Impaired Dendritic Expression and Plasticity of h-Channels in the fmr1 (cid:1) /y Mouse Model of Fragile X Syndrome

SUMMARY Despite extensive research into both synaptic and morphological changes, surprisingly little is known about dendritic function in fragile X syndrome (FXS). We found that the dendritic input resistance of CA1 neurons was signiﬁcantly lower in fmr1 (cid:1) /y versus wild-type mice. Consistent with elevated dendritic I h , voltage sag, rebound, and resonance frequency were signiﬁcantly higher and temporal summation was lower in the dendrites of fmr1 (cid:1) /y mice. Dendritic expression of the h-channel subunit HCN1, but not HCN2, was higher in the CA1 region of fmr1 (cid:1) /y mice. Interestingly, whereas mGluR-medi-ated persistent decreases in I h occurred in both wild-type and fmr1 (cid:1) /y mice, persistent increases in I h that occurred after LTP induction in wild-type mice were absent in fmr1 (cid:1) /y mice. Thus, chronic upregulation of dendritic I h in conjunction with impairment of homeostatic h-channel plasticity represents a dendritic channelopathy in this model of mental retardation and may provide a mechanism for the cognitive impairment associated with FXS.


INTRODUCTION
Fragile X syndrome (FXS) is the most common form of inherited mental retardation with a variety of phenotypes including impaired cognitive ability, problems with working memory, autistic behavior, and increased incidence of epilepsy. FXS is characterized by the absence of the RNA-binding protein fragile X mental retardation protein (FMRP) (Bell et al., 1991). Mice lacking FMRP, such as the fmr1 knockout (fmr1 À/y ) mouse, perform poorly in hippocampal-dependent spatial learning tasks (Bakker et al., 1994). Interestingly, the performance of fmr1 À/y mice was similar to mice lacking proteins involved in long-term potentiation (LTP), a proposed cellular correlate of learning (e.g., Silva et al., 1992).
Several FMRP-target mRNAs are localized in neuronal dendrites and encode for voltage-gated channels or proteins involved in the regulation of channel expression and/or function (Darnell et al., 2011;Bassell and Warren, 2008). Incorrect gene expression, channel function, and/or loss of posttranslational channel regulation can have substantial effects on the control of cellular excitability. The physiological regulation of cellular excitability and dendritic integration through plasticity of voltage-gated ion channels is hypothesized to maintain the input-output characteristics of a neuron within normal limits (Turrigiano and Nelson, 2000). However, changes in dendritic channels in neurological disorders or disease can also alter the input-output relationship and result in a pathophysiological state (Bernard et al., 2004;Jung et al., 2007). One channel that undergoes activity-dependent plasticity and is altered in neurological disease is the h-channel (van Welie et al., 2004;Brager and Johnston, 2007;Fan et al., 2005;Jung et al., 2007;Shin et al., 2008).
h-channels are widely distributed in the central nervous system. In the hippocampus, h-channels are composed primarily of HCN1 and HCN2 subunits (Santoro et al., 2000). In CA1 pyramidal neurons the density of h-channels increases with distance from the soma along the apical dendrite (Magee, 1998). The high density of h-channels in the dendrites allows I h (the current mediated by h-channels) to significantly contribute to the total membrane conductance and thereby exert strong influence over neuronal function in the subthreshold voltage range near rest. Subtle modifications in the physiology of h-channels can produce significant changes in synaptic integration and neuronal excitability (Santoro et al., 2000;Magee, 1999;Poolos et al., 2002).
Using whole-cell recording from CA1 pyramidal neuron apical dendrites, we show that dendritic physiology in the fmr1 À/y mouse is altered in a manner consistent with elevated dendritic I h . Differences in hippocampal HCN1 protein expression were consistent with our physiological results. Interestingly, activitydependent increases in I h following LTP induction were absent in fmr1 À/y mice. Elevated dendritic expression of I h coupled with impaired h-channel plasticity might thus indicate some of the cellular mechanisms underlying the cognitive impairments associated with FXS.
ion channels, strongly influence the manner in which synaptic inputs are combined and propagated from the dendrites to the soma. In particular a proportion of the total population of h-channels is active at rest and contributes to the resting membrane potential (V m ), input resistance (R N ), and membrane time constant (t m ) of CA1 neurons (Magee, 1998). We used somatic and dendritic whole-cell recording to test whether there were significant differences in the V m , R N , and t m of CA1 neurons between fmr1 À/y and wild-type (WT) mice. We found that V m was significantly more depolarized, R N was lower, and t m was faster, in the dendrites versus the soma for both WT and fmr1 À/y mice; consistent with a higher density of h-channels in the apical dendrite. There were no significant differences between WT and fmr1 À/y mice using somatic recordings (see Figure S1 available online). When we compared the dendrites between fmr1 À/y and WT mice, however, we found that the dendritic V m of fmr1 À/y mice was significantly depolarized versus WT (knockout [ko], À61 ± 0.6mV; WT, À64 ± 0.6mV; p < 0.05). Furthermore, dendritic R N was significantly lower ( Figures 1A-1C) and dendritic t m was significantly faster (WT, 17 ± 2 ms; ko, 8 ± 1 ms; p < 0.05) in fmr1 À/y mice versus WT.
To test the hypothesis that the differences in dendritic properties were due to elevated I h , we took advantage of several unique characteristics of h-channels and measured electrophysiological parameters that are sensitive to the amount of I h (Brager and Johnston, 2007). Voltage sag (sag) during step current injections can be due to the relatively slow activation/ deactivation kinetics I h (Magee, 1998). We found that dendritic sag from hyperpolarizing current steps was 38% higher in fmr1 À/y mice versus WT ( Figures 1D-1F). Rebound amplitude following a hyperpolarization is due in part to the slow closing of h-channels. The amount of rebound from a given steadystate hyperpolarization is increased with larger I h (Brager and Johnston, 2007). Dendritic rebound was 43% higher in fmr1 À/y mice versus WT ( Figures 1G-1I). The same kinetic properties that underlie sag and rebound allow I h to act as a resonator conductance, and differences in resonance frequency (f R ) reflect differences in I h (Hutcheon and Yarom, 2000;Narayanan and Johnston, 2007). Similar to sag and rebound, dendritic f R was 42% higher in fmr1 À/y mice versus WT ( Figures 1J-1L). I h reduces temporal summation by repolarizing the membrane during an excitatory postsynaptic potential (EPSP) (Magee, 1998). Consistent with elevated I h , we found that dendritic temporal summation was significantly lower in fmr1 À/y mice versus WT ( Figures 1M-1O). When measured from the soma, we found no significant difference in sag, rebound, f R , or summation between WT and fmr1 À/y mice ( Figure S1).
As a further test of the contribution of I h to the differences in dendritic properties, we compared WT and fmr1 À/y dendritic recordings before and after application of the h-channel blocker ZD7288. Extracellular application of 20 mM ZD7288 significantly increased R N in both WT and fmr1 À/y dendrites ( Figure 1C). Furthermore, block of I h significantly decreased sag, rebound, and f R and increased temporal summation in both WT and fmr1 À/y dendrites ( Figures 1F, 1I, 1L, and 1O). After block of I h by ZD7288, dendritic properties were no longer significantly different between WT and fmr1 À/y dendrites. These electrophys-iological results suggest that dendritic, but not somatic, I h is significantly higher in fmr1 À/y mice.
Elevated HCN1 Levels in the CA1 Region of the fmr1 À/y Mouse Hippocampus Our electrophysiological data support the hypothesis that I h is higher in the dendrites of CA1 neurons from fmr1 À/y mice versus WT mice. We sought to further test this hypothesis by examining h-channel proteins using immunohistochemistry and western blotting. We used antibodies to label the hchannel subunits HCN1 and HCN2. As expected, the intensity of HCN1 and HCN2 labeling increased with distance from the cell body layer indicative of the h-channel gradient. Measuring the intensity along a line perpendicular to, but not including, the pyramidal cell body layer in CA1 allows for a comparison of the distribution of HCN1 and HCN2 staining between slices (Shin et al., 2008). We found that the intensity of the HCN1 fluorescence, but not HCN2, was higher in the distal dendrites of fmr1 À/y mice versus WT mice (Figures 2A-2F). The difference in HCN1 staining between fmr1 À/y and WT mice persisted when the HCN1 signal was normalized to dendritic marker MAP2 ( Figure S2). We used western blot analysis to determine if there was higher HCN protein levels in the dendritic field of the CA1 region of the fmr1 À/y mouse versus WT. We found that HCN1, but not HCN2, protein levels were significantly higher in homogenates from the dendritic region of CA1 in fmr1 À/y versus WT ( Figures 2G and 2H). These results suggest that the higher dendritic I h in fmr1 À/y mice is due in part to an increase in the distal dendritic expression of the h-channel subunit HCN1.
Intrinsic Plasticity Is Impaired in fmr1 À/y Mice Homeostatic mechanisms that maintain neuronal output are essential for the stabilization of neural networks during classical Hebbian synaptic plasticity (Turrigiano, 2011). We previously demonstrated that activation of group I metabotropic glutamate receptors (mGluRs) results in long-term depression (LTD) and an accompanying decrease of I h in CA1 neurons (Brager and Johnston, 2007). In the hippocampus of fmr1 À/y mice, group I mGluR-dependent LTD is enhanced, whereas NMDA receptordependent LTD is unaffected (Huber et al., 2002). To test whether mGluR-dependent h-channel plasticity is altered in fmr1 À/y mice, we measured intrinsic properties before and after activation of group I mGluRs (100 mM DHPG for 10 min). Application of DHPG to either WT or fmr1 À/y CA1 neurons produced a transient depolarization that recovered by 30 min after wash ( Figure 3A). This transient activation of mGluRs decreased EPSP slope by 23% in WT and 40% in fmr1 À/y mice 30 min after DHPG application ( Figures 3B and 3C). In agreement with our previous observations in rat hippocampus (Brager and Johnston, 2007), we found that mGluR activation significantly increased R N by 85% in WT mice and by 83% in fmr1 À/y mice ( Figure 3D). Consistent with a decrease in I h , f R , rebound, and sag were significantly decreased after DHPG application in WT, and f R and rebound were significantly decreased in fmr1 À/y mice ( Figures 3E-3G). Based on these results, we conclude that mGluR-dependent plasticity of I h occurs in both WT and fmr1 À/y mice. (B) Group data showing that dendritic R N was significantly lower in fmr1 À/y mice (n = 6) versus WT (n = 6).
(C) Group data showing that application of ZD significantly increased R N in both WT (n = 4) and fmr1 À/y mice (n = 4).
(D) Representative dendritic whole-cell recordings showing that there was more sag (arrow) in the dendrites of fmr1 À/y mice versus WT.
(E) Group data showing that dendritic sag was significantly higher in fmr1 À/y mice. (F) Group data showing that application of ZD significantly decreased sag in both WT and fmr1 À/y mice. (G) Representative dendritic whole-cell recordings (left) and plot used to calculate rebound slope (right) from WT and fmr1 À/y mice.
(H) Group data showing that dendritic rebound slope was significantly higher in fmr1 À/y mice. (I) Group data showing that application of ZD significantly decreased rebound in both WT and fmr1 À/y mice. (J) Representative voltage traces and ZAP used to determine f R (dashed line). Note the rightward shift in the fmr1 À/y (red) versus the WT (black).
(K) Group data showing that dendritic f R was significantly higher in fmr1 À/y mice. (L) Group data showing that application of ZD significantly decreases f R in both WT and fmr1 À/y mice. (M) Representative traces used to measure temporal summation of fmr1 À/y mice versus WT.
(N) Group data showing that dendritic temporal summation was significantly less in fmr1 À/y mice.
(O) Group data showing that application of ZD significantly increased dendritic summation in both WT and fmr1 À/y mice. Group data are presented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.005 significantly different from WT. yp < 0.05 significantly different from baseline.
In CA1 neurons, theta-burst firing of action potentials paired with EPSPs (theta-burst pairing, TBP) results in an NMDAdependent increase in I h (Fan et al., 2005). To test whether the higher dendritic I h in fmr1 À/y mice occludes the TBP-dependent increases in I h , we induced TBP-LTP in CA1 neurons from both WT and fmr1 À/y mice. CA1 neurons from both WT and fmr1 À/y mice displayed a significant increase in EPSP slope following TBP ( Figures 4A and 4B). In agreement with previous studies (Fan et al., 2005;Narayanan and Johnston, 2007), TBP was accompanied by a significant decrease in R N in WT mice (baseline, 82 ± 7 MU; TBP, 68 ± 6 MU; n = 5, p < 0.05) ( Figures  4C and 4D). Interestingly, there was no significant change in R N in fmr1 À/y mice. Consistent with an increase in I h , TBP significantly increased f R , rebound, and sag in WT mice ( Figures  4E-4G). In contrast, there was no significant change in f R , rebound, or sag in fmr1 À/y neurons after TBP. We thus conclude that the activity-dependent expression of I h plasticity that normally occurs following TBP-LTP is absent in fmr1 À/y mice.

DISCUSSION
Neurons from patients and animal models of FXS show significant morphological and synaptic alterations. Here, we show that a significant change to the intrinsic properties of the dendrites of CA1 neurons occurs in the fmr1 À/y mouse model of FXS. Electrophysiological data support the hypothesis that these changes are due in part to a dendritic enhancement of the hyperpolarization-activated nonselective cationic current, I h , in the distal dendrites. Furthermore, this elevation in I h appears to be due to increased distal dendritic expression of the HCN1 subunit. Interestingly, activity-dependent increases in I h are absent, whereas activity-dependent decreases in I h remain. These data represent the first physiological description of a channelopathy involving dendritic h-channels associated with FXS.

Impact on Dendritic Function
I h plays a complex role in controlling membrane excitability. What is the impact of elevated I h in FXS on dendritic function? Elevated I h will reduce dendritic excitability by decreasing R N as well as reducing temporal and spatial summation of subthreshold EPSPs. I h also constrains distal dendritic spikes (E) A plot of intensity as a function of location showing that distal dendritic HCN1 staining was higher in fmr1 À/y mice versus WT mice.
(F) Intensity profile showing that there was no difference in HCN2 staining between fmr1 À/y and WT mice. (G) The total amount of HCN1 protein in the CA1 region of the fmr1 À/y hippocampus was significantly higher than WT.
(H) Representative western blot from WT and fmr1 À/y mice for HCN2 and tubulin. There was no significant difference in the total amount of HCN2 protein between fmr1 À/y and WT hippocampus. Group data are presented as mean ± SEM.**p < 0.01.
by increasing inactivation of both T-and N-type Ca 2+ channels (Tsay et al., 2007). The normal gradient of h-channels regulates the propagation of signals between the dendrites and the soma, and changes in the amount of dendritic I h will strongly influence somato-dendritic coupling (Magee, 1998(Magee, , 1999Poolos et al., 2002). The increase in dendritic I h in FXS would decrease somato-dendritic coupling, reducing the likelihood of synaptic inputs eliciting an action potential at the soma. Elevated I h will not only have inhibitory influences on dendritic excitability but also increase membrane excitability by depolarizing the membrane potential closer to the threshold for dendritic spike initiation and by augmenting the depolarizing rebound potential that occurs following a hyperpolarization. We found that rebound from hyperpolarization was higher in the dendrites of fmr1 À/y mice. This increase in rebound either alone (Chen et al., 1999) or in concert with the activation of lowthreshold voltage-dependent calcium channels could elicit a dendritic spike or dendritic plateau potential (Berger et al., 2003;Tsay et al., 2007). There is a high density of GABAergic synapses and GABA B receptors in the distal dendrites of CA1 neurons (Otmakhova and Lisman, 2004). It is, therefore, possible that in fmr1 À/y mice the normal inhibitory input into the distal dendrites of CA1 neurons will in fact result in significant dendritic excitation.
These two opposing actions of elevated I h on cellular excitability will have the complicated impact of decreasing R N , (D) Group data showing that R N significantly increased after DHPG washout in both WT and fmr1 À/y mice. (E) Group data showing that f R significantly decreased after DHPG washout in both WT and fmr1 À/y mice. (F) Group data showing that rebound slope significantly decreased after DHPG washout in both WT and fmr1 À/y mice. (G) Group data showing that sag significantly decreased after DHPG washout in WT but not fmr1 À/y mice. Group data are presented as mean ± SEM.*p < 0.05; **p < 0.01. temporal and spatial summation, and somato-dendritic coupling, making it harder for synaptic inputs to summate and elicit a somatic action potential, whereas also increasing the likelihood of local dendritic regenerative events. The exact impact of elevated dendritic I h on information processing in FXS requires further investigation.

I h Plasticity in FXS
Transient and persistent changes in I h , which alters neuronal intrinsic excitability, occur during and after induction of synaptic plasticity (Brager and Johnston, 2007;Fan et al., 2005). This coordinated change in synaptic and intrinsic excitability is not completely understood, but one current hypothesis is that intrinsic plasticity can alleviate the inherent instability of synaptic plasticity. Following induction of LTP there is a persistent increase in I h throughout the dendrites of CA1 neurons (Fan et al., 2005;Narayanan and Johnston, 2007). We found that although TBP-dependent LTP was not different between WT and fmr1 À/y mice, I h plasticity was absent in the latter. One possibility is that changes in the intracellular signaling molecules occlude any increase in I h after TBP-LTP. For example, CAMKII levels are elevated in fmr1 À/y mice (Zalfa et al., 2003), and the activation of CAMKII is required for TBP-dependent I h plasticity (Fan et al., 2005). It is possible that the already higher levels of CAMKII lead to elevated dendritic I h and that TBP-LTP is unable to further increase in I h in fmr1 À/y mice. The concomitant increase in I h that normally occurs with LTP is believed to be one mechanism by which homeostatic regulation of neural network stability can be achieved (Narayanan and Johnston, 2007). Our results suggest that despite the appearance of normal (B) The time course of EPSP slope change after TBP for WT (n = 6) and fmr1 À/y mice (n = 4). (C) The time course of RN change after TBP for WT (n = 6) and fmr1 À/y mice (n = 4). (D) Group data showing that R N significantly decreased after TBP in WT but not fmr1 À/y mice. (E) Group data showing that f R significantly increased after TBP in WT but not fmr1 À/y mice. (F) Group data showing that rebound slope significantly increased after TBP in WT but not fmr1 À/y mice. (G) Group data showing that sag significantly increased after TBP in WT but not fmr1 À/y mice. Group data are presented as mean ± SEM.*p < 0.05; **p < 0.01; ***p < 0.005. synaptic potentiation, the deficit in intrinsic plasticity may lead to inherent network destabilization. The combined plasticity of I h and synaptic strength results in preferential tuning of the dendrites to selected synaptic inputs. This form of homeostatic plasticity is believed to be critical for information storage and processing by neuronal networks (Turrigiano and Nelson, 2000). A loss of normal h-channel plasticity, with or without accompanying deficits in synaptic plasticity, could have profound effects on neuronal signal processing.

Potential Mechanisms
How does the absence of FMRP lead to an increase in dendritic I h and HCN1 protein? FMRP may bind to and regulate the translation of HCN1 mRNA. Classically, FMRP acts as a translational repressor (but see Gross et al., 2011). If HCN1 mRNA is a target of FMRP, then the loss of FMRP will directly result in elevated HCN1 expression. Indeed, our data show that expression of HCN1 protein is higher in the CA1 region of the fmr1 À/y mouse hippocampus. FMRP may also interact directly with ion channel proteins. In the olfactory bulb and brain stem, FMRP can bind to and regulate the gating of the sodium-activated potassium channel, Slack (Brown et al., 2010). FMRP may bind directly to h-channel proteins or regulate the binding between HCN subunits and auxiliary proteins such as tetratricopeptide repeat-containing Rab8b-interacting protein (TRIP8b) (Santoro et al., 2004). The interaction between TRIP8b and HCN subunits regulates both the gating and surface expression of h-channels.
The elevated I h may also be a compensatory change due to the loss of FMRP-dependent regulation of another protein. FMRP is necessary for normal synapse elimination during development (Pfeiffer and Huber, 2007). One possibility is that excessive synaptic excitation may lead to a homeostatic increase in I h . Upregulation of hippocampal h-channels can occur following high levels of AMPA receptor activation (van Welie et al., 2004). Increased excitatory input from CA3 Schaeffer collaterals due to elevated neurotransmitter release (Deng et al., 2011) could lead to excess CA1 AMPA receptor activation in fmr1 À/y mice. Additionally, synaptic inputs from layer III of the entorhinal cortex (EC) onto the distal dendrites of CA1 neurons regulate the normal h-channel gradient (Shin and Chetkovich, 2007). Increased excitatory input from the EC, due to improper synaptic pruning, could lead to a compensatory increase in I h . Alternatively, elevated I h may be in response to loss of another ion channel. Invertebrate studies suggest that there may be homeostatic coregulation of I h and I A including functional reciprocity (Harris-Warrick et al., 1995). There are conflicting reports about Kv4.2 expression, the putative subunit of I A , in the fmr1 À/y hippocampus. One study reported that FMRP promotes translation of Kv4.2 mRNA and that Kv4.2 protein levels are reduced in fmr1 À/y mice (Gross et al., 2011). More recently, a second study reported that FMRP acts as a translational repressor and that Kv4.2 expression is higher in the fmr1 À/y hippocampus compared to WT (Lee et al., 2011). It is possible that changes in Kv4.2 expression results in a compensatory increase in I h , in part by increased HCN1 expression. We are currently investigating these hypotheses as potential mechanisms underlying the change in dendritic function in FXS.
In summary we found that the dendrites of CA1 neurons from fmr1 À/y mice have altered intrinsic properties consistent with an elevation of I h . This elevation is due in part to the increased distal dendritic expression of the h-channel subunit HCN1. Furthermore, this elevation in dendritic I h appears to occlude normal homeostatic intrinsic plasticity of I h that occurs during LTP. These results suggest that altered dendritic processing and the potential for network instability may underlie some of the neurological deficits associated with FXS.

EXPERIMENTAL PROCEDURES Acute Hippocampal Slices and Electrophysiology
All experiments were conducted in accordance with the University's IACUC. Hippocampal slices (300 mm) were prepared from 2-to 3-month-old male WT and fmr1 À/y mice (C57BL/6) as described previously (Fan et al., 2005) (see Extended Experimental Procedures). Slices were placed in a holding chamber filled with ACSF containing (125 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 2 CaCl 2 , 2 MgCl 2 , and 12.5 dextrose, bubbled continuously with 95% O 2 /5% CO 2 ) warmed to 35 C for 30 min and then kept at RT. Hippocampal slices were placed into a submerged recording chamber and perfused with ACSF (as above except 3.0 KCl and 1.0 MgCl 2 ) at 31 C-33 C and viewed with a Zeiss Axioskop. For physiological measurements (Figure 1), 10 mM DNQX, 50 mM AP5, 2 mM SR95531, and 5 mM CGP52432 were present throughout. For mGluRs experiments (Figure 3), AP5 and 10 mM MK-801 were included in the ACSF (Brager and Johnston, 2007). For LTP experiments (Figure 4), GABA A -and GABA B -mediated IPSPs were blocked by SR95531 and CGP55845, and area CA3 was removed. All drugs were made from stock solutions in water or DMSO (final concentration of DMSO %0.1%). Drugs were obtained from Ascent Scientific (Bristol, UK).
Pipettes (4-6 MU) were pulled from borosilicate glass and filled with solution containing 120 mM K-gluconate, 20 mM KCl, 10 mM HEPES, 4 mM NaCl, 4 mM MgATP, 0.3 mM Na-GTP, and 7 mM K 2 -phosphocreatine (pH 7.3). Series resistance (R S ) was monitored throughout the recording, and experiments with R S >30 MU were discarded. EPSPs of 4-6mV were elicited using tungsten-stimulating electrodes placed <20 mm from the apical dendrite $100-150 mm from the soma. Action potentials were elicited by current injection into the soma (1-2 nA for 2 ms).
Data were sampled at 20 kHz, filtered at 3 kHz, and digitized by an ITC-18 interface. EPSPs were quantified by linear fit to the initial slope. H-sensitive parameters were measured and analyzed as described in the Extended Experimental Procedures.

Immunohistochemistry
Two to 3-month-old male fmr1 À/y and WT mice were anesthetized and perfusion fixed 4% paraformaldehyde (PFA) at 4 C. The brain was removed, postfixed in PFA for 2 hr at RT, and placed in 30% sucrose in PBS overnight at 4 C. A total of 50 mm near-horizontal thin sections containing the hippocampus were cut on a freezing microtome. Slices were washed and then incubated in a blocking buffer solution containing 5% Normal Goat Serum, 0.1% Triton X-100, and PBS (PBS-T), and then incubated in primary antibody (1:2,000 concentration in PBS-T) against either HCN1 or HCN2 overnight at 4 C. Sections were then washed with PBS then placed in PBS-T containing a secondary antibody overnight at 4 C. The slices were then washed in PBS and mounted on microscope slides. The sections were visualized on a fluorescent microscope; images were captured by CCD camera, and analyzed using ImageJ (National Institutes of Health). The intensity of staining was quantified by measuring the pixel intensity along a line perpendicular to the cell body axis in CA1 (Shin et al., 2008).

Western Blotting
Area CA1 was isolated from three to five individual hippocampal slices frozen on dry ice. Tissue was homogenized in lysis buffer containing protease inhibitors. Homogenates were centrifuged at 80 3 g for 10 min, resolved using SDS-PAGE, and transferred to nitrocellulose membranes. Membranes were blocked in blocking buffer (LI-COR) overnight at 4 C or for 1 hr at RT. Membranes were incubated for 2 hr at RT or overnight at 4 C in primary antibodies diluted in blocking buffer (gp a-HCN1, 1:2,000; gp a-HCN2, 1:2,000; mouse a-tubulin 1:20,000). Membranes were washed in PBS plus 0.05% Tween 20 (PBS+) and then incubated in species-appropriate secondary in PBS+ for 45 min at RT. Membranes were then washed in PBS+ followed by PBS and visualized on an Odyssey imaging system (LI-COR). Blots were analyzed using ImageJ. HCN protein levels were normalized to tubulin levels as a control for protein loading. Homogenates from four fmr1 À/y and four WT mice were used for four separate western blots.

SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and two figures and can be found with this article online at doi:10.1016/j.celrep. 2012.02.002.

LICENSING INFORMATION
This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 Unported License (CC-BY; http://creativecommons. org/licenses/by/3.0/legalcode).