.............................................................. Mechanism for the learning deficits in a mouse model of neurofibromatosis type 1 Rui M. Costa*, Nikolai B. Federov*†, Jeff H. Kogan*†, Geoffrey G. Murphy*, Joel Stern*, Masuo Ohno*, Raju Kucherlapati‡, Tyler Jacks§ & Alcino J. Silva* * Departments of Neurobiology, Psychiatry and Psychology, BRI, University of California at Los Angeles, Los Angeles, California 90095-1761, USA ‡ Department of Molecular Genetics, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York, New York 10461, USA § Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA ............................................................................................................................................................................. Neurofibromatosis type I (NF1) is one of the most common single-gene disorders that causes learning deficits in humans 1 . Mice carrying a heterozygous null mutation of the Nf1 gene (Nf1 +/2 ) show important features of the learning deficits associ- ated with NF1 (ref. 2). Although neurofibromin has several known properties and functions, including Ras GTPase-activat- ing protein activity 3,4 , adenylyl cyclase modulation 5,6 and micro- tubule binding 7 , it is unclear which of these are essential for learning in mice and humans. Here we show that the learning deficits of Nf1 +/2 mice can be rescued by genetic and pharmaco- logical manipulations that decrease Ras function. We also show that the Nf1 +/2 mice have increased GABA (g-amino butyric acid)-mediated inhibition and specific deficits in long-term potentiation, both of which can be reversed by decreasing Ras function. Our results indicate that the learning deficits associated with NF1 may be caused by excessive Ras activity, which leads to impairments in long-term potentiation caused by increased GABA-mediated inhibition. Our findings have implications for the development of treatments for learning deficits associated with NF1. Visual –spatial problems are among the most common cognitive deficits detected in individuals affected with NF1 (refs 1, 8). Studies have shown that Nf1 +/2 mice have abnormal spatial learning 2 when tested in the hidden version of the water maze — a task that is sensitive to hippocampal lesions 9 . Studies suggest that an upregula- tion of Ras activity may account for the learning deficits in NF1 both in mice 10 and humans 11 . To test this hypothesis, we crossed the Nf1 +/2 mice (C57B6/N background) with mice heterozygous for a null mutation in the K-ras gene 12 (K-ras +/2 , 129T2/SvEmsJ), and tested the F 1 descendants (hybrid isogenic background) in the hidden version of the water maze task. Because K-ras 2/2 mice, like Nf1 2/2 (ref. 13), die in utero 12 , we used K-ras +/2 mice, which are viable and show no apparent developmental defects. Mice were trained with two trials per day. In all the following experiments, no differences between genotypes and/or treatments were observed in acquisition (Fig. 1a, d, g), floating, thigmotaxic behaviour or swimming speed (data not shown). Spatial learning was assessed in a probe trial 14 (training day 7), in which the platform was removed from the pool. The time spent searching in the training quadrant was different among the different genotypes (analysis of variance, ANOVA, F 3,61 = 7.27, P , 0.05). Wild-type mice spent significantly more time searching in the training quadrant than did Nf1 +/2 mice (Fisher’s protected least significant difference, PLSD, P , 0.05; Fig. 1b), confirming that the Nf1 +/2 mice have impaired spatial learning. K-ras +/2 mice were also impaired (P , 0.05; Fig. 1b); however, mice carrying heterozygous mutations in both the Nf1 and the K-ras genes (Nf1 +/2 /K-ras +/2 ) spent as much time as wild- type mice in the training quadrant (P . 0.05; Fig. 1b), and significantly more time than Nf1 +/2 and K-ras +/2 mice (P , 0.05). This indicates that the learning deficits in Nf1 +/2 mice can be rescued by the K-ras +/2 mutation. Furthermore, both wild-type (paired t-test, t 23 = 5.935, P , 0.05) and Nf1 +/2 /K-ras +/2 mice searched selectively for the missing platform; that is, they searched closer to the exact platform position than to the opposite position in the pool 15 (t 10 = 4.09, P , 0.05; Fig. 1c), whereas Nf1 +/2 and K-ras +/2 mice did not (Nf1 +/2 , t 14 = 1.83; K-ras +/2 , t 14 = 0.149; P . 0.05). Mutations that decrease Ras function do rescue the deficits of the Nf1 +/2 mice because a null heterozygous mutation in N-ras (N-ras +/2 ), another mammalian ras gene with overlapping function and similar expression to that of K-ras 12 , also reverses the spatial learning deficits of the Nf1 +/2 mice (Fig. 1e, f; see Methods). During the probe trial, there was a significant effect of genotype on searching strategy (F 3,32 = 2.83, P , 0.05; Fig. 1e). Nf1 +/2 /N-ras +/2 mice were indistinguishable from wild type (Fisher’s PLSD, P . 0.05), and searched significantly more in the training quadrant than did Nf1 +/2 mice (P , 0.05). Also, analysis of proximity scores showed that both wild-type and Nf1 +/2 /N-ras +/2 mice searched selectively (wild type, t 9 = 3.621; Nf1 +/2 /N-ras +/2 , t 8 = 5.84; P , 0.05; Fig. 1f), whereas Nf1 +/2 mice did not (t 14 = 1.83, P . 0.05). The N-ras +/2 mutation by itself did not produce any detectable phenotype (P . 0.05; Fig. 1e), consistent with previous studies indicating that Ras signalling is less impaired in N-ras than in K-ras mutants 12 . The ras mutations that reversed the Nf1 +/2 spatial learning deficits should result in a decrease in Ras function throughout the life of the mice. To determine whether decreases in Ras function specifically during training could reverse the learning deficits of the Nf1 +/2 mice, we used a farnesyl-transferase inhibitor (FTI; BMS 191563; ref. 16). This agent reduces Ras signalling by blocking farnesylation, a post-translational modification that is essential for Ras function 17 . FTIs have been shown to rescue other NF1 pheno- types, such as Schwann cell proliferation in human and murine cells 18,19 . Wild-type and Nf1 +/2 mice were injected intraperitoneally every day 60 min before training with either saline or 5 mg per kg (body mass) BMS 191563; the training conditions and the genetic background of the animals were the same as in the Nf1 +/2 /K-ras +/2 experiment. Analysis of the day 7 probe trial showed an effect of genotype– treatment interaction on searching strategy (F 3,32 = 2.83, P , 0.05; Fig. 1h). Nf1 +/2 mice treated with BMS191563 searched signifi- cantly longer in the training quadrant than Nf1 +/2 mice treated with saline (P , 0.05), and were indistinguishable from wild-type mice (P . 0.05). Moreover, analysis of proximity scores showed that Nf1 +/2 mice treated with BMS191563 searched selectively (t 18 = 4.13, P , 0.05; Fig. 1i), as did wild-type mice (t 17 = 4.33, P , 0.05), whereas Nf1 +/2 mice treated with saline did not (t 19 = 0.266, P . 0.05). Administration of 5 mg per kg BMS191563 did not adversely affect wild-type animals, as these mice searched equiva- lently to wild-type mice treated with saline (P . 0.05; Fig. 1h). The finding that the learning deficits of Nf1 +/2 mice can be reversed with an FTI confirms our genetic results, and shows that these learning deficits are reversible in adult mice. Learning is thought to occur through activity-dependent synap- tic modifications in neuronal networks 20 . To determine whether changes in synaptic plasticity might account for the spatial (hippo- campal-dependent 9 ) learning impairments of Nf1 +/2 mutants, we examined long-term potentiation (LTP) at Schaffer collateral/CA1 This advance online publication (AOP) Nature paper should be cited as “Author(s) Nature advance online publication, 16 January 2002 (DOI 10.1038/nature711)”. Once the print version (identical to the AOP) is published, the citation becomes “Author(s) Nature volume, page (year); advance online publication, 16 January 2002 (DOI 10.1038/nature711)”. nature † Present address: Memory Pharmaceuticals Corporation, 100 Philips Parkway, Montvale, New Jersey 07645, USA advance online publication letters to nature NATURE | 16 January 2002 | DOI 10.1038/nature711 | www.nature.com 1 © 2002 Macmillan Magazines Ltd