HCN channels are a novel therapeutic target for cognitive dysfunction in Neurofibromatosis type 1

Abstract

Cognitive impairments are a major clinical feature of the common neurogenetic disease neurofibromatosis type 1 (NF1). Previous studies have demonstrated that increased neuronal inhibition underlies the learning deficits in NF1, however, the molecular mechanism underlying this cell-type specificity has remained unknown. Here, we identify an interneuron-specific attenuation of hyperpolarization-activated cyclic nucleotide-gated (HCN) current as the cause for increased inhibition in Nf1 mutants. Mechanistically, we demonstrate that HCN1 is a novel NF1-interacting protein for which loss of NF1 results in a concomitant increase of interneuron excitability. Furthermore, the HCN channel agonist lamotrigine rescued the electrophysiological and cognitive deficits in two independent Nf1 mouse models, thereby establishing the importance of HCN channel dysfunction in NF1. Together, our results provide detailed mechanistic insights into the pathophysiology of NF1-associated cognitive defects, and identify a novel target for clinical drug development.

Figure 1

Characterization of the Nf1^9a−/9a− mutant. (a) Results of the reverse transcription-PCR in cortex and hippocampus in WT and Nf1^9a−/9a− mice showing loss of Nf1 exon 9a-containing transcripts in Nf1^9a−/9a− mice (representative image, quantitative PCR experiment performed on 7 WT and 9 mutant mice). (b) Western blots of cortical and hippocampal lysates of the WT and Nf1^9a−/9a− mice showing reduction of total neurofibromin levels in the mutants (representative image, experiment performed on 10 WT and 11 mutant mice). (c) Nf1 mRNA (Nf1^9a−/9a−: n = 9; WT: n = 7, P < 0.01) and neurofibromin protein (Nf1^9a−/9a−: n = 11; WT: n = 10, P < 0.001) in the hippocampus of WT and Nf1^9a−/9a− mice. (d) In the water maze probe trial, Nf1^9a−/9a− mice show no significant spatial learning, while WT mice show a clear preference for the target quadrant (TQ, black column) (Nf1^9a−/9a−: n = 15, P = 0.6; WT: n = 14, P < 0.001, paired t-test between target and average of other quadrants). Bars represent target, adjacent right, opposite, and adjacent left quadrants, respectively. The heat-plots are a visual representation of all search tracks, in which the color indicates the mean time spent at a certain position. The black circle indicates the target platform position used during training. (e) LTP deficit at Schaffer collateral-CA1 synapses in Nf1^9a−/9a− mice induced by TBS (WT: n = 14 and Nf1^9a−/9a−: n = 15; P < 0.02) but not by an (f) High frequency stimulation (WT: n = 13 and Nf1^9a−/9a−: n = 15, P = 0.8). (g) Picrotoxin (PTX) restores TBS-LTP in Nf1^9a−/9a− mice (WT: n = 14 and Nf1^9a−/9a−: n = 17, P = 0.9). Insets are field excitatory postsynaptic potentials (fEPSPs) in WT and Nf1^9a−/9a− mice taken at the points shown in each graph. Calibration: 5 ms and 0.2 mV. Statistical analyses were performed by Student’s t-test. *P < 0.05, **P < 0.01. Data represent the mean ± s.e.m.

Figure 2

Enhanced inhibition in Nf1^9a−/9a− mice. (a) Evoked inhibitory postsynaptic potentials (IPSPs) show larger amplitude in mutants (two-way analysis of variance (ANOVA), F_1,16 = 5.3; P < 0.05; n = 9 for both groups) and paired-pulse ratio is reduced in Nf1^9a−/9a− compared with WT controls (0.6 ± 0.04, n = 10 vs 0.8 ± 0.05, n = 7; P < 0.02). (b) Representative traces (top) and cumulative distributions of miniature inhibitory postsynaptic current (mIPSC) inter-event intervals (bottom) showing a significant leftward shift of the curve in Nf1^9a−/9a− pyramidal neurons under high KCl (Kolmogorov–Smirnov test, P < 0.001), but not with normal KCl (Kolmogorov–Smirnov test, P = 0.3). (c) Group data of mIPSC frequency in baseline and high KCl conditions. High KCl results in a significant increase in mIPSC frequency for both WT (6 ± 0.5 Hz vs 9 ± 0.9 Hz, n = 12; P < 0.004) and Nf1^9a−/9a− mice (6 ± 0.5 Hz vs 12 ± 1.0 Hz, n = 11; P < 0.0005). Overall, the ratio of mean frequency of the events in high KCl to that in baseline condition is higher in mutants (P < 0.001), whereas the mean amplitude remains unchanged. (d) Evoked excitatory postsynaptic potentials (EPSPs) show no difference in amplitude (two-way ANOVA, F_1,17 = 0.03; P = 0.9, WT: n = 9, Nf1^9a−/9a−: n = 10) or paired-pulse ratio (Nf1^9a−/9a−: 1.5 ± 0.5, n = 10; WT: 1.5 ± 0.1, n = 9; P = 0.8). (e) Representative traces (top) and cumulative distributions of miniature excitatory postsynaptic current (mEPSC) inter-event intervals (bottom) for which there is no significant difference in the cumulative distribution during baseline or high KCl conditions (P = 0.35). (f) Group data of mEPSC frequency in baseline and high KCl conditions (WT: 0.9 ± 0.1 Hz vs 1.26 ± 0.1 Hz, n = 9, P < 0.05; Nf1^9a−/9a−: 0.8 ± 0.1 Hz, vs 1.25 ± 0.1 Hz, n = 10, P < 0.05). The ratio of mean frequency of mEPSC in high KCl to that in baseline conditions is unchanged (P = 0.9). Calibration: 500 ms and 100 pA. *P < 0.05, **P < 0.01. Data represent the mean ± s.e.m.

Figure 3

Hyperpolarization-activated cyclic nucleotide-gated 1 (HCN1) is identified as an Neurofibromatosis type 1 (NF1)-interacting protein and enriched in parvalbumin (PV) neurons. (a) Immunoprecipitation (IP) using an antibody directed against neurofibromin on hippocampal lysates from WT, Hcn1^−/− and Nf1^9a−/9a− mice (representative image; IP on WT and Nf1^9a−/9a− lysates has been successfully repeated four times, and two times on Hcn1^−/− lysates. Immunoblotting analysis was performed using the antibodies as indicated (left panel). HCN1, HCN2 and TRIP8b were all co-precipitated with NF1 in the WT and Nf1^9a−/9a− samples. In the absence of HCN1 (Hcn1^−/−), the NF1 antibody is unable to co-immunoprecipitation (co-IP) HCN2 and TRIP8b. Right panel: hippocampal lysates used as input for IP experiments, showing normal HCN1, HCN2 and TRIP8b expression in Nf1^9a−/9a− mice. (b) Co-IP using a monoclonal antibody against HCN1 on lysates made from HEK293T cells co-transfected with constructs expressing HCN1 and the N-terminal half of NF1 (with or without exon 9a) fused to GFP (representative image; transfection and IP were repeated two times). Immunoblot analysis was performed with antibodies directed against HCN1 and GFP. (c and d) Immunofluorescence co-labeling of PV (red) and NF1 (green) (c), or PV (red) and HCN1 (green) (d) in hippocampal sections of WT mice, showing high expression of NF1 and HCN1 in PV-expressing neurons.

Figure 4

I_h is selectively attenuated in Nf1^9a−/9a− PV neurons. (a) Top: PV immunoreactivity of a biocytin-filled interneuron shown by fluorescent double labeling (representative image, repeated for each included interneuron). Bottom: Representative voltage responses of PV and pyramidal neurons to depolarizing and hyperpolarizing current injections. (b) Sample voltage–clamp recordings from a PV neuron and a pyramidal neuron. The ZD7288-sensitive I_h currents were obtained by subtracting current traces before and after application of (30 µm) ZD7288. Currents were evoked by 800 ms hyperpolarizing voltage steps from a holding potential of −50 to −120 mV in 10 mV increments. (c) The current–voltage (I–V) relationship of I_h is significantly shifted toward smaller I_h in Nf1^9a−/9a− PV neurons (F_1,16 = 4.1, P < 0.05, n = 7 and n = 11 for WT and Nf1^9a−/9a−, respectively). (d) The activation curve of I_h obtained by plotting the normalized tail current amplitude as a function of the voltage step and fitted with a Boltzmann function (solid lines). There was a significant hyperpolarizing shift in the voltage-dependent activation of I_h in Nf1^9a−/9a− PV neurons (V_1/2 in WT was −81.2 ± 2.0 and −88.4 ± 1.9 in Nf1^9a−/9a−, P < 0.01). (e) The mean number of action potentials generated in response to depolarizing current pulses in the PV neurons from Nf1^9a−/9a− is significantly higher (F_1,22 = 4.4, P < 0.05). (f) No significant differences were found in the I–V relationship (F_1,19 = 1.4, P = 0.2) or in (g) the voltage dependence of I_h activation of pyramidal neurons between genotypes (P = 0.3; n = 9 and n = 10 for WT and Nf1^9a−/9a−, respectively). (h) Excitability of pyramidal neurons is unchanged in Nf1^9a−/9a− mice (F_1,32 = 0.1, P = 0.8; n = 16 and n = 18 for WT and Nf1^9a−/9a−, respectively). Statistical analyses were performed by Student’s t-test and two-way ANOVA followed by Tukey's test. *P < 0.05, **P < 0.01. Data represent the mean ± s.e.m.

Figure 5

Lamotrigine (LTG) rescues the electrophysiological and behavioral phenotypes in Nf1^9a−/9a− mice. (a) Left: I_h amplitude is significantly increased by LTG in Nf1^9a−/9a− mice (F_1,21 = 9.9, P < 0.05, n = 12). Right: LTG shifts I_h activation to more depolarized potentials in PV neurons from Nf1^9a−/9a− mice. Control data are the same as shown in Figures 4c and d. Dashed line shows activation curve for wild-type (WT) littermates. (b) LTG decreases membrane excitability in PV neurons from Nf1^9a−/9a− mice and this effect is blocked by ZD7288. Left: sample recordings of APs elicited by a depolarizing current step before and after application of LTG in the absence and presence of ZD7288. Right: summary graph of change in firing rates caused by LTG, in the absence and presence of ZD7288, as a function of injected current. (c) Left: LTP at SC-CA1 synapses is impaired in vehicle-treated slices from Nf1^9a−/9a− mice (n = 9) compared with WT mice (n = 6, P < 0.01). Right: LTG rescues the LTP deficit in Nf1^9a−/9a− mice (n = 10 and n = 16 for WT and Nf1^9a−/9a−, respectively, P = 0.2). (d) LTG rescues the learning deficits in Nf1^9a−/9a− mice in a water maze probe trial, (two-way analysis of variance (ANOVA), genotype × drug interaction, F_1,43 = 6.0, P < 0.05). The black and white columns represent the target quadrant (TQ) and average time spent in other quadrants, respectively. Although vehicle-treated Nf1^9a−/9a− mice (n = 9) spend only 27% of their time in the TQ, LTG-treated Nf1^9a−/9a− mice (n = 12) show 45% TQ preference (P < 0.05). LTG has no significant effect on spatial learning in WT mice (vehicle, n = 12 and LTG, n = 14, P = 0.3). (e) LTG rescues the spatial learning deficits of Nf1^+/− mutants (two-way ANOVA, genotype × drug interaction, F_1,40 = 5.2, P < 0.05). Vehicle-treated Nf1^+/− mice show no preference for the TQ (25%, n = 9), whereas LTG-treated Nf1^+/− mice show robust learning (44%, n = 10; P < 0.05). LTG has no significant effect on WT mice (Vehicle, n = 13, and LTG, n = 12; P = 0.4). The heatplots are visual representation of all search tracks, in which the color indicates the mean time spent at a certain position. Statistical analyses were performed by Student’s t-test and two-way ANOVA followed by Tukey's test. *P < 0.05. Data represent the mean ± s.e.m.