Neurofibromatosis type 1 alternative splicing is a key regulator of Ras/ERK signaling and learning behaviors in mice

Abstract

Appropriate activation of the Ras/extracellular signal-regulated kinase (ERK) protein signaling cascade within the brain is crucial for optimal learning and memory. One key regulator of this cascade is the Nf1 Ras GTPase activating protein (RasGAP), which attenuates Ras/ERK signaling by converting active Ras is bound to guanosine triphosphate, activating Ras into inactive Ras is bound to guanosine diphosphate, inactivating Ras. A previous study using embryonic stem cells and embryonic stem cell-derived neurons indicated that Nf1 RasGAP activity is modulated by the highly regulated alternative splicing of Nf1 exon 23a. In this study, we generated Nf123aIN/23aIN mice, in which the splicing signals surrounding Nf1 exon 23a were manipulated to increase exon inclusion. Nf123aIN/23aIN mice are viable and exon 23a inclusion approaches 100% in all tissues, including the brain, where the exon is normally almost completely skipped. Ras activation and phosphorylation of ERK1/2 downstream of Ras are both greatly increased in Nf123aIN/23aIN mouse brain lysates, confirming that exon 23a inclusion inhibits Nf1 RasGAP activity in vivo as it does in cultured cells. Consistent with the finding of altered Ras/ERK signaling in the brain, Nf123aIN/23aIN mice showed specific deficits in learning and memory compared with Nf1+/+ mice. Nf123aIN/23aIN mice performed poorly on the T-maze and Morris water maze tests, which measure short- and long-term spatial memory, respectively. In addition, Nf123aIN/23aIN mice showed abnormally elevated context-dependent fear and a diminished ability to extinguish a cued fear response, indicating defective associative fear learning. Therefore, the regulated alternative splicing of Nf1 is an important mechanism for fine-tuning Ras/ERK signaling as well as learning and memory in mice.

Figure 1

Mutations designed to increase Nf1 exon 23a inclusion do not affect mouse viability. (A) Mouse pups were genotyped by PCR. (B) The genotypes of weaning-age offspring from crosses between Nf1^23aIN/+ mice. Total numbers of mice are shown along with percent of total mice in parentheses; χ² test: P=0.95.

Figure 2

Nf1^23aIN mutations increase endogenous Nf1 exon 23a inclusion in mouse tissues. RT-PCR showing endogenous Nf1 exon 23a inclusion in mouse tissues. PCR primers are located on surrounding exons. Error bars represent standard error. N ≥ 3. Brain: *P = 7.3 × 10⁻⁸, **P = 1.6 × 10⁻⁸; Heart: *P = 1.0 × 10⁻², **P = 1.5 × 10⁻²; Lung: *P = 1.0 × 10⁻³, **P = 1.5 × 10⁻⁴; Kidney: *P = 2.8 × 10⁻⁴, **P = 5.9 × 10⁻⁵; Liver: *P = 2.1 × 10⁻⁵, **P = 1.0 × 10⁻⁵; Spleen: *P = 5.4 × 10⁻⁴, **P = 1.2 × 10⁻⁵; Muscle: *P = 5.8 × 10⁻³, **P = 1.1 × 10⁻⁴; Testes: *P = 1.3 × 10⁻², **P = 4.5 x10 ⁻⁶.

Figure 3

Increased endogenous Nf1 exon 23a inclusion leads to increased Ras activation in mouse brain. (A) GST-Raf1RBD was used to pull down active Ras (Ras-GTP) from mouse brain lysates, followed by western blot analysis using anti-Ras antibody. Mock pull-downs were performed using GST. Four percent of total lysate was loaded in the total Ras lanes. U1 70K is a loading control. (B) Western blot analysis showing Nf1 protein expression in mouse brain lysates. U1 70K is a loading control.

Figure 4

Western blot analysis showing the expression of downstream targets of Ras in lysates from whole mouse brains. (A) Phospho-ERK1/2 (∼42/44 kDa) and total ERK1/2 (∼42/44 kDa). Error bars represent standard errors. N = 3. *P = 1 × 10⁻², **P = 5.85 × 10⁻⁶. (B) Phospho-Akt (∼60 kDa) and total Akt (∼60 kDa). Error bars represent standard errors. N = 3. (C) Phospho-S6 (∼32 kDa), total S6 (∼32 kDa) and γ-tubulin (∼48 kDa, a loading control). Error bars represent standard errors. N = 3.

Figure 5

Spatial learning and memory tests; n = 9 for each genotype. Error bars represent standard error. (A) T-maze test. Mice were trained for 10 min in a T-maze where one of the arms was blocked and were then tested after 2 h in the same T-maze, except all arms were open. The percent of time spent exploring the previously blocked arm was measured (unfamiliar arm choice %). The dotted line indicates the unfamiliar arm choice percentage expected by chance. *Nf1^23aIN/+ and Nf1^23aIN/23aIN were significantly lower than Nf1^+/+ (F_2,27= 3.82, P = 0.0346). (B and C) Morris water maze test. Four trials were performed each day for four days. (B) Latency to reach the hidden platform was recorded (Gene: F_2,27 = 3.302, P = 0.0521, Days: F_3,81 = 50.209, P < 0.0001, Gene×Days: F_6,81= 4.35, P = 0.0007]. *Day 3: Nf1^23aIN/23aIN was significantly higher than Nf1^+/+ (P < 0.05), Day 4: Nf1^23aIN/23aIN was significantly higher than Nf1^+/+ (P < 0.01). There was no significant difference between genotypes on Day 1 or Day 2. (C) Following the final session, the platform was removed and the animals were allowed to swim for 60 s. The percent time spent in the quadrant where the platform was previously located was measured. *Nf1^23aIN/23aIN was significantly lower than Nf1^+/+ (F_2,27= 3.945, P = 0.0314).

Figure 6

Fear learning tests. Error bars represent standard error. (A) Mice were trained to associate a conditioned stimulus (CS, a 30-s-long pure tone of 5 kHz and 80 dB) with an unconditioned stimulus (US, a one second long electrical shock of 0.5 mA), over four trials [Gene: F_2,27 = 0.903, P = 0.4171, Trial: F_4,108 = 91.114, P < 0.0001, Gene×Trial: F_8,108 = 0.794, P = 0.609]. Freezing was significantly higher than baseline for CS 2, CS 3 and CS 4 (Ps < 0.01). There was no significant difference in freezing between genotypes. (B) Twenty-four hours later, mice were evaluated for their learned aversion for an environment associated with the shock (context-dependent fear), as measured by the percent time freezing in the environment where they had previously been shocked. *Nf1^23aIN/+ and Nf1^23aIN/23aIN were significantly higher than Nf1^+/+ (F_2,27 = 12.997, P = 0.0001). (C) Two hours later the animals were reintroduced into a contextually altered box, the same tone (30 s, 5 kHz, 80 dB) was delivered 10 times without an electrical shock, and freezing behavior was measured [Gene: F_2,27= 1.30, P = 0.289, Block: F_4,108 =10.273, P < 0.0001, Gene×Block: F_8,108 = 2.934, P = 0.0053]. *Block 4 and 5: Nf1^23aIN/+ and Nf1^23aIN/23aIN were significantly higher than Nf1^+/+ (Ps < 0.05).