thought to reduce the
amplitude and duration of distal synaptic potentials and
to remove the location dependence of synaptic summa-
tion that would otherwise arise as a result of the passive
Magee, 1998, 1999; Williams and Stuart, 2000, 2003).
The dendritic gradient of Ihis of further interest be-
cause different regions of the apical dendritic tree re-
dritic regions of pyramidal cells in CA1 contain the
highest density of HCN1 channels and are innervated
by direct perforant path (temporoammonic) inputs from
layer III of the entorhinal cortex. These inputs may be
of particular functional importance for representation of
spatial information by CA1 neurons (Brun et al., 2002).
The more proximal regions of the apical dendrite, where
the HCN1 density is lower, are innervated by Schaffer
Matthew F. Nolan,1Gae ¨l Malleret,1
Josh T. Dudman,1Derek L. Buhl,2Bina Santoro,1
Emma Gibbs,1Svetlana Vronskaya,1
Gyo ¨rgy Buzsa ´ki,2Steven A. Siegelbaum,1,3,5,6
Eric R. Kandel,1,4,5,6,* and Alexei Morozov1,7
1Center for Neurobiology and Behavior
New York, New York 10032
2Center for Molecular and Behavioral Neuroscience
Newark, New Jersey 07102
3Department of Pharmacology
4Departments of Physiology, Biochemistry and
Biophysics, and Psychiatry
5Howard Hughes Medical Institute
6Kavli Institute for Brain Sciences
New York, New York 10032
7Unit of Behavioral Genetics
National Institute of Mental Health
Bethesda, Maryland 20892
The importance of long-term synaptic plasticity as a
cellular substrate for learning and memory is well es-
tablished. By contrast, little is known about how learn-
ing and memory are regulated by voltage-gated ion
channels that integrate synaptic information. We in-
vestigated this question using mice with general or
forebrain-restricted knockout of the HCN1 gene,
which we find encodes a major component of the hy-
perpolarization-activated inward current (Ih) and is an
important determinant of dendritic integration in hip-
pocampal CA1 pyramidal cells. Deletion of HCN1 from
forebrain neurons enhances hippocampal-dependent
learning and memory, augments the power of theta
oscillations, and enhances long-term potentiation (LTP)
of CA1 pyramidal neurons, but has little effect on LTP
at the more proximal Schaffer collateral inputs. We
suggest that HCN1 channels constrain learning and
memory by regulating dendritic integration of distal
synaptic inputs to pyramidal cells.
Considerable progress has been made toward identi-
fying the molecular mechanisms mediating long-term,
important for memory storage (Bliss and Collingridge,
1993; Bliss et al., 2003; Milner et al., 1998). However, to
alter neuronal output these synaptic changes must be
cell. Voltage-gated channels active at subthreshold po-
tentials are ideally situated to influence this readout and
Consistent with this lower channel density, pharmaco-
logical blockade of Ihdoes not alter long-term potentia-
tion (LTP) or long-term depression (LTD) at the Schaffer
collateral synapses (Gasparini and DiFrancesco, 1997;
Wang et al., 2003). Given the high density of HCN1 at
the distal perforant path inputs, we were interested in
determining whether this channel may selectively regu-
late synaptic transmission and plasticity at these direct
cortical inputs (Colbert and Levy, 1993; Remondes and
Using mice with general (HCN1?/?) and forebrain-
restricted (HCN1f/f,cre) knockout of the HCN1 gene, we
examined the influence of HCN1 channels on CA1 pyra-
campal network activity, and hippocampal-dependent
forms of learning and memory. We provide evidence
to pyramidal cells with no change in LTP at Schaffer
collateral inputs, suppress theta frequency network ac-
tivity, and constrain spatial learning and memory. Modi-
fied dendritic integration resulting in enhanced LTP at
perforant path synapses provides a possible mecha-
nism for enhanced learning by HCN1f/f,cremice.
water maze experiment was comparable to control
(HCN1f/f) mice (Figure 1A).
HCN1 Constrains Acquisition of a Spatial
We next asked if deletion of HCN1 from forebrain neu-
rons alters the performance of tasks requiring spatial
in a new location that could only be determined by the
mouse using spatial cues placed around the room and
which remained the same during subsequent trials. A
priming procedure, during which mice were placed on
the platform at the new location for 15 s, was carried
out 15 min prior to the first spatial learning trial. Mice
were then trained with four trials per day over 4 days.
Comparison of the performance of the mice on each
day indicated that HCN1f/f,cremice learned the spatial
task faster than control mice, although both groups of
mice reached a similar final level of performance and
performed similarly when their memory of the platform
location was tested with a probe trial indicating that
they both use a spatial strategy to solve the task (Figure
1B). Comparison of swimming speed, time spent float-
ing, or thigmotaxis did not reveal any effect of forebrain
deletion of HCN1 during training on days 3–6, indicating
that motor or motivational effects do not contribute to
the differences in learning (Supplemental Figure S2A on
the Cell website).
The short-term effects of information remembered
during the 15 min between trials can be distinguished
by comparing the performance of control and HCN1f/f,cre
mice across individual trials, rather than days (Figure
1C). During the four training trials of the first spatial
training day (day 3), the path length for HCN1f/f,cremice
to reach the platform was consistently shorter than for
during the first trial. However, the largest difference be-
tween the two groups of mice was on trial 1 of day 4,
during which the performance of HCN1f/f,cremice was
significantly enhanced compared with HCN1f/fmice (p ?
0.01). This is because the performance of HCN1f/fmice
returned toward the pre-training level on the first trial
that the memory of the platform location used to com-
by the HCN1f/fmice. By contrast, HCN1f/f,cremice main-
tained a similar level of performance between these two
trials, indicating that they retained the memory of the
platform location from the previous day.
HCN1 Constrains Performance of Long-Term
Components of a Water Maze Spatial Memory Task
when Training Frequency Is Reduced
To determine if HCN1 channels also influence longer-
term memory processes that operate cumulatively over
group of mice on a more demanding task in which they
were trained to find a hidden platform with only one trial
per day (Figure 1D). The platform was placed in a new
location and approximately 15 min before the first trial
the priming procedure was again carried out. The path
length for HCN1f/f,cremice to reach the platform was re-
duced compared with HCN1f/fmice on both the first
day and on days 5–8, indicating that deletion of HCN1
Deletion of HCN1 Channels
from Forebrain Neurons
We examined two HCN1 mutant mouse lines (Nolan et
al., 2003): HCN1?/?mice in which HCN1 is deleted from
the entire mouse and HCN1f/f,cremice in which deletion
of HCN1 is limited to the forebrain. Wild-type (HCN1?/?)
and floxed (HCN1f/f) mice, used as littermate controls
for experiments with HCN1?/?and HCN1f/f,cremice, re-
spectively, expressed HCN1 mRNA and protein with
similar patterns to that described previously for wild-
type mice (Lorincz et al., 2002; Santoro et al., 1997,
2000). In the forebrain, high expression of HCN1 mRNA
was found in layer V of the neocortex and in the pyrami-
labeling of the protein occurring in regions containing
the distal apical dendrites of pyramidal cells. By con-
trast, HCN1 mRNA and protein were absent from the
brains of HCN1?/?mice and greatly reduced from the
forebrain of HCN1f/f,cremice (Nolan et al., 2003; Supple-
mental Figure S1 at http://www.cell.com/cgi/content/
Contribution of Forebrain HCN1 Channels
to Behaviors Involving Learning and Memory
We previously found that HCN1?/?mice have a profound
deficit in motor learning. Thus, on the first 2 days of a
water maze experiment in which mice learn to navigate
to a submerged platform, whose location is indicated
by a flag, these mice do not learn to swim directly to
the platform but instead maintain a tendency to swim
in loops (Nolan et al., 2003). By contrast, HCN1f/f,cremice
do not appear to have any deficits in motor learning and
their performance in this visible platform phase of the
HCN1 Channels Constrain Learning and Memory
Figure 1. Short- and Long-Term Memory Is Enhanced by Deletion of HCN1 from Forebrain Neurons
The path length for mice to reach the platform in a water maze is plotted against the day of the experiment (D) or the trial number (T). The
insets show the percentage of time spent in each quadrant of the pool during the corresponding probe trial.
(A) Data from the visible platform, nonspatial version of the water maze task (effect of genotype p ? 0.05).
(B) The path length plotted against the day of the experiment for a four trials a day, hidden platform, spatial version of the water maze task.
ANOVA indicated a significant difference between HCN1f/f,cre(n ? 13) compared with HCN1f/fmice (n ? 12) in path length and latency [F(1,24) ?
9.48, p ? 0.005 for path length; F(1,24) ? 7.05, p ? 0.01 for latency]. Probe trial performance on day 7 was similar in both groups of mice
(p ? 0.29).
(C) The data from the same experiment as in (B) plotted as a function of the trial number on each day.
(D) Data from a one trial per day, hidden platform version of the water maze task. The path length for HCN1f/f,cremice to reach the platform
on the first session was reduced compared with HCN1f/fmice (p ? 0.04). The path length (genotype effect p ? 0.03) and latency (genotype
effect p ? 0.04) for HCN1f/f,cremice to reach the platform was also reduced during the second week of testing.
(E) Data from a different group of mice with a four trials a day, hidden platform version of the water maze task, without priming prior to the
first trial on day 1. There was no significant difference between HCN1f/fmice (n ? 13) and HCN1f/f,cremice (n ? 14) (ANOVA, latency p ? 0.21,
path p ? 0.19).
(F) Data from the same group of mice as in (E) tested with one trial a day hidden platform water maze task, without priming. Forebrain deletion
of HCN1 enhanced performance of the mice during the second week of the experiment (genotype effect of ANOVA for 2nd week, path length
p ? 0.031, latency p ? 0.028).
enhances performance of short- and long-term compo-
nents of the task.
The Enhancement of Short-Term, but not Long-
Term, Memory Requires that the Mice Be
Primed Prior to the First Trial
We next asked if the modulation by HCN1 of short- and
long-term components of learning could be dissociated
from one another. We repeated the water maze experi-
ments in a second group of animals, but without the
was no significant difference between the performance
of control and HCN1f/f,crelittermates when trained in the
water maze task with four trials per day (Figure 1E).
When the same mice were then trained to find the plat-
form in a new location with one trial per day, again
without the priming procedure, the early enhancement
of learning by the HCN1f/f,cremice was again absent (Fig-
ure 1F). However, the performance of HCN1f/f,cremice
was enhanced on days 5 to 7 of the experiment (Figure
1F). A probe trial after day 8 did not reveal a significant
difference between the HCN1f/fand HCN1f/f,cremice, indi-
cating that both groups of mice use spatial strategies
to solve the task.
These data suggest that forebrain HCN1 channels
modulate both short- and long-term memory. The re-
duced path length to reach the hidden platform during
the initial trial following priming suggests that HCN1f/f,cre
during the priming procedure. The priming-indepen-
dent, reduced path length to the hidden platform during
later trials of the more difficult one trial per day experi-
ment suggests that HCN1f/f,cremice also have enhanced
Anxiety, Attention, and Fear Conditioning
Are Not Altered by Deletion of HCN1
from Forebrain Neurons
Can the enhanced spatial learning shown by the
HCN1f/f,cremice be accounted for by changes in anxiety
or attention? There was no significant difference be-
tween the behavior of HCN1f/f,creand control mice in an
elevated plus maze indicating that the two groups of
mice do not differ in levels of anxiety. Nor was there a
difference in pre-pulse inhibition, a simple test of sen-
sory-motor gating, suggesting that basal attention is
similar in the two groups of mice (Supplemental Figures
S2B and S2C).
Does the spatial learning enhancement extend to
ence between HCN1f/f,creand control mice in cued fear
conditioning, an amygdala-dependent form of memory,
or in a contextual-fear conditioning task, which is
thought to involve a ventral hippocampal-dependent
form of memory (Supplemental Figures S2D and S2E).
Thus, the enhancement of learning and memory in
HCN1f/f,cremice does not appear to extend to contextual
fear conditioning or to involve changes in anxiety or at-
ever, the peak to trough area during the falling phase
of the theta wave was significantly different between
HCN1?/?and HCN1?/?animals during both wheel run-
ning (p ? 0.01) and REM sleep (p ? 0.01) (Figures 2C
and 2D), indicating that the deletion of HCN1 attenuates
theta wave asymmetry. Thus, HCN1 channels are not
obligatory for the generation of hippocampal network
rhythms in vivo, but they may affect the proper physio-
logical expression of theta frequency activity.
Role of HCN1 in the Physiological Properties
of CA1 Pyramidal Cells
Since CA1pyramidal neurons representspatial informa-
tion, are important for learning and memory, and
strongly express HCN1 in their distal dendrites, we fo-
cused our cellular analysis on the contribution of HCN1
to the integrative properties of these neurons.
Rapid and Full Activation of Pyramidal Cell Ih
Requires HCN1 Channels
with cloned HCN channels suggests that the current is
subunits (Chen et al., 2001; Ulens and Tytgat, 2001).
We investigated the contribution of HCN1 to Ihisolated
pharmacologically at room temperature. Qualitatively
similar results were obtained with standard recording
S4B). Hyperpolarizing voltage steps activated Ihin CA1
mately two thirds. As knockout of HCN2 reduces the
amplitude of Ihby approximately one third (Ludwig et
al., 2003), the sum of the residual Ihfrom HCN1 and
HCN2 knockout mice can account for all of the wild-
type current. Deletion of HCN1 also dramatically slowed
the activation and deactivation kinetics of Ih. The activa-
tion timecourse of Ihfrom HCN1?/?or HCN1f/fmice re-
quired a sum of two exponentials for an adequate fit,
well fit with a single exponential function (Figures 3E–
3G). The slower kinetics of the residual Ihin mice with
deletion of HCN1 are similar to those of cloned HCN2
channels, whereas the kinetics of Ihin HCN2 knockout
mice resemble those of cloned HCN1 channels (Ludwig
et al., 2003). Thus, HCN1 and HCN2 appear to be the
major determinants of Ihin CA1 pyramidal cells.
HCN1 Channels Contribute to the Resting
Membrane Properties of CA1 Pyramidal Cells
We investigated the contribution of HCN1 to the physio-
logical properties of CA1 pyramidal cells using current
4, Supplemental Figures S4C and S4D, and Table 1).
did not fire spontaneous action potentials. Deletion of
HCN1 shifted the resting potential to more negative val-
ues, increased the input resistance, and prolonged the
membrane time constant. Activation of Ihin response to
negative current steps can drive a depolarizing mem-
brane potential “sag” (Figures 4C and 4D). The ampli-
polarization to the steady-state hyperpolarization, was
not altered by knockout of HCN1. However, when the
sag was quantified as the ratio of the peak hyperpolar-
Effect of HCN1 Deletion on Hippocampal Network
Activity: Enhancement of Theta Band
Network Activity In Vivo
What is the neuronal basis for the enhancement of spa-
tial learning? We focused our electrophysiological anal-
ysis on the CA1 region of the hippocampus, based on
its importance for spatial learning and memory and the
strong expression of HCN1 in this region. Since Ihhas
been suggested to influence network oscillations that
and McBain, 1996), we asked if deletion of HCN1 modi-
fies hippocampal network oscillations. Delta, theta,
gamma, and fast (“ripple”) oscillatory field potentials
were recorded in vivo from the CA1 pyramidal layer.
Spectral analysis of the spontaneous field potentials
differences between HCN1?/?and HCN1?/?mice. Both
low-frequency power and fast “ripple” power were re-
markably similar (see Supplemental Figure S3). How-
ever, power in the theta frequency band (4–9 Hz), re-
flecting oscillations thought to be important in encoding
and storing spatial information, was selectively en-
hanced in HCN1?/?animals during both wheel running
(p ? 0.01) and REM sleep (p ? 0.001)(Figures 2A and
2B), while frequencies in the gamma band (30–80 Hz)
were not affected.
HCN1 channel activation also affects the shape of
theta waves when analyzed in the time domain. No sig-
nificant difference was found in the area from the trough
to the peak of the rising phase of the theta wave. How-
HCN1 Channels Constrain Learning and Memory
Figure 2. HCN1 Is Not Required for Hippocampal Network Rhythms but May Selectively Suppress Theta Frequency Activity
(A and B) Power spectra of hippocampal local field activity from the CA1 pyramidal layer. Group data (mean ? SEM) during wheel running
and REM sleep. Note significantly larger theta power peaks (p ? 0.01) during both behaviors.
(C and D) Peak-triggered theta wave averages (?SEM) during wheel running and REM sleep. The area between peak and trough (?0–90 ms)
was significantly different between HCN1?/?and HCN1?/?mice (p ? 0.01) during both wheel running and REM sleep.
ization to the mean hyperpolarization between 400 and
significantly, reflecting the slower kinetics of the resid-
Responses to Low-Frequency Inputs Are
Preferentially Attenuated by HCN1 Channels
Previous studies have suggested that Ihcontributes to
a resonant peak in the voltage response of pyramidal
neurons to oscillatory current inputs at around theta
frequencies (Hu et al., 2002; Pike et al., 2000). Since this
appears at odds with the enhanced theta power seen
upon deletion of HCN1, we determined the response of
CA1 pyramidal cells to injection of oscillating currents
of linearly increasing frequency(Strohmann et al., 1994).
In CA1 pyramidal cells from HCN1f/fmice, the amplitude
of the membrane potential oscillation appeared con-
stant or increased slightly up to frequencies of approxi-
mately 4–8 Hz and then decayed steeply (Figure 5A).
By contrast, the amplitude of the membrane potential
response of CA1 pyramidal cells from HCN1f/f,cremice
was larger than that of HCN1f/fmice at low frequencies
but not at higher frequencies (Figure 5B).
The relationship between membrane impedance and
input frequency, obtained from the voltage responses
channels preferentially attenuate responses to low-fre-
quency inputs (Figure 5C and Supplemental Figure S5),
indicating that Ihcontributes to resonance by reducing
membrane impedance at frequencies below the theta
range. Similar effects of HCN1 on the frequency-imped-
ance relationships of CA1 pyramidal cells were found
in a comparison of HCN1?/?and HCN1?/?mice (data
not shown). Thus, deletion of HCN1 causes a general
enhancement in the voltage response to low-frequency
oscillatory currents, consistent with the enhancement
in theta power.
Properties of Schaffer Collateral and Perforant
Path Inputs to CA1 Pyramidal Cells
Given the important role of HCN1 in determining the
subthreshold properties of CA1 neurons, we next exam-
ined the effect of HCN1 deletion on synaptic responses
to activation of the two major excitatory inputs to CA1
neurons: the direct cortical input via the perforant path
and the hippocampal input from the Schaffer collateral
pathway (Figure 6). For Schaffer collateral inputs, the
relationship between somatic excitatory postsynaptic
for HCN1f/f,crecompared with HCN1f/fmice, indicating
that HCN1 has little influence on the amplitude of the
postsynaptic response to a single presynaptic stimulus.
Figure 3. HCN1 Is Required for Rapid and Full Activation of Ihin CA1 Pyramidal Cells
(A) Examples of membrane currents (bottom) from HCN1?/?(left) and HCN1?/?(right) mice, evoked by hyperpolarizing voltage steps (top).
Voltage steps have duration 5 s and are from a holding potential of ?50 mV to potentials down to ?115 mV in 5 mV increments.
(B) Mean amplitude of Ihtail currents plotted against test potential. Ihfrom HCN1?/?mice is reduced to approximately 40% of that in HCN1?/?
mice (tail currents following steps to ?115 mV: HCN1?/??144.7 ? 18.4 pA, n ? 4; HCN1?/??59.1 ? 15.4 pA, n ? 5; p ? 0.009).
(C and D) Membrane currents evoked by hyperpolarizing voltage steps as in (A) and mean tail currents as in (B), except data are from HCN1f/f
(n ? 6) and HCN1f/f,cremice (n ? 6). Deletion of HCN1 reduces Ihto approximately 30% of its control amplitude (tail current amplitudes following
steps to ?115 mV: HCN1f/f?126.6 ? 7.1; HCN1f/f,cre?40.6 ? 3.9; p ? 0.003). In (B) and (D) the shallow slope of the I-V relationship is probably
due to poor space clamp of currents originating from distal dendritic regions.
(E and F) Activation of current responses to steps from ?50 mV to ?110 mV (bottom traces), from HCN1f/fand (E) and HCN1f/f,cremice (F). The
residuals obtained from fitting the current activation with a single exponential (top) or with the sum of two exponentials (middle) are also shown.
(G) Mean time constants obtained from fitting the activation of Ihwith a single exponential (top) or the sum of two exponentials (bottom) are
plotted against the test membrane potential. Deletion of HCN1 caused an approximately 2.5-fold slowing of the activation time constant (tact)
obtained by fitting with a single exponential (activation time constants at ?115 mV: HCN1f/f282.3 ? 10.5 ms; HCN1f/f,cre725.4 ? 9.7 ms; p ?
2.6e-05). The time constants for deactivation (tdeact) of Ihwere also slowed by deletion of HCN1 (for steps from ?115 mV to ?50 mV, tdeact ?
138.3 ? 5.8 ms for HCN1f/fversus 319.5 ? 9.0 ms for HCN1f/f,cre, p ? 0.003).
However, the area of subthreshold Schaffer collateral
EPSPs was increased in pyramidal cells from HCN1f/f,cre
mice, indicating that HCN1 channels do modulate the
decay of Schaffer collateral EPSPs. By contrast, both
the amplitude and area of somatically recorded perfor-
rise times and decay kinetics in cells from HCN1f/f,cre
compared with HCN1f/fmice. This difference in EPSP
kinetics was larger for perforant path compared to
Schaffer collateral EPSPs, suggesting that the higher
expression of HCN1 channels in the distal dendrites of
CA1 pyramidal cells enables them to exert a greater
influence on integration of distal synaptic inputs. We
HCN1 Channels Constrain Learning and Memory
Figure 4. HCN1 Influences the Subthreshold
Membrane Properties of CA1 Pyramidal Cells
(A and B) Examples of membrane potential
responses (top) of CA1 pyramidal cells from
a HCN1f/f(A) and a HCN1f/f,cremouse (B) to
a series of current steps (middle). The peak
hyperpolarization (open symbols) and steady-
state membrane potential (closed symbols)
are plotted as a function of the current step
amplitude (bottom). Dotted lines indicate the
resting membrane potential.
(C and D) Comparison of the onset (C) and
offset (D) of similar amplitude responses to
currents steps demonstrates the faster mem-
brane time constant of pyramidal cells from
HCN1f/fcompared with HCN1f/f,cremice. Note
the membrane potential “sag” following the
peak hyperpolarization of the HCN1f/fre-
sponse. Records are from the data shown in
(A) and (B).
did not find any difference between the two groups of
mice in the relationship between stimulus intensity and
the slope of extracellular field EPSPs in either pathway,
indicating that deletion of HCN1 does not alter basal
transmitter release or the excitatory postsynaptic cur-
rent (Figures 6F–6G and Supplemental Figure S6).
HCN1 Channels Preferentially Attenuate LTP
at Distal Perforant Path Synapses and Not
Proximal Schaffer Collateral Synapses
Could changes in dendritic integration caused by dele-
tion of HCN1 alter the induction of synaptic plasticity?
We compared the effect of HCN1 deletion on LTP of
Schaffer collateral and perforant path inputs to CA1
(Figure 7). We found no difference between HCN1f/fand
HCN1f/f,cremice in the magnitude of LTP in the Schaffer
collateral pathway induced by a theta-burst protocol or
with a number of other standard LTP protocols and
recording conditions (Supplemental Figure S6). The ab-
inputs is also consistent with previous pharmacological
studies (Wang et al., 2003). In contrast to the Schaffer
collateral pathway, the amplitude of LTP at synapses in
the perforant path was dramatically enhanced in slices
from HCN1f/f,cremice compared with HCN1f/fmice. Thus,
the high density of HCN1 channels found in the distal
dendrites of CA1 pyramidal cells, which receive the per-
forant path synaptic inputs, appears to act as an inhibi-
tory constraint on the induction of LTP in this pathway.
in the proximal apical dendrites and do not appear to
influence LTP at Schaffer collateral inputs, which syn-
apse onto this part of the apical dendritic tree.
Our behavioral and electrophysiological experiments,
using mice with general and forebrain-restricted dele-
tion of HCN1 channels, provide evidence that dendritic
integration by forebrain pyramidal neurons can con-
strain both learning and memory as well as synaptic
plasticity. We suggest that HCN1 channels in the distal
dendrites of pyramidal cells could act as inhibitory con-
Table 1. Comparison of the Membrane Properties of CA1 Pyramidal Cells from HCN1f/fand HCN1f/f,creMice
Mean ? SEMn Mean ? SEMnp
IR ?ve (M?)
IR ?ve (M?)
?68.0 ? 0.9
100.0 ? 7.5
116.0 ? 9.1
22.0 ? 1.65
0.88 ? 0.01
0.88 ? 0.01
?73.6 ? 0.7
138.0 ? 12.6
161.5 ? 13.6
33.2 ? 3.6
0.86 ? 0.01
0.92 ? 0.01
Resting membrane potential (Em), input resistance measured from responses to negative current steps (IR ?ve), input resistance measured
from responses to positive current steps (IR ?ve), estimated membrane time constant (Tm), and the size of the membrane potential sag
during a negative current step (sag ratio), measured either at steady-state (ss) or from the mean membrane potential between 400 and 500
ms after the onset of the current step (early).
Figure 5. HCN1 Preferentially Attenuates Responses of CA1 Pyramidal Cells to Low-Frequency Inputs
(A and B) The frequency-response properties of CA1 pyramidal cells from HCN1f/f(A) and HCN1f/f,cremice (B) characterized by analysis of
membrane potential responses to oscillating current inputs.
(A1 and B1) Examples of membrane potential responses (top) to oscillating current inputs (bottom).
(A2 and B2) Plot of membrane impedance as a function of input frequency calculated from the data in (A1) and (B1).
(C) Mean impedance magnitude (Z) and ratio of impedance magnitude in HCN1f/fto that in HCN1f/f,cremice (Z ratio) plotted as a function of
input frequency. Dashed lines indicate standard error of the mean. At frequencies ?1.6 Hz the impedance of pyramidal cells from HCN1f/f,cre
mice (n ? 10) is significantly (p ? 0.05) higher than HCN1f/fcontrols (n ? 10). The preferential attenuation of low-frequency inputs by HCN1
channels is illustrated by the much smaller z ratio for input frequencies less than 2 Hz compared with input frequencies above 5 Hz.
straints on learning by damping postsynaptic changes
in membrane potential that at distal dendrites might
otherwise trigger synaptic plasticity.
provide clear evidence that the same channel can have
distinct functional roles in different forms of learning
depending on the cellular context and neuronal circuitry
in which the channel participates.
HCN1 Channels Constrain Short-
and Long-Term Memory
Several previous studies have found an enhancement
in performance of learning and memory tasks using ge-
netic modifications that either downregulate intracellu-
lar signalingpathways that constrainsynaptic plasticity,
overexpress molecules involved in the expression of
synaptic plasticity,or modify the propertiesof receptors
involved in the induction of synaptic plasticity (Abel et
al., 1998; Malleret et al., 2001; Routtenberg et al., 2000;
Tang et al., 1999). Other genetic modifications that in-
performance, indicating that the relationship between
the two phenomena is complex and unpredictable
(Sanes and Lichtman, 1999). So far genetic deletion of
other neuronal voltage-gated ion channels has been
et al., 2003). By contrast, the performance of HCN1f/f,cre
mice was enhanced in specific components of behav-
ioral tasksinvolving spatial referencememory. Together
with the finding that HCN1?/?mice have a profound
motor learning deficit (Nolan et al., 2003), these data
HCN1 Channels Constrain LTP at Perforant Path
Synapses by Influencing the Integrative
Properties of Pyramidal Neurons
rative properties of CA1 pyramidal cells. They act as a
temporal filter that preferentially attenuates low-fre-
filter that preferentially dampens distal inputs. Thus,
HCN1 channels open at the resting membrane potential
shunt synaptic inputs, and voltage-dependent gating of
HCN1 channels further opposes membrane responses.
The greater effect of HCN1 deletion on perforant path
compared with Schaffer collateral inputs can be ex-
to distal dendrites (Lorincz et al., 2002), supporting a
preferential role for the channel in integration of distal
dendritic inputs (Magee, 1998). As pyramidal cells in
the subiculum and in layer V of the neocortex have a
dendritic distribution of Ihand HCN1 subunits similar
to CA1 pyramidal neurons, HCN1 channels may have
similar functions in these neurons (Berger et al., 2001;
Lorincz et al., 2002; Williams and Stuart, 2000).
HCN1 Channels Constrain Learning and Memory
Figure 6. Effects of HCN1 Knockout on Schaffer Collateral and Perforant Path EPSPs in CA1 Pyramidal Cells
(A and B) Schematic illustration of the recording configuration and examples of somatically recorded EPSPs evoked by stimulation of Schaffer
collateral inputs (A) and perforant path inputs (B). Subthreshold Schaffer collateral EPSPs from HCN1f/fmice (n ? 8) and HCN1f/f,cremice (n ?
10) had 10%–90% rise times of 6.3 ? 0.9 ms and 6.7 ? 1.6 ms, respectively (p ? 0.84), decay times of 38.6 ? 7.0 ms and 51.3 ? 4.8 ms (p ?
0.14), and areas of 366.2 ? 55.9 mV.ms and 617.2 ? 74.6 mV.ms (p ? 0.02). The mean amplitude for HCN1f/fand HCN1f/f,creSchaffer collateral
EPSPs respectively, was 10.6 ? 1.6 mV and 11.5 ? 1.7 mV (p ? 0.7) with a stimulus of duration of 0.1 ms and intensity of 6 ? 0.8 V and 5.7 ?
0.5 V (p ? 0.75). Subthreshold perforant path EPSPs from HCN1f/fmice (n ? 8) and HCN1f/f,cremice (n ? 12) evoked by stimuli of amplitude
15 V and duration 0.1 ms had 10%–90% rise times of 7.14 ? 0.52 ms and 8.82 ? 0.80 ms, respectively (p ? 0.14), halfwidths of 38.8 ? 6.4
ms and 57.6 ? 4.6 ms (p ? 0.03). The mean amplitude for HCN1f/fand HCN1f/f,creperforant path EPSPs, respectively, was 1.6 ? 0.4 mV and
2.4 ? 0.3 mV (p ? 0.1) in response to stimuli with amplitude of 15 V and 1.7 ? 0.5 mV and 3.59 ? 0.6 mV (p ? 0.04) in response to stimuli
with amplitude of 25 V.
(C) Plot of the relationship between stimulus strength and the slope of EPSPs in CA1 pyramidal cells, from HCN1f/fmice (n ? 6) and HCN1f/f,cre
mice (n ? 8), evoked by stimulation of Schaffer collateral inputs.
(F and G) Plot of relationship between stimulus strength and the slope of field EPSPs recorded from stratum radiatum (F) (HCN1f/f, n ? 10,
HCN1f/f,cre, n ? 16) or stratum lacunosum moleculare (G) (HCN1f/f, n ? 8, HCN1f/f,cre, n ? 11) in response to activation of Schaffer collateral or
perforant path inputs, respectively.
Can the absence of HCN1 channels from distal den-
drites of CA1 pyramidal cells account for the enhanced
perforant path LTP or the modified theta field activity?
The modified integrative properties of CA1 pyramidal
cells may lead to a larger postsynaptic response that
enhances the induction of perforant path LTP and in-
creases the size of theta oscillations. One potential
mechanism is provided by the findings that the NMDA
than that of Schaffer collateral input and is opposed
by Ih(Otmakhova and Lisman, 2004; Otmakhova et al.,
2002). Thus, the removal of HCN1 channels may en-
hance theta-related depolarization and promote the in-
duction of LTP by allowing increased activation of distal
NMDA receptors at perforant path synapses. Deletion
of HCN1 may also preferentially enhance the firing of
local dendritic Ca2?spikes, which are of particular im-
portance for intrinsic theta oscillations and for induction
of LTP at distal synapses on CA1 pyramidal neurons
(Golding et al., 2002; Kamondi et al., 1998).
We found no evidence for any role of HCN1 in basal
synaptic transmission, and ultrastructural studies have
Figure 7. HCN1 Constrains LTP at Perforant
Path, but Not Schaffer Collateral Inputs to
CA1 Pyramidal Cells
(A and B) The mean slope of Schaffer collat-
eral (A) (HCN1f/f, n ? 7, HCN1f/f,cre, n ? 9) and
perforant path (B) (HCN1f/f, n ? 7, HCN1f/f,cre,
n ? 11) fEPSPs normalized to the average
LTP is induced at time zero. Insets show rep-
resentative field EPSPs from HCN1f/fand
HCN1f/f,cremice. Scale bars in (A) are 10 ms
and 0.5 mV and in (B) are 20 ms and 0.2 mV.
There was a significant difference between
HCN1f/fand HCN1f/f,cremice in the amplitude
(p ? 0.001), but not Schaffer collateral inputs
(p ? 0.05).
localized HCN1 to postsynaptic dendrites, but not pre-
synaptic axonal inputs to CA1 or subicular pyramidal
neurons (Lorincz et al., 2002; Notomi and Shigemoto,
2004), supporting a postsynaptic integrative role of
HCN1 in modulation of perforant path inputs. Although
contribution to the enhanced LTP can be ruled out since
the LTP experiments were performed in the presence
of GABAAand GABABblockers. Although the perforant
path input to CA1 is the single most important contribu-
tor to the extracellular theta field in the hippocampus
(Buzsaki, 2002), actions of HCN1 channels in interneu-
rons in these in vivo experiments cannot be ruled out.
However, it is unlikely that the modified theta is due to
the loss of HCN1 from septal neurons, which provide
an input to the hippocampus that is important for theta
rhythm generation, as HCN1 expression levels in the
medial septum are low (Santoro et al., 2000) and block-
ates, rather than increases, the power of hippocampal
theta activity (Xu et al., 2004).
Selective Control of Plasticity at Perforant Path
Inputs to the Distal Dendrites of CA1 Pyramidal
Cells: A Possible Mechanism for the Influence
of HCN1 Channels on Learning and Memory
How might deletion of HCN1 from CA1 pyramidal cells
enhance performance of learning and memory tasks?
The hippocampal trisynaptic circuit, from layer II of the
entorhinal cortex via the dentate gyrus to CA3 and then
to CA1, has been thought of as the major pathway for
the relay of information from the associational areas of
the neocortex through the hippocampus (Anderson et
al., 1971). However, synaptic plasticity in the trisynaptic
layer III of the entorhinal cortex to neurons in CA1 may
also be of functional importance (Figure 8) (Brun et al.,
2002; Huang et al., 1995; Witter et al., 2000). Our study
shows that synaptic plasticity at these direct entorhinal
inputs, but not at the more proximal Schaffer collateral
inputs, is strongly regulated by HCN1 channels. Thus,
the enhanced performance in spatial learning and mem-
HCN1 Channels Constrain Learning and Memory
Figure 8. The Organization of Medial Temporal Lobe Circuits
Information flows through the hippocampal trisynaptic circuit from layer II of the entorhinal cortex (EC) via the dentate gyrus (DG) to CA3 and
then to CA1. The input from CA3 to CA1, via the Schaffer collateral pathway, targets the proximal dendrites of CA1 pyramidal neurons and
is indicated by the dashed line. In addition, CA1 neurons receive a direct input from the EC to their distal apical dendrites via the perforant
path indicated by the bold line. The distal apical dendrites of CA1 pyramidal neurons contain the highest density of HCN1 channels. This
figure was modified from Huang et al., 1995.
ory tasks in the HCN1f/f,cremice may reflect the removal
of the inhibitory constraint on LTP at the perforant path
inputs to CA1 neurons. It has been suggested that CA1
pyramidal neurons compare sensory information pro-
vided via the direct cortical perforant path inputs with
predictions made by the dentate/CA3 region on the ba-
sis of previously stored information (Lisman, 1999). In-
creased LTP of perforant path inputs in the absence of
HCN1 may, thus, facilitate recall when stored associa-
tions from CA3 match information about a familiar envi-
ronment provided by the perforant path input to CA1.
Because HCN1 was deleted throughout the forebrain,
we cannot definitively assign the behavioral phenotype
to any one cell type. In addition to CA1 pyramidal neu-
rons, HCN1 is also strongly expressed in pyramidal neu-
pyramidal neurons, and in neocortical layer V pyramidal
neurons. Since HCN1 also shows its characteristic gra-
dient of increasing expression in the apical dendrites
of these pyramidal cells, deletion of HCN1 may also
enhance plasticity at their distal synaptic inputs. More-
over, the present data do not rule out a contribution of
HCN1 in nonpyramidal neurons, for example entorhinal
stellate cells, to the behavioral phenotype. Further spa-
tial restriction of the HCN1 deletion will be required to
distinguish these possibilities.
In summary, our findings suggest that learning and
memory is constrained by the spatial and temporal in-
tegrative properties of neuronal dendrites. The ability to
enhance the performance of a behavior by deleting an
ion channel is surprising and leads to the question of
what evolutionary advantageous function the channel
may perform. HCN1 channels may maximize the infor-
mation capacity of pyramidal cells by normalizing the
somatic waveform of synaptic inputs so that temporal
the ability to store information during different behav-
ioral states may shift the relative importance of proximal
versusdistal synapticinputsbymodulating HCN1chan-
nel activity. Both possibilities are compatible with our
findings as the demands placed on forebrain neurons
during the behavioral tasks that we have used may be
well below the maximum capacity to which they have
evolved. Thus, the dendritic mechanisms for normaliza-
tion or modulation of synaptic inputs may have evolved
at a modest cost to the ability to modify the strength of
The generation of the lines of mice used in this study has been
described previously (Nolan et al., 2003). HCN1f/f,creand HCN1f/flit-
termates with mixed average 50/50% 129SVEV/C57 background
were used in experiments. Mice were maintained and bred under
standard conditions, consistent with National Institutes of Health
guidelines and approved by the Institutional Animal Care and Use
Committee. Mean values are stated as ? standard error of the mean
(SEM). On figures in which error bars are absent the SEM is smaller
than the symbol size.
For all behavioral tasks, mutant and control littermates (males, 3
months old) were used. Statistical analyses used ANOVAs with ge-
notype as the between-subject factor. The experimenter was blind
to the genotype in all studies.
leret et al., 1999) with three training phases: 2 days with a visible
platform followed by 4 days with a hidden platform in quadrant
number 3, followed by 8 days with the hidden platform in quadrant
number 1. For the first two phases, four trials, 120 s maximum
duration and 15-min intertrial interval, were given daily. For the third
phase, only one trial was given each day. In the first experiment,
15 min before the first trial of each phase, the mice were primed by
allowing them to rest on the platform for 15 s. The second experi-
ment used a new group of mice and the same experimental para-
digm, expect that the priming procedure was omitted. Probe trials,
during which the platform was removed from the maze, lasted 60
s. The trajectories of mice in the maze were recorded with a video
tracking system (HVS Image Analysis System VP-118).
phosphocreatine (10). Series resistances were ?15 M? for voltage-
clamp experiments and ?40 M? for current-clamp experiments.
There was no significant difference between the series resistance
of recordings between experimental groups in either configuration.
Series resistance in voltage-clamp recordings was compensated
by 50%–80%. For current-clamp recordings appropriate bridge and
electrode capacitance compensation were applied. Membrane cur-
rent and voltagewere filtered at 1–2KHz and 4–20 KHzand sampled
at 5–10 KHz and 10–50 KHz for voltage- and current-clamp experi-
Glutamatergic EPSPs were investigated in the presence of either
bicuculline or picrotoxin and CGP55845 to block GABAAand GABAB
receptors, respectively. CA3 was removed from the slice to prevent
emergence of epileptiform discharges after blocking GABA recep-
tors. A concentric bipolar stimulating electrode was placed in stra-
tum radiatum to evoke Schaffer collateral inputs or stratum lacuno-
sum molecularae to evoke perforant path inputs. A patch electrode
containing ACSF was placed in the same layer as the stimulating
electrode at a distance of 100–200 ?m to record the corresponding
field EPSP. Selective activation of perforant path axons was con-
firmed by observation of a reversal in the polarity of the field EPSP
when it was recorded from stratum radiatum (Colbert and Levy,
1993). Synaptic responses to various intensities of stimulation
(3–20 V, 0.1 ms) were recorded locally with field electrodes and
in the whole-cell configuration from the soma of individual CA1
pyramidal cells. The stimulus strength was then set to give a field
EPSP with slope approximately 50% of the maximum. A 10–15 min
baseline wasobtained beforegiving fourtheta-burst trainsof stimuli
at 20 s intervals. Each theta-burst train consisted of five bursts at
a frequency of 5 Hz with each burst containing 10 stimuli at a fre-
quency of 100 Hz. Synaptic responses were then recorded for a
further 40 min and in most cases for ?60 min. Protocols used for
additional LTP experiments are described in Supplemental Figure
S6 on the Cell website.
Data were analyzed in IGOR pro (Wavemetrics) using custom
written routines. To determine impedance-frequency relationships,
membrane potential response to current inputs that oscillated with
Membrane impedance was calculated from the ratio of the Fourier
current. The field EPSP slope was determined by fitting a straight
line to the initial rising phase following the fiber volley. An identical
time window (width 1–2 ms) was used for slope measurements of
all field EPSPs from the same experiment. Statistical significance
was tested with Student’s t test.
In Vivo Electrophysiology
Male HCN1?/?(n ? 9) and HCN1?/?(n ? 6) mice were implanted with
chronic recording tetrodes (four 12-?m polyimide-coated nichrome
wires) or 60-?m wires under anesthesia (Buhl et al., 2003). The
electrodes were gradually moved to the CA1 pyramidal layer. The
position of the electrode in the CA1 pyramidal layer was determined
by the presence of fast oscillations (“ripples”) in association with
synchronous discharge of neurons (Buzsa ´ki et al., 2003). Once in
the layer, maximum ripple amplitudes were used as an online refer-
ence for consistent electrode placement between animals.
Electrical activity was recorded from each animal during slow
wave sleep (SWS) and rapid eye movement sleep (REM) in its home
cage and during the wake cycle while the animal was running in a
wheel. The mice were allowed to freely explore the apparatus which
consisted of a running wheel (29.5 cm in diameter) and adjacent
box (30 cm ? 40 cm ? 35 cm; Czurko et al., 1999). Field potential
and unit activity were recorded after being amplified (2000?) and
band-pass filtered (1–5 kHz; model 12-64 channel; Grass Instru-
ments,Quincy, Massachusetts),digitized with14-bit resolutioncon-
tinuously at 20 kHz, and recorded on a PC using custom Labview
software (National Instruments, Austin, Texas). Running speed of
the mouse was measured from the output of an optical encoder
recordings were made, recordings from each hemisphere were
treated as independent data.
The beginning, middle, and end of CA1 ripple events were de-
tected by applying an amplitude threshold (mean ?7 SD) to the
previously bandpass-filtered (150–250 Hz) EEG. Theta and gamma
transform to the previously bandpass-filtered (4–12 Hz and 30–80
was then calculated for every 10? of the theta cycle. All comparisons
of ripples, theta, and gamma events were taken from the same
electrode location in the CA1 pyramidal layer. At the end of the
experiment, the animal was perfused and the location of the elec-
trode tips were histologically verified (Buhl et al., 2003). Differences
between groups were compared with an unpaired Student’s t test.
on the manuscript. This work was supported by the Howard Hughes
Medical Institute, The Kavli Insitute for Brain Sciences, grants from
NIH to E.R.K. (MH045923), S.A.S. (NS36658), and G.B. (NS34994,
ate Research Fellowship (J.T.D.), and the New York State Research
Foundation for Mental Hygiene. E.R.K. is one of the four founders
of Memory Pharmaceuticals and Chairman of its Scientific Advisory
Board. Memory Pharmaceuticals is concerned with developing
drugs for age-related memory loss. Some of these drugs are also
potentially useful in depression and schizophrenia.
In Vitro Electrophysiololgy
Horizontal brain slices were prepared from 4- to 10-week-old mice.
Mice were decapitated, their brains rapidly removed and placed in
cold (2?C ?4?C) modified ACSF of composition (mM) NaCl (86),
NaH2PO4 (1.2), KCl (2.5), NaHCO3 (25), Glucose (25), CaCl2(0.5),
MgCl2(7), sucrose (75). The hemisected brain was glued to an agar
block and cut submerged under cold modified ACSF into 400 ?m
sections with a Vibratome 3000 sectioning system. Slices were
transferred to a storage container filled with standard ACSF at 33?C
?35?C for 20–30 min and then allowed to cool to room temperature
(20?C ?22?C). The standard ACSF had the following composition
(mM): NaCl (124), NaH2PO4(1.2), KCl (2.5), NaHCO3(25), Glucose
(20), CaCl2(2), MgCl2(1). The modified ACSF for isolation of Ihhad
the following composition (mM): NaCl (115), NaH2PO4 (1.2), KCl (5),
NaHCO3 (25), glucose (20), CaCl2 (2), MgCl2 (1), BaCl2 (1), CdCl2
(0.1), 4-AP (1), TEA (5), NBQX (0.005), bicuculline (0.02), and TTX
(0.0005). For recording, slices were transferred to a submerged
chamber at room temperature for isolation of Ihand at 33?C ?35?C
for all other experiments. Whole-cell recordings were obtained from
tion with DIC optics, using 2–5 M? resistance electrodes filled with
intracellular solution of composition (mM): KMethylsulfate (120), KCl
(20), HEPES (10), MgCl2(2), EGTA (0.1), Na2ATP (4), Na2GTP (0.3),
Received: June 6, 2004
Revised: August 27, 2004
Accepted: October 4, 2004
Published: November 23, 2004
Abel, T., Martin, K.C., Bartsch, D., and Kandel, E.R. (1998). Memory
suppressor genes: inhibitory constraints on the storage of long-
term memory. Science 279, 338–341.
tion of hippocampal pathways. Exp. Brain Res. 13, 222–238.
Berger, T., Larkum, M.E., and Luscher, H.R. (2001). High I(h) channel
HCN1 Channels Constrain Learning and Memory
density in the distal apical dendrite of layer V pyramidal cells in-
creases bidirectional attenuation of EPSPs. J. Neurophysiol. 85,
Bi, G.Q., and Poo, M.M. (1998). Synaptic modifications in cultured
hippocampal neurons: dependence on spike timing, synaptic
strength, and postsynaptic cell type. J. Neurosci. 18, 10464–10472.
Bliss, T.V., and Collingridge, G.L. (1993). A synaptic model of mem-
ory: long-term potentiation in the hippocampus. Nature 361, 31–39.
Bliss, T.V., Collingridge, G.L., and Morris, R.G. (2003). Introduction.
Long-term potentiation and structure of the issue. Philos. Trans. R.
Soc. Lond. B Biol. Sci. 358, 607–611.
Brun, V.H., Otnass, M.K., Molden, S., Steffenach, H.A., Witter, M.P.,
Moser, M.B., and Moser, E.I. (2002). Place cells and place recogni-
tion maintained by direct entorhinal-hippocampal circuitry. Science
Buhl, D.L., Harris, K.D., Hormuzdi, S.G., Monyer, H., and Buzsaki,
G. (2003). Selective impairment of hippocampal gamma oscillations
Buzsa ´ki, G. (2002). Theta oscillations in the hippocampus. Neuron
A. (2003). Hippocampal network patterns of activity in the mouse.
Neuroscience 116, 201–211.
polarization-activated pacemaker current defined by coassembly of
HCN1 and HCN2 subunits and basal modulation by cyclic nucleo-
tide. J. Gen. Physiol. 117, 491–504.
Colbert, C.M., and Levy, W.B. (1993). Long-term potentiation of
perforant path synapses in hippocampal CA1 in vitro. Brain Res.
Czurko, A., Hirase, H., Csicsvari, J., and Buzsaki, G. (1999). Sus-
tained activation of hippocampal pyramidal cells by ‘space clamp-
ing’ in a running wheel. Eur. J. Neurosci. 11, 344–352.
Fisahn, A.,Yamada, M.,Duttaroy, A., Gan,J.W., Deng,C.X., McBain,
C.J., and Wess, J. (2002). Muscarinic induction of hippocampal
gamma oscillations requires coupling of the M1 receptor to two
mixed cation currents. Neuron 33, 615–624.
Gasparini, S., and DiFrancesco, D. (1997). Action of the hyperpolar-
ization-activated current (Ih) blocker ZD 7288 in hippocampal CA1
neurons. Pflugers Arch. 435, 99–106.
Golding, N.L., Staff, N.P., and Spruston, N. (2002). Dendritic spikes
as a mechanism for cooperative long-term potentiation. Nature
Hu, H., Vervaeke, K., and Storm, J.F. (2002). Two forms of electrical
resonance at theta frequencies, generated by M-current, h-current
and persistent Na? current in rat hippocampal pyramidal cells. J.
Physiol. 545, 783–805.
Huang, Y.Y., Kandel, E.R., Varshavsky, L., Brandon, E.P., Qi, M.,
Idzerda, R.L., McKnight, G.S., and Bourtchouladze, R. (1995). A
genetic test of the effects of mutations in PKA on mossy fiber LTP
and its relation to spatial and contextual learning. Cell 83, 1211–
Kamondi, A., Acsady, L., Wang, X.J., and Buzsaki, G. (1998). Theta
oscillations in somata and dendrites of hippocampal pyramidal cells
in vivo: activity-dependent phase-precession of action potentials.
Hippocampus 8, 244–261.
Kubota, M., Murakoshi, T., Saegusa, H., Kazuno, A., Zong, S.Q., Hu,
Q.P., Noda, T., and Tanabe, T. (2001). Intact LTP and fear memory
but impaired spatial memory in mice lacking Ca(v)2.3 (alpha(1E))
channel. Biochem. Biophys. Res. Commun. 282, 242–248.
of memory sequences by reciprocal dentate-CA3 interactions. Neu-
ron 22, 233–242.
Lorincz, A., Notomi, T., Tamas, G., Shigemoto, R., and Nusser, Z.
(2002). Polarized and compartment-dependent distribution of HCN1
in pyramidal cell dendrites. Nat. Neurosci. 5, 1185–1193.
Ludwig, A., Budde, T., Stieber, J., Moosmang, S., Wahl, C., Holthoff,
K., Langebartels, A., Wotjak, C., Munsch, T., Zong, X., et al. (2003).
Absence epilepsy and sinus dysrhythmia in mice lacking the pace-
maker channel HCN2. EMBO J. 22, 216–224.
Lupica, C.R., Bell, J.A., Hoffman, A.F., and Watson, P.L. (2001).
brane potential and GABA release in hippocampal interneurons. J.
Neurophysiol. 86, 261–268.
Maccaferri, G., and McBain, C.J. (1996). The hyperpolarization-acti-
vated current (Ih) and its contribution to pacemaker activity in rat
CA1 hippocampal stratum oriens-alveus interneurones. J. Physiol.
Magee, J.C. (1998). Dendritic hyperpolarization-activated currents
modify the integrative properties of hippocampal CA1 pyramidal
neurons. J. Neurosci. 18, 7613–7624.
Magee, J.C. (1999). Dendritic lh normalizes temporal summation in
hippocampal CA1 neurons. Nat. Neurosci. 2, 508–514.
Magee, J.C., and Johnston, D. (1997). A synaptically controlled,
associative signal for Hebbian plasticity in hippocampal neurons.
Science 275, 209–213.
Malleret, G., Hen, R., Guillou, J.L., Segu, L., and Buhot, M.C. (1999).
5-HT1B receptor knockout mice exhibit increased expolaratory ac-
tivity and enhanced spatial performance in the Morris water maze.
J. Neurosci. 19, 6157–6168.
Malleret, G., Haditsch, U., Genoux, D., Jones, M.W., Bliss, T.V.,
Vanhoose, A.M., Weitlauf, C., Kandel, E.R., Winder, D.G., and Man-
suy, I.M. (2001). Inducible and reversible enhancement of learning,
memory, and long-term potentiation by genetic inhibition of cal-
cineurin. Cell 104, 675–686.
Markram, H., Lubke, J., Frotscher, M., and Sakmann, B. (1997).
Regulation of synaptic efficacy by coincidence of postsynaptic APs
and EPSPs. Science 275, 213–215.
Meiri, N., Ghelardini, C., Tesco, G., Galeotti, N., Dahl, D., Tomsic,
D., Cavallaro, S., Quattrone, A., Capaccioli, S., Bartolini, A., and
Alkon, D.L. (1997). Reversible antisense inhibition of Shaker-like
Kv1.1 potassium channel expression impairs associative memory
in mouse and rat. Proc. Natl. Acad. Sci. USA 94, 4430–4434.
Milner, B., Squire, L.R., and Kandel, E.R. (1998). Cognitive neurosci-
ence and the study of memory. Neuron 20, 445–468.
B., Yin, D., Thompson, R.F., Siegelbaum, S.A., Kandel, E.R., and
Morozov, A. (2003). The hyperpolarization-activated HCN1 channel
Purkinje cells. Cell 115, 551–564.
Notomi, T., and Shigemoto, R. (2004). Immunohistochemical local-
ization of Ih channel subunits, HCN1–4, in the rat brain. J. Comp.
Neurol. 471, 241–276.
Otmakhova, N.A., and Lisman, J.E. (2004). Contribution of Ih and
GABAB to synaptically induced afterhyperpolarizations in CA1: A
brake on the NMDA response. J. Neurophysiol. 92, 2027–2039.
Otmakhova, N.A., Otmakhov, N., and Lisman, J.E. (2002). Pathway-
specific properties of AMPA and NMDA-mediated transmission in
CA1 hippocampal pyramidal cells. J. Neurosci. 22, 1199–1207.
Pike, F.G., Goddard, R.S., Suckling, J.M., Ganter, P., Kasthuri, N.,
and Paulsen, O. (2000). Distinct frequency preferences of different
types of rat hippocampal neurones in response to oscillatory input
currents. J. Physiol. 529, 205–213.
ulates plasticity and spiking in CA1 pyramidal neurons. Nature
Robinson, R.B., and Siegelbaum, S.A. (2003). Hyperpolarization-
activated cation currents: from molecules to physiological function.
Annu. Rev. Physiol. 65, 453–480.
Routtenberg, A., Cantallops, I., Zaffuto, S., Serrano, P., and Nam-
gung, U. (2000). Enhanced learning after genetic overexpression of
a brain growth protein. Proc. Natl. Acad. Sci. USA 97, 7657–7662.
Sanes, J.R., and Lichtman, J.W. (1999). Can molecules explain long-
term potentiation? Nat. Neurosci. 2, 597–604.
Santoro, B., Grant, S.G., Bartsch, D., and Kandel, E.R. (1997). Inter-
active cloning with the SH3 domain of N-src identifies a new brain
Cell Download full-text
specific ion channel protein, with homology to eag and cyclic nucle-
otide-gated channels. Proc. Natl. Acad. Sci. USA 94, 14815–14820.
G.R., and Siegelbaum, S.A. (2000). Molecular and functional hetero-
geneity of hyperpolarization-activated pacemaker channels in the
mouse CNS. J. Neurosci. 20, 5264–5275.
Strohmann, B., Schwarz, D.W., and Puil, E. (1994). Subthreshold
frequency selectivity in avian auditory thalamus. J. Neurophysiol.
Tang, Y.P., Shimizu, E., Dube, G.R., Rampon, C., Kerchner, G.A.,
Zhuo, M., Liu, G., and Tsien, J.Z. (1999). Genetic enhancement of
learning and memory in mice. Nature 401, 63–69.
and HCN2 pacemaker channels. J. Biol. Chem. 276, 6069–6072.
tional changes in spatial dendritic integration accompanying long-
term synaptic modifications. Neuron 37, 463–472.
Williams, S.R., and Stuart, G.J. (2000). Site independence of EPSP
time course is mediated by dendritic I(h) in neocortical pyramidal
neurons. J. Neurophysiol. 83, 3177–3182.
Williams, S.R., and Stuart, G.J. (2003). Voltage- and site-dependent
control of the somatic impact of dendritic IPSPs. J. Neurosci. 23,
Witter, M.P., Wouterlood, F.G., Naber, P.A., and Van Haeften, T.
(2000). Anatomicalorganization ofthe parahippocampal-hippocam-
pal network. Ann. N Y Acad. Sci. 911, 1–24.
Xu, C., Datta, S., Wu, M., and Alreja, M. (2004). Hippocampal theta
rhythm is reduced by suppression of the H-current in septohippo-
campal GABAergic neurons. Eur. J. Neurosci. 19, 2299–2309.