Neurofibromin regulation of ERK signaling modulates GABA release and learning

Abstract

We uncovered a role for ERK signaling in GABA release, long-term potentiation (LTP), and learning, and show that disruption of this mechanism accounts for the learning deficits in a mouse model for learning disabilities in neurofibromatosis type I (NF1). Our results demonstrate that neurofibromin modulates ERK/synapsin I-dependent GABA release, which in turn modulates hippocampal LTP and learning. An Nf1 heterozygous null mutation, which results in enhanced ERK and synapsin I phosphorylation, increased GABA release in the hippocampus, and this was reversed by pharmacological downregulation of ERK signaling. Importantly, the learning deficits associated with the Nf1 mutation were rescued by a subthreshold dose of a GABA(A) antagonist. Accordingly, Cre deletions of Nf1 showed that only those deletions involving inhibitory neurons caused hippocampal inhibition, LTP, and learning abnormalities. Importantly, our results also revealed lasting increases in GABA release triggered by learning, indicating that the mechanisms uncovered here are of general importance for learning.

Figure 1. Cre-deletions of neurofibromin gene in inhibitory neurons cause learning deficits

A. Nf1^{syn I+/−} (n=13), Nf1^GFAP+/− (n=12) and their littermates controls (WT, n=16; Nf1^floxed/+, n=15; syn I-cre, n=14 and GFAP-Cre, n=11) were trained with 2 trails per day in the Morris water maze. A probe trial conducted after 7 days of training revealed that Nf1^{syn I+/−} mice showed no preference for the target quadrant, but Nf1^GFAP+/− mice and their littermates did (F _{(3, 48)} =0.045, p=0.987 and F _{(3, 41)} =7.873, p<0.001 for Nf1^{syn I+/−} and the Nf1^GFAP+/− mice, respectively; one-way ANOVA). Since the Nf1^{syn I+/−} and Nf1^GFAP+/− are in the same genetic background, the same group of WT and Nf1^floxed/+ mice (including littermates of Nf1^synI+/− and Nf1^GFAP+/− mice) were used as controls. B. The distribution of target quadrant occupancy during the day 7 probe trial showed that Nf1^Dlx5/6+/− (n=8) had no preference for the target quadrant, but Nf1^{αCaMKII+/−} (n=8) and their littermate controls (WT, n=8; Nf1^floxed/+, n=7; Dlx5/6-cre, n=7; and αCaMKII-Cre, n=8) searched selectively in the target quadrant(F _{(3, 28)} =12.102, p<0.001 and F _{(3, 28)} =0.458, p=0.7141; for Nf1^{αCaMKII+/−} and the Nf1^Dlx5/6+/− mice, respectively; one-way ANOVA). The same WT and Nf1^floxed/+ mice were used as controls because Nf1^Dlx5/6+/− and Nf1^{αCaMKII+/−} mutations were on the same genetic background. The figure shows % search times for each of the four quadrants: target (in black), adjacent left, opposite, and adjacent right quadrants (in that order). All statistical comparisons are presented in Table S2 and Table S3.

Figure 2. Increased mIPSC frequency in Nf1^+/−; mutants in high KCl

A. Representative traces of sIPSCs and mIPSCc recorded from CA1 pyramidal neurons of WT and Nf1^+/− mice with normal ACSF (2.5mM KCl). Calibration: 100 pA, 100 ms; B. Average of more than 100 events of sIPSCs or mIPSCs from WT or Nf1^+/− neurons. Calibration: 50 ms and 10pA. C and D: Frequency of mIPSC(e) and mEPSCs(f) recorded in normal and high K+ ACSF. The individual points represent recordings from different neurons. Solid black squares represent WT and open squares represent Nf1^+/− neurons. Solid lines (WT) and dashed lines (Nf1^+/−) represent the change ratio of the mIPSC or mEPSC frequency of neurons in 12.5 mM KCl versus 2.5 mM KCl ACSF. E and F. Comparison of normalized mIPSC (C) frequencies and mEPSC (D) in WT and Nf1^+/−; Data are presented as mean ± SEM. C. High KCl (12.5 mM) in ACSF resulted in a significant increase in mIPSC frequency in both WT and Nf1^+/− mice (WT: 13.89 ±0.69Hz vs. 21.98±0.95 Hz, n=30, paired t-test, t=8.306, P<0.001; Nf1^+/−: 16.26±0.65Hz vs. 43.21±1.11 Hz, n=29, t=36.525, p<0.001). E. The ratio of mIPSC frequency recorded in high KCl to that in control solution was significantly larger in Nf1^+/− mice (2.76 ± 0.34 n=29) than in their WT littermates (1.65 ± 0.229 n=30) (t=7.906; P<0.001). D. High KCl (12.5 mM) in ACSF increased the frequency of mEPSC in both WT (1.15± 0.14 Hz in normal ACSF and 2.12±0.17 in ACSF with 12.5 mM KCl, n=19, paired t-test, t=−13.585, p<0.001) and Nf1^+/− (1.25±0.104 Hz in normal ACSF and 2.25±0.16 in ACSF with 12.5 mM KCl, n=16, paired t-test, t=−7.906, p<0.001). F. The ratio of mEPSC frequency recorded in 12.5 mM KCl to that in control solution was not significantly different between WT and Nf1^+/− neurons (2.12±0.15, n=19 in WT and 2.03±0.24 n=16, in Nf1^+/−, Student’s t-test, t=0.294, p=0.7708). G and H: Deletion of the Nf1 gene from inhibitory neurons caused increased mIPSC frequency(G and H). G. Representative traces of mIPSCc recorded from CA1 pyramidal neurons of WT and Cre-mediated Nf1 mutants mice with normal ACSF (2.5mM KCl) and 12.5mM KCl ACSF. Calibration: 50 pA, 100 ms; H. The figure shows normalized frequency changes of miniature IPSCs recorded from CA1 pyramidal neurons in 12.5 mM KCl versus 2.5 mM KCl ACSF of different Nf1 mice with Cre-driven deletions (WT, n=20; Nf1^floxed/+, n=19; Syn-Cre, n=14; αCaMKII-Cre, n=18; Dlx5/6-Cre, n=13; Nf1^{syn I+/−}, n=15; Nf1^{αCaMKII+/−}, n=15; Nf1^Dlx5/6+/−, n=14) Nf1^{syn I+/−} and Nf1^Dlx5/6+/−, showed higher mIPSC frequencies than WT when the KCl concentration in ACSF was 12.5 mM KCl (WT: 13.52±0.35Hz vs. 24.31±0.71Hz; Nf1^{syn I+/−}: 15.27±0.62Hz vs. 41.26±0.37Hz; Nf1^Dlx5/6+/−: 12.46±0.42Hz vs. 35±0.51 Hz; the first value listed for each mutant line is for 2.5 mM and the second is for 12.5 mM KCl). In contrast, mIPSCs in Nf1^GFAP+/− and Nf1^{αCaMKII+/−} mice, as well as in all controls lines, did not differ from WT mice in either normal or 12.5 mM KCl ACSF

Figure 3. Increased mIPSCs frequency in Nf1^+/− mice is reversed by MEK inhibitors

A. Increased mIPSCs frequency caused by high K+ ACSF perfusion is reversed by the MEK inhibitor U0126 (normalized frequency: WT w/o U0126, 1.650.058, WT w/ U0126, 1.320±063; Nf1^+/− w/o U0126, 2.650α0.082, Nf1^+/− w/ U0126, 1.89±0.067. p>0.05 for WT w/ U0126 vs. Nf1^+/− w/ U0126); B. Increased mIPSCs frequency caused by high K+ ACSF perfusion is reversed by the MEK inhibitor PD 98095(normalized frequency: WT w/o PD98095, 1.790.057, WT w/PD98095, 1.52±0.069; Nf1^+/− w/o PD98095, 2.75±0.09, Nf1^+/− w/ PD98095, 1.58±0.047. p>0.05 for WT w/ PD98095 vs. Nf1^+/− w/ PD98095).

Figure 4. Increased ERK-dependent synapsin I phosphorylation in Cre-mediated Nf1 mutants

A. Nf1^+/− mice have deficits in the acquisition of contextual fear conditioning Nf1^+/− mice (n=16) and WT controls (n=14) were trained with a contextual fear conditioning protocol using one-trial per day for 5 consecutive days. The average freezing levels during the first 30 seconds of each training day and 24 hours after the last training trial were plotted. WT mice freeze significantly more compare to WT (F_(1,140)=3.927, P<0.05). Error bars indicate SEM. B. Contextual conditioning increases the phosphorylation of ERK and synapsin I (sites 4/5). Representative Western blots indicating protein bands visualized with antibodies to dually phosphorylated ERK1/2, total ERK1/2, synapsin I at sites 4/5, and total ERK1/2. + symbols denote contextual conditioning (shocks were delivered during the contextual exposure). - symbols denote that the no shock was delivered. C and D: Quantification of relative phosphorylated ERK1/2, synapsin I at sites 4/5; For each experiment, both phosphorylated and total MAPK levels were normalized to those observed in the control group of wild-type mice. 3–6 mice in each group. Values are mean ± SEM. E. Increased ERK phosphorylation in inhibitory neurons of Nf1^+/−and WT mice after Morris water maze training After spatial learning, double immunofluorescent staining shows phosphorylated ERK (Red) in inhibitory neurons (labeled with GAD65/67, Green) of both WT and Nf1^+/− mice. The arrows point to examples of neurons positive for both GAD65/67 and phosphorylated ERK in the hippocampal CA1 region of WT and Nf1^+/− mice. Scale bar, 50 µm. F. Quantification of the relative number of ERK phosphorylation positive inhibitory neurons in the CA1 region of Nf1^+/− mutants and WT controls. The overall number of ERK phosphorylation positive inhibitory neurons in the CA1 region is higher in Nf1^+/− mutant (6.2±2.2%, n=4) mice than in WT controls (3.0±1.3%, n=4, t-test, p=0.0414); after 5 training trials in the Morris water maze, the numbers of ERK phosphorylation positive inhibitory neurons in the CA1 region increased in both Nf1^+/− (36.2±9.21%, n=5) and WT (26.3±7.43%) mice.

Figure 5. Impaired hippocampal CA1 LTP in Nf1^Dlx5/6+/− but not Nf1^{αCaMKII+/−} mutants

LTP induced by a 5-theta-bursts tetanus (at 10 min). Each point indicates the field EPSP slope normalized to the average baseline response before the tetanus delivered at time 0. (Solid square, WT; Open circle, Nf1^Dlx5/6+/− and open triangle, Nf1^{αCaMKII+/−}). Between 40 and 50 min after the tetanus, the Nf1^Dlx5/6+/− mice (n = 6 mice) showed 124.5% ±4.5% potentiation, the Nf1^{αCaMKII+/−} mice (n = 9 mice) showed 142.1%± 6.7% potentiation, whereas WT mice (n = 6 mice) showed 148.6% ±4.1% (F_(1,10)= 18.713, P < 0.01; F_(1,13) =0.613, P=0.4477 for Nf1^Dlx5/6+/− and Nf1^{αCaMKII+/−} respectively when compared with WT mice, one way ANOVA)

Figure 6. Spatial learning and memory deficits of Nf1^+/− mice are reversed by a sub-threshold dose of picrotoxin

A. Mean percent of time spent in each quadrant during a probe trial. Analysis of the day 7 probe trial showed an interaction between genotype and treatment (F_(3,71) = 7.937, P<0.01, two-way ANOVA). 0.01mg/Kg picrotoxin did not affect the searching time of WT mice in the target quadrant (39.23±2.20% for WT treated with saline vs. 40.40±2.64% for WT treated with picrotoxin). Nf1^+/− mice treated with picrotoxin searched significantly longer in the training quadrant than Nf1^+/− mice treated with saline (27.94±2.05% for Nf1^+/− treated with saline vs. 36.83±2.19% for Nf1^+/− treated with picrotoxin), and were indistinguishable from WT mice with or without treatment. The figure shows % search times for each of the four quadrants: target (in black), adjacent left, opposite, and adjacent right quadrants (in that order). B. Number of times the mice crossed the exact position where the platform was during the training (solid bar), compared with the number of crossings in the opposite position in the pool (open bar). 0.01mg/Kg picrotoxin did not affect the proximity of WT mice to the target quadrant (2.68±043 times for WT treated with saline vs. 3.00α0.39 for WT treated with picrotoxin). Nf1^+/− mice treated with picrotoxin crossed the platform site more times than Nf1+/− mice treated with saline (1.11±0.30 for Nf1^+/− treated with saline vs. 2.89±0.63 for Nf1^+/− treated with picrotoxin), and were indistinguishable from WT mice with or without treatment. C. Average proximity to the exact position where the platform was during training (solid bar), compared with proximity to the opposite position in the pool (open bar). 0.01mg/Kg picrotoxin did not affect the proximity of WT mice in the target quadrant (42.82±1.84 cm in target quadrant for WT treated with saline vs. 45.44±1.75 cm in target quadrant for WT treated with picrotoxin). Nf1^+/− mice treated with picrotoxin searched significantly closer to the platform than Nf1^+/− mice treated with saline (53.69±2.32 cm for Nf1^+/− treated with saline vs. 44.67±2.15 cm for Nf1^+/− treated with picrotoxin), and were indistinguishable from WT mice with or without treatment. D. Distribution of performance of each group of mice in the probe trial. The percentage of mice spending less than 25%, between 25–45% and more than 45% of the time in the training quadrant, during the day 7 probe trial, is plotted for each group. (WT, n=19; Nf1^+/−, n=19; WT+PTX, n=19; Nf1^+/− +PTX, n=18.)

Figure 7. Spatial learning increases mIPSC frequency in hippocampal CA1 neurons

1 hr after water maze training, mIPSC frequency(A), amplitude(B), decay-time constant(C) and 10%–90% rise time(D) of the Learning group were compared with Swimming control and Cage control levels; (A) mIPSC frequency of the Learning group were significantly above control levels. (F_{(2, 44})=8.954, p < 0.001, one way ANOVA; **P<0.01, ***P<0.001); (B), (C) and (D): mIPSC amplitude(b), decay time constant(c) and 10%-90% rise time(d) of the Learning group were not significantly different from control levels.