Mechanisms underlying cognitive deficits in a mouse model for Costello Syndrome are distinct from other RASopathy mouse models

RASopathies, characterized by germline mutations in genes encoding proteins of the RAS-ERK signaling pathway, show overlapping phenotypes, which manifest themselves with a varying severity of intellectual disability. However, it is unclear to what extent they share the same downstream pathophysiology that underlies the cognitive deficits. Costello syndrome (CS) is a rare RASopathy caused by activating mutations in the HRAS gene. Here we investigated the mechanisms underlying the cognitive deficits of HRas G12V/G12V mice. HRas G12V/G12V mice showed robust upregulation of ERK signaling, neuronal hypertrophy, increased brain volume, spatial learning deficits, and impaired mGluR-dependent long-term depression (LTD). In contrast, long-term potentiation (LTP), which is affected in other RASopathy mouse models was unaffected. Treatment with lovastatin, a HMG-CoA-Reductase inhibitor which has been shown to rescue the behavioral phenotypes of mouse models of NF1 and Noonan syndrome, was unable to restore ERK signaling and the cognitive deficits of HRas G12V/G12V mice. Administration of a potent mitogen-activated protein kinase (MEK) inhibitor rescued the ERK upregulation and the mGluR-LTD deficit of HRas G12V/G12V mice, but failed to rescue the cognitive deficits. Taken together, this study indicates that the fundamental molecular and cellular mechanisms underlying the cognitive aspects of different RASopathies are remarkably distinct, and may require disease specific treatments.


Results
Increased RAS-ERK signaling in HRas G12V/G12V mice. To better understand the role of HRAS protein in neuronal functioning, we studied a mouse model for Costello syndrome (CS) 16 . This mouse carries a germline targeted point mutation introduced into the HRAS gene: glycine (G) to valine (V) at amino acid 12 (G12V), leading to expression of a constitutively active HRAS protein (from here referred to as HRAS G12V protein or HRas G12V/ G12V mice). The G12V mutation has been found in a number of CS patients and is associated with very severe disease manifestations 14,18,19 . As described previously, mutant HRas G12V/G12V mice were born at the expected Mendelian ratio, were fertile, survived at rates comparable to those of their wild-type littermates, and displayed several of the phenotypic abnormalities observed in CS patients, including facial dysmorphia 16,17 . In examining protein expression, we observed a 50% decrease in HRAS protein in hippocampal lysates of HRas G12V/G12V mice (t 8 = 5.34, P < 0.01; Fig. 1A,B), which was confirmed by immunohistochemistry (Fig. 1C). Hence, the level of HRAS G12V protein in homozygous HRas G12V/G12V mice might resemble the HRAS G12V protein level of heterozygous patients.
To evaluate whether constitutively active HRAS leads to hyperactivation of the RAS-ERK pathway, we assessed the level of ERK phosphorylation (pERK) in hippocampal lysates of adult HRas G12V/G12V mice. Immunoblotting revealed a significant increase in phosphorylated ERK isoforms ERK1 (pERK 44) and ERK2 (pERK 42) (t 8 = −5.92, P < 0.01, whereas total ERK was similar between the genotypes (Fig. 1A,B). In accordance with this result, immunohistochemistry revealed increased amounts of pERK throughout the HRas G12V/G12V brain, particularly in the hippocampus and cortex (Fig. 1D,E).
Hypertrophy of brain and pyramidal neurons in HRas G12V/G12V mice. Similar to the previous descriptions of the CS mouse model, we observed that most HRas G12V/G12V mice could be clearly distinguished from their wild-type (WT) littermates by their blunt nose and prominent forebrain 16,17 . After careful dissection of the brain from the skull, we observed a difference in the overall shape of the brain of HRas G12V/G12V mice ( Fig. 2A). Brain weight was increased by 22% in HRas G12V/G12V mice (F 1,11 = 31.59, P < 0.01; Fig. 2B), while total body weight was unchanged (F 1,11 = 0.15, P = 0.71). Measurement of the surface area of several brain regions on level-matched coronal brain sections revealed that HRas G12V/G12V mice had a significantly larger surface area of cortex (t 6 = −13,05, P < 0.01), striatum (t 6 = −10.27, P < 0.01), and corpus callosum (t 6 = −12.49, P < 0.01, Fig. 2D). Moreover, both the cortex (t 6 = −4.05, P < 0.01) and corpus callosum (t 6 = −2.56, P < 0.05) were significantly thicker in HRas G12V/G12V mice (Fig. 2C,E). In addition, analysis of neuronal cell size was performed in the deep lamina of the S1 somatosensory cortex in transverse sections stained for DAPI, NeuN and FOXP2, which is expressed by a subset of neurons in layer V and VI. The diameter of both NeuN+FOXP2 negative and NeuN+FOXP2 positive neurons was significantly larger in HRas G12V/G12V mice (upper-layer neurons: t 6 = −3.58, P < 0.05; deep-layer neurons t 6 = −4.76, P < 0.01; Fig. 2F-H). These results are in agreement with previous observations from a transgenic HRas G12V mouse model, which showed increased brain volume and larger pyramidal cells in layers V-VI of the motor and somatosensory cortex in adult mice 20,21 .
Scientific RepoRts | 7: 1256 | DOI: 10.1038/s41598-017-01218-0 Learning impairment in HRas G12V/G12V mice in the Morris water maze (MWM). Learning deficits are among the most frequent cognitive limitations observed in CS patients 22 as well as other RASopathies 3 , and are also observed in mouse models of CS, NF1, NS, and Legius syndrome. In particular, impairments of hippocampus-dependent spatial learning in the Morris water maze (MWM) have been reported in every RASopathy mouse model examined thus far (Table 1). During the MWM training, mice learn to use visual cues positioned outside a pool to find a hidden escape platform submerged beneath the surface of the opaque water. Previously, it has been shown that HRas G12V/G12V mice in the mixed 129-C57BL/6 background showed a learning deficit in the MWM test 17 . However, it has been frequently been observed that behavioral phenotypes are lost upon backcrossing. In fact expression of NF1, the most important regulator of RAS, is 30% lower in animals on the 129 strain compared to the animals on the C57BL/6 strain 23 , and the learning deficits in NF1 mice are exclusively observed in the mixed (F1) 129-C57BL/6 background 8,24,25 . Hence, to determine whether the MWM phenotype of HRas G12V/G12V mice is still present upon backcrossing the mutants for over 20 generations in the C57BL/6 background, we repeated the MWM experiment. We observed that both genotypes showed a reduction in their latency to find the hidden platform, but HRas G12V/G12V mice exhibited significantly longer latencies compared to wild-type mice (between subject comparison day 1-9: F 1,30 = 12.78, P < 0.01; day 10-14: F 1,15 = 6.53, P < 0.05; Fig. 3A). There were no significant differences in swimming speed (between subject comparison, effect of genotype: F 1,30 = 0.28, P = 0.60) or thigmotaxis (F 1,30 = 190, P = 0.18).
After 5 days of hidden platform training, the platform was removed and mice were subjected to a probe trial to test spatial memory. Wild-type mice demonstrated a focused search pattern with the majority of their time spent in close proximity to the trained position of the platform (target quadrant (TQ) versus other quadrants; t 16 = −3.46, P < 0.01). In contrast, HRas G12V/G12V mice showed no significant preference for the target quadrant compared to the other three quadrants and were significantly different from wild-type mice (TQ vs. other a significant increase in pERK (t 8 = −5.92, P < 0.01) and a significant decrease in expression of HRAS protein in HRas G12V/G12V mice compared to their wild-type littermates (t 8 = 5.34, P < 0.01). Data are presented as mean ± SEM of percentage of WT level set as 100% from 5 mice per genotype (n = 5/genotype). Statistical test: unpaired two-tailed t-test, *P < 0.05, **P < 0.01. (C) Visualisation of β-galactosidase activity of the IRES-β-geo cassette integrated into the mutated HRAS gene (LacZ; blue) combined with an HRAS staining (DAB; brown) indicating successful recombination and expression of HRAS G12V , however lower amounts of HRAS protein are seen in HRas G12V/G12V mice compared to WT littermates. Right panel: Zoomed-in image of hippocampus. (D,E) DAB staining using pERK 42/44 and ERK 42/44 antibodies revealed increased levels of ERK phosphorylation in HRas G12V/G12V mouse brain slices compared to WT (D), but no difference in the level of expression and distribution of total ERK between genotypes (E).
In order to assess the cognitive flexibility of HRas G12V/G12V mice, the platform was then relocated to the opposite quadrant for 5 additional training days, beginning on day 10. A probe trial on day 14 revealed that wild-type mice had successfully learned the new platform position (TQ vs. OQ: t 9 = −3.89, P < 0.01), whereas HRas G12V/G12V mice again showed no preference for the target quadrant (TQ vs. OQ: t 7 = −1.18, P = 0.14; between subject    : P = 0.14; between subject comparison: effect of genotype P = 0.17). Data are presented as mean ± SEM from wild-type n = 17 mice, HRas G12V/G12V n = 15 mice (for A,B) and from wild-type n = 10 mice, HRas G12V/G12V n = 8 mice (for C). Statistical test: repeated measure one-way MANOVA, paired two-tailed t-test and one-way MANOVA. Arrows in A indicate day of probe trials. comparison: F 1,16 = 2.01, P = 0.17; Fig. 3A,D). Therefore, we conclude that HRas G12V/G12V mice have reduced spatial learning abilities, which may also impair cognitive flexibility (reversal MWM phenotype).

Baseline synaptic transmission and LTP are unaffected in HRas G12V/G12V mice.
To determine whether changes in synaptic function account for the spatial learning impairments in HRas G12V/G12V mice, we measured baseline synaptic transmission using whole-cell recordings from hippocampal CA1 pyramidal neurons. No significant differences were observed in the frequency or amplitude of spontaneous inhibitory or excitatory postsynaptic currents (sIPSCs: frequency: t 44  The underlying mechanism for the cognitive deficits in the NF1 RASopathy mouse model has been previously shown to result from a shift in the excitatory/inhibitory balance, in which inhibitory synaptic transmission is abnormally enhanced during high frequency firing 24,25 . Therefore, in order to investigate if a similar mechanism might be present in HRas G12V/G12V mice, we recorded spontaneous miniature inhibitory postsynaptic currents (mIPSCs) in the presence of depolarizing levels of extracellular potassium (increased from 2.5 to 12.5 mM KCl), as previously shown for Nf1 mice 24,25 . As expected, high extracellular potassium (12.5 mM) resulted in a significantly increased frequency of mIPSCs in both HRas G12V/G12V and WT mice (frequency: F 1,49 = 22.28, P < 0.05, amplitude: F 1,49 = 1.21, P = 0.28; Fig. 4H,I). However, no differences in frequency or amplitude were observed between genotypes (effect of the genotype at 2.5 mM KCl condition frequency: F 1,49 = 1.07, P = 0.30, amplitude: To further examine whether the HRas G12V/G12V mutation and the resulting increase of pERK signaling might affect short-or long-term forms of synaptic plasticity, we recorded field excitatory postsynaptic potentials (fEPSPs) from hippocampal Schaffer collateral (SC)-CA1 synapses. Consistent with the whole-cell data, baseline synaptic transmission (defined by the relationship between the postsynaptic fEPSP slope and presynaptic fiber volley amplitude) was similar between HRas G12V/G12V mice and their WT littermates (fiber volley: F 1,53 = 0.15, P = 0.70, fEPSP slope: F 1,53 = 17, P = 0.68; Fig. 5A). Furthermore, paired-pulse facilitation (PPF), which represents a presynaptically-mediated short-term form of plasticity, was also normal in HRas G12V/G12V mice (F 1,14 = 0.36, P = 0.56, Fig. 5B). Since long-term plasticity (LTP) induced by theta-burst stimulation (TBS) is impaired in Nf1 8, 25 , Ptpn11 10 , Spred1 mice 11 and conditional Kras G12V mice 26 (see Table 1), we tested LTP induced by a theta-burst stimulation protocol and a high frequency (100 Hz) stimulation protocol. Notably, both TBS-induced LTP and 100Hz-induced was similar between genotypes (TBS: F 1,20 = 0.92, P = 0.51; 100 Hz: F 1,14 = 0.80, P = 0.61; Fig. 5C,D). Taken together, these findings suggest that although constitutively active HRAS impairs spatial learning, it does not affect hippocampal synaptic transmission nor short-or long-term synaptic potentiation.
Impaired mGluR-LTD in HRas G12V/G12V mice. Metabotropic glutamate receptor-dependent long-term depression (mGluR-LTD), a form of synaptic plasticity resulting in decreased synaptic efficacy, has been shown to strongly depend on activity of the RAS-ERK pathway 27,28 . Therefore, we hypothesized that hippocampal mGluR-dependent LTD might be affected in HRas G12V/G12V mice. Hippocampal slices from HRas G12V/G12V mice and their wild-type littermates were treated with DHPG (50 μM, 5 min), a specific agonist of group I (mGluR1 and mGlur5) metabotropic glutamate receptors, to induce LTD. This resulted in a substantial initial depression of synaptic transmission in WT mice and significantly less pronounced depression in HRas G12V/G12V mice (F 1,34 = 6.72, P < 0.05; Fig. 5E). This depression was followed by a robust long-lasting depression in fEPSCs amplitude in WT mice whereas it quickly returned to near baseline levels in HRas G12V/G12V mice, indicating that mGluR-dependent LTD cannot be stably expressed in HRas G12V/G12V mice (F 1,34 = 13.29, P < 0.01; Fig. 5E). In order to determine the pathway and receptor specificity of the LTD deficit in HRas G12V/G12V mice, N-methyl D-aspartate receptor-dependent LTD (NMDAR-LTD) was induced using low-frequency stimulation (LFS). Notably, LFS-induced LTD was similar between HRas G12V/G12V mice and WT littermates (F 1,26 = 0.40, P = 0.53; Fig. 5F), indicating that HRAS G12V specifically affects mGluR-dependent LTD.
Lovastatin does not rescue the spatial learning deficit in HRas G12V/G12V mice. As shown in Fig. 1, a major biochemical alteration exhibited in HRas G12V/G12V mice is hyper activation of the RAS-ERK pathway, resulting from the constitutive activation of HRAS. The G12V mutation reduces the intrinsic GTPase activity of HRAS protein and thereby prolongs its GTP-bound state 29,30 . Considering the fact that transition of GTDto GTP-bound RAS relies on proper anchoring to the plasma membrane by means of a lipid (farnesyl)-based plasma-membrane-targeting motif, interfering with its membrane recruitment is an established strategy to reduce RAS activity. Notably, treatment with lovastatin, a specific inhibitor of the rate-limiting enzyme in cholesterol biosynthesis (HMG-CoA reductase), can reduce RAS farnesylation and activity, and has been previously shown to reverse spatial learning deficits in the Nf1 31 and Ptpn11 10 mouse models for RASopathies. Therefore, we tested whether the same lovastatin treatment would also rescue the spatial learning deficit of HRas G12V/G12V mice. Mice were treated with lovastatin (10 mg/kg) once daily for 3 days prior to and during hidden platform Morris water maze training (daily injection 6 h prior to the training sessions). Although latencies to find the platform significantly decreased in all groups during training (within-subjects comparison, effect of time: F 1,58 = 46.80, P < 0.01), HRas G12V/G12V mice showed significantly longer latencies compared to wild-type littermates (between-subjects comparison, effect of genotype: F 1,58 = 5.89, P < 0.05, Fig. 6A). Moreover, there was no effect of Lovastatin treatment (genotype*treatment interaction: F 1,58 = 0.23, P = 0.61). The effect of lovastatin on spatial learning in both groups was assessed in a probe trial given after the 7 th day of the training. As shown in Fig. 6B, wild-type mice To further characterise the effect of lovastatin treatment, we quantified pERK (42/44) in hippocampal lysates of WT and in HRas G12V/G12V mice following the completion of behavioural testing. As described above, pERK1/2 level was significantly higher in HRas G12V/G12V mice compared to wild-type littermates (effect of the genotype: F 1,10 = 6.48, P < 0.05), but unaffected by lovastatin treatment (effect of the treatment: F 1,10 = 0.21, P = 0.65; interaction genotype*treatment: F 1,20 = 0.08, P = 0.78; Fig. 6C,D). Taken together, these data suggest that treatment with Scientific RepoRts | 7: 1256 | DOI:10.1038/s41598-017-01218-0 lovastatin using the same dosing regimen that was sufficient to rescue special learning and pERK hyper-activation in the Nf1 31 and Ptpn11 10 RASopathy models, is not effective in rescuing the spatial learning deficit of HRas G12V/ G12V mice.

Acute inhibition of MEK rescues deficit in mGluR-LTD in HRas G12V/G12V mice. Previous studies have
demonstrated an important role of the RAS-ERK pathway in mediating the expression of mGluR-LTD 27,32 . Hence we hypothesized that the mGluR-LTD deficit exhibited by HRas G12V/G12V mice might be caused by hyperactive ERK. If this model is valid, the impaired mGluR-LTD in HRas G12V/G12V slices should be able to be rescued by a MEK inhibitor, which would decrease ERK activation. Therefore, we induced mGluR-LTD by administering hippocampal slices with DHPG in addition to pre-treatment with either vehicle or a subthreshold dose of a selective and membrane-permeable MEK-inhibitor (U0126, 1 μM) which was previously shown not to affect WT mGluR-LTD at this concentration 28 . Application of DHPG induced stable depression of synaptic efficacy in wild-type slices, but only a brief depression in HRas G12V/G12V slices treated with vehicle solution (effect of genotype, vehicle treated group: F 3,63 = 8.53 P < 0.01; Fig. 7), thereby replicating the mGluR-LTD deficit in HRas G12V/G12V mice. Notably, pre-incubation of slices with the MEK-inhibitor U0126 rescued the LTD deficit in HRas G12V/G12V (WT vehicle vs. HRas G12V/G12V treated, Bonferroni's post-hoc test P < 0.05) without any discernible effect on LTD in WT littermates (WT vehicle vs. WT treated, Bonferroni's post-hoc test P > 0.05). Together, these results indicate that the deficit in mGluR-LTD observed in HRas G12V/G12V mice is indeed caused by hyperactive RAS-ERK signaling, and that acute inhibition of this pathway with a MEK inhibitor restores normal induction and expression of mGluR-LTD in HRas G12V/G12V mice.
Restoring pERK levels in HRas G12V/G12V mice by a MEK inhibitor does not rescue the spatial learning deficit or increased brain weight. Given that the mGluR-LTD deficit of HRas G12V/G12V slices can be rescued by a MEK inhibitor, we investigated whether the learning phenotype observed in HRas G12V/G12V could be reversed by treatment with a MEK inhibitor. For these experiments, we used the allosteric MEK inhibitor PD0325901 which has good bioavailability and is often used for in vivo studies, including studies targeting the brain 9, 33-35 . Although previous studies have shown a rescue of several phenotypes in an adult mouse model of Noonan syndrome using PD0325901 at 5 mg/kg bodyweight 35 , we sought to identify a lower effective dose, since inhibition of the RAS-ERK pathway results in learning deficits 36 . Mice were injected daily for 7 days with 2 or 5 mg/kg of PD0325901 and sacrificed 2 h after their final scheduled injection to assess pERK levels. As shown in Fig. 8A,B, pERK was significantly higher in vehicle-treated HRas G12V/G12V mice compared to vehicle treated WT littermates. Treatment with the MEK-inhibitor showed a dose-dependent effect on pERK in both HRas G12V/G12V mice as well as wild-type littermates (effect of genotype: F 1,34 = 6.88, P < 0.05). At only 2 mg/kg of PD0325901, a significant effect of treatment was already observed (F 1,34 = 38.95, P < 0.01), in which pERK levels in HRas G12V/ G12V mice were similar to vehicle-treated wild-type littermates. Hence, we chose to proceed with the 2 mg/kg daily dose, beginning 3 days prior to water maze training. All groups of mice showed a significant reduction in the latency to find the platform during training (effect of time: F 1,55 = 49.50, P < 0.01). However, HRas G12V/G12V mice exhibited significantly longer search latencies, regardless of treatment (effect of genotype: F 1,55 = 8.97, P < 0.01; Fig. 8C).
Spatial learning of wild-type and HRas G12V/G12V littermates, treated either with vehicle or PD0325901, was assessed in a probe trial after 5 days of training, a time point in which the wild-type mice showed already a strong preference for the target quadrant. Although it was notable that under these conditions vehicle treated HRas G12V/ G12V mice also showed a significant preference for the target quadrant, we again observed a significant effect of genotype with respect to time spent in the target quadrant, replicating the spatial learning deficit in HRas G12V/G12V mice (effect of genotype: F 1,51 = 6.07, P < 0.05; Fig. 8D). In addittion, a significant effect of treatment was observed this time (F 1,51 = 7.41 P < 0.05; Fig. 7D), but this was mediated by a decrease in performance of the treated mice. There was no significant interaction between genotype and treatment (F 1,51 = 1.21, P = 0.28).
In order to confirm that daily PD0325901 treatment reduced pERK levels in HRas G12V/G12V mice to that of their wild-type littermates, we performed Western blot analysis of hippocampal lysates taken from WT and HRas G12V/G12V mice immediately following the conclusion of the probe trial. Quantification of Western blots revealed a significantly higher level of pERK in vehicle-treated HRas G12V/G12V mice compared to vehicle treated wild-type mice (effect of genotype F 1,37 = 47.83, P < 0.01) with a significant decrease in pERK after PD0325901 treatment in both WT and HRas G12V/G12V mice (effect of treatment F 1,37 = 35.49, P < 0.01; genotype*treatment F 1,37 = 0.81,  Although PD0325901 significantly reduced performance of wild-type mice, it showed no effect on mutant mice and no interaction of genotype and treatment was observed (P = 0.28). TQ = target quadrant, RQ = adjacent right quadrant, LQ = adjacent left quadrant, OQ = opposite quadrant. Heat plots present the data of all tracks P = 0.38; Fig. 8E,F). Importantly, pERK levels in PD0325901-treated HRas G12V/G12V mice were similar to their vehicle-treated WT littermates, and thus reduced to normal levels.

Discussion
Here we report that a mouse model for Costello syndrome (CS), which carries an activating mutation in the HRAS gene, shows hyperactivation of the RAS-ERK pathway, learning deficits, structural brain abnormalities, and alterations in synaptic functioning. Although we studied homozygous HRas G12V/G12V mice, these mice express the HRAS G12V protein at approximately 50%, and hence their levels of oncogenic HRAS protein resemble that of heterozygous CS patients. The reduced HRAS protein expression is possibly caused by the IRES-β-geo cassette inserted into the 3′UTR of the HRas G12V targeting vector, as reduced expression is not observed in an independent line of HRas G12V mice 37 .
Previously it has been shown that pERK was not increased in this mouse line 16,17 , whereas we consistently observed a highly significant increase in pERK levels in hippocampal lysates (Fig. 1). This could be due to brain-region specific variability of ERK expression and activation levels resulting in no significant difference observed in total brain lysates 38 . Indeed, immunolabeling as shown by Viosca et al. clearly indicate increased pERK staining in CA3 mossy fiber region of hippocampus suggesting that, at least, locally pERK levels are enhanced.
We have deliberately focused our study on CS, as the RAS protein is the central protein of the RAS-ERK signaling pathway. In contrast to germline KRAS G12V mutants, the germline HRas G12V mutation is not lethal, which allowed us to directly study the molecular, cellular, and behavioral consequences of a constitutively active HRAS protein present from germline. Furthermore, it also enabled us to determine if there were differences in phenotype-or treatment efficacy compared to other mouse models of RASopathies caused by mutations in up-stream regulators of HRAS protein, such as NF1, SHP2/PTPN11, and SPRED1.
In accordance with several previous studies using adult transgenic HRas G12V overexpressing mice, we observed brain and neuronal hypertrophy in HRas G12V/G12V mutant mice 20,21 . Moreover, an enlarged corpus callosum was observed after bi-allelic inactivation of Nf1 9 . These morphological changes could not be rescued by acute treatment with the MEK-inhibitor PD0325901 during adulthood. A recent study reported that aberrant RAS signaling in the developing brain (E14-P7) resulted in an increased production of cortical neurons and morphological deficits 39 . In addition, other studies have demonstrated a clear requirement for intact ERK activity in controlling cortical development, including fibroblast growth factor (FGF)-stimulated cortical progenitor differentiation and neurogenesis 2,40 . Moreover, a recent study showed increased proteoglycan expression in astrocytes derived from iPSCs from CS patients as well as in astrocytes of mutant HRas G12S mice 41 . These studies highlight the importance of proper control of the RAS-ERK signaling during brain development, and suggest that adult treatment could be too late to restore the behavioural phenotype.
As summarized in Table 1, mouse models for Costello Syndrome (CS), Neurofibromatosis I (NF1), Noonan Syndrome (NS), and Legius Syndrome show similar phenotypes including hyperactivity of the RAS-ERK pathway, learning deficits in the water maze, and impaired synaptic plasticity, which confirm the widespread observation of overlapping disease manifestations across patients with different RASopathies. Given the shared biochemical alteration of increased hyperactivity of the RAS-ERK pathway, and the indisputable role of RAS-ERK signaling in LTP and memory formation 36,42 , it seemed reasonable to assume that the plasticity deficits observed in these RASopathy mouse models might all be caused by the same mechanism. However, a closer look at the specific cellular mechanisms reveals substantial differences between the RASopathy mouse models. For instance, neither the cell-type specific increase of inhibition found in Nf1 mutants 24 and a conditional Kras G12V mutant 26 , nor the increase of glutamatergic synaptic transmission observed in the NS model 10 were observed in HRas G12V/ G12V mice. In addition, we and others showed that LTP induced by theta-burst stimulation is impaired in all RASopathy mouse models published so far: Nf1 8, 25 , Ptpn11 10 , Spred1 mice 11 and conditional Kras G12V mice 26 . However, despite the strong ERK activation, LTP was unaffected in HRas G12V/G12V mice. The phenotypic differences between the mutants could be caused by differences in RAS-ERK activation (Table 1), by differences in cell-type specific expression, and/or by differential regulation during development. These possibilities could also explain increased LTP in transgenic mice expressing the same HRAS protein (HRAS G12V ) driven by the CAMK2 promoter 43 resulting in transgenic expression restricted to glutamatergic pyramidal neurons in the cortex and combined, in which the color indicates the mean time spent at that position in the pool. Data are presented as mean ± SEM from: vehicle group: wild-type n = 14 mice; HRas G12V/G12V n = 14 mice; treated group: wild-type n = 13 mice; HRas G12V/G12V n = 13 mice. Statistical test: repeated measures two-way ANOVA and two-way MANOVA, respectively. (E) Western blot of hippocampal tissue taken from WT and HRas G12V/G12V mice after water maze training and concurrent treatment with either vehicle or PD0325901. (F) Quantification of Western blots. A significant effect of genotype (F 1,37 = 47.83 P < 0.01) and treatment (P < 0.01) on pERK was observed. No interaction effect of genotype and treatment was found (P = 0.37). Data are presented as mean ± SEM from n = 10 mice per group. Statistical test: unpaired two-tailed t-test.
hippocampus. Moreover, there could be subtle differences in the intracellular location of HRAS, NF1, SPRED1, PTPN11 and KRAS which diverge in their relative subcellular signaling mechanisms 44,45 . Lastly, some of these proteins may have RAS-independent functions. For example, it was recently reported that NF1 interacts with hyperpolarization-activated cyclic nucleotide-gated (HCN) channels for which mutations in Nf1 cause attenuations of I h current selectively in GABAergic interneurons 25 . Treatment with the HCN channel agonist lamotrigine during adulthood rescued the cognitive deficits in two different Nf1 mouse models, thereby establishing an important function of HCN channels in mediating the NF1 phenotype. In contrast, HCN function is unaffected in HRas G12V/G12V mice 25 .
Previous studies have shown that mGluR-dependent LTD is sensitive to changes in ERK activity. High concentrations of the MEK-inhibitor U0126 (20 µM) impair mGluR-LTD 27, 28 , while a lower dose of U0126 (1 µM), which did not affect mGluR-LTD in WT mice, was sufficient to rescue the mGluR-LTD deficit in HRas G12V/G12V mice. This counter-intuitive result suggests that mGluR-LTD is sensitive to both hypo-and hyper activation of RAS-ERK pathway. Possibly, due to the high levels of RAS-ERK signaling in our mouse model, mGluR-mediated LTD was inhibited, and pre-treatment with a low dose of U0126 brought ERK signaling back into the normal range. To determine the mechanism underlying the impairment of mGluR-LTD in HRas G12V/G12V mice will require further investigation. In addition, it will be important to assess whether mGluR-LTD is altered the NF1, Legius Syndrome, or Noonan syndrome RASopathy mouse models.
Until recently, many symptoms of neurodevelopmental disorders, including RASopathies and mTOR-pathies 3 , were thought to be untreatable in adulthood due to their neurodevelopmental alterations, including gross brain dysmorphology and/or neuronal wiring/connectivity. Remarkably, candidate therapies implemented in NF1 and Noonan syndrome mouse models have shown extensive phenotypic rescue, including behavioral, synaptic, and molecular alterations. For instance, spatial learning deficits in adult Nf1 +/− mice were rescued by treatment with farnesyl-transferase inhibitor (FTI), which reduces RAS activity by interfering with the farnesylation of RAS, a post-translational modification that allows RAS to be targeted to the membrane 8,46 . Moreover, a behavioral rescue was shown after treatment of adult Nf1 +/− and Ptpn11 D61G/+ mice for 3-4 weeks with lovastatin, which affects the synthesis of farnesyl 10,31 . Furthermore, lovastatin reduced pERK levels in Nf1 +/− and Ptpn11 D61G/+ mice, and reduced pERK-dependent protein synthesis in the mouse model of Fragile X Syndrome, indicating its potency in restoring alterations of the RAS-ERK pathway 10,31,47 . Unfortunately however, we found that implementation of the identical lovastatin treatment regimen was unable rescue learning deficits and pERK levels in adult HRas G12V/ G12V mice (Fig. 6). It is likely that the G12V mutation, leads to much higher activation of the RAS-ERK pathway than mutations in upstream regulators of RAS protein, such as NF1 or PTPN11/SHP2 and therefore might be less responsive to lovastatin. In support of this, extensive studies of the effects of lovastatin as a potential cancer therapy have revealed that the concentration of statins required to significantly reduce farnesylation of RAS proteins with oncogenic mutations such as G12V are impossible to achieve in vivo without substantial side effects 48,49 .
A recent study has shown that the MEK inhibitor SL327 restored learning, LTP, and ERK signaling in adult mouse models for Noonan Syndrome, caused by mutations in PTPN11 10 (Table 1). Surprisingly, even though we observed that the mGluR-LTD deficit in HRas G12V/G12V mice can be rescued by a lower concentration of MEK inhibitor, treatment of HRas G12V/G12V mice with 2 mg/kg of PD0325901 (a MEK-inhibitor which has good bioavailability and is currently in clinical trials) 9, 33-35 successfully restored pERK levels but was insufficient to rescue the spatial learning impairment. Recently, it was also shown that treating adult conditional KRas G12V mice with PD0325901 failed to rescue the behavioural phenotype 26 . There are several possible explanations for our failure to rescue the behavioural phenotype of HRas G12V/G12V mice. First, although we chose a dose that resulted in a normalization of pERK levels, it may be possible that we have not used an optimal dose. Second, it might be that the observed morphological brain abnormalities interfere with normal learning, despite acutely normalized levels of RAS-ERK signaling. Lastly, it could be that successful rescue of the learning deficit of HRas G12V/G12V mice with lovastatin or MEK inhibitors requires treatment earlier in neurodevelopment. We recently demonstrated that adult gene-reactivation in a mouse model for Angelman Syndrome rescued LTP but not the cognitive and motor impairments, while early developmental gene reactivation led to a full rescue of all phenotypes, thereby suggesting a therapeutic critical period for achieving effective treatment in certain neurodevelopmental disorders 50 .
Our findings in the HRas G12V/G12V mouse model, as well as emerging findings in the NF1 and Noonan syndrome mouse models 10,25 , suggest that the mechanisms underlying learning deficits in RASopathies can be highly distinct, for which a universal treatment effective across all RASopathies remains elusive. Further studies are required to uncover the additional mechanisms underlying this divergence, and to identify the most optimal therapeutic strategies.

Materials and Methods
Animals. Mice used for this study were previously described and generously provided by Mariano Barbacid and co-workers 16 . These animals express a G → T missense mutation in the second base of the twelfth codon of the endogenous HRAS gene locus, which replaces the normal GGA (glycine) sequence with GTA (valine) causing a constitutively active, GTP-bound HRAS G12V protein. In order to be able to monitor HRAS expression, an internal ribosomal entry site-β-gal-neomycin resistance fusion protein (IRES-β-geo) cassette was inserted into the 3′ UTR of the HRAS gene. This allows visualization of HRAS gene expression patterns, by histological X-gal staining methods. HRas +/G12V mice were backcrossed (20-25 crosses) and maintained in the C57BL/6J background, and were intercrossed to generate homozygous (HRas G12V/G12V ), heterozygous (HRas G12V/WT ), and wild-type (HRas WT/WT ) littermates. All experiments were performed with HRas WT/WT and HRas G12V/G12 mice, as more robust neurological deficits were found among homozygous HRas G12V/G12 mice than in the analysis of heterozygous HRas G12V/WT mice, suggesting the existence of a gene dose effect 17 . Both males and females mice were used in all experiments.
All mice were between 8 and 14 weeks of age at the start of the behavioral experiments, housed in groups of 2-4 per cage, on a regular 12 h light/dark cycle (lights on between 7:00 am and 7:00 pm), and fed standard laboratory food ad libitum. Experimenters were blind to genotype and treatment. All experiments were approved by the Dutch Ethical Committee (DEC) and were in accordance with the institutional animal care and use committee guidelines.
Immunohistochemistry. Mice were deeply anesthetized with pentobarbital and perfused transcardially with Phosphate Buffered Saline (PBS) followed by freshly prepared 4% paraformaldehyde (PFA, Sigma). Brains were carefully removed and fixed for two hours in 4% PFA, followed by 24 h incubation in 0.1 M phosphate buffer (PB) and 10% sucrose. Brains were then embedded in gelatine blocks (10% gelatine, 10% sucrose), post-fixed in 10% formaldehyde and 30% sucrose for 3 h, and kept in 30% sucrose overnight at 4 °C. The embedded brains were sectioned coronally using a freezing microtome with a section thickness of 40 μm and collected in PBS. Sections were stained with Cytochrome C or thionin, processed for LacZ staining (to visualize the expression of the HRAS gene mutation), or processed for immunohistochemistry. LacZ staining was performed by overnight incubation in 1 ml X-gal solution containing: 5% X-gal, 5 mM K3F3(CN)6, 5 mM K4F3(CN)6.3H2O, 2 mM MgCl2, 0.01% Nadeoxycholate, 0.02% NP40 in PBS. In cases where X-gal treatment was combined with immunohistochemistry, X-gal incubation preceded the other immunohistochemical steps.
Field EPSP recordings were performed as previously described 25,52 . Briefly, field excitatory postsynaptic potentials (fEPSPs) were evoked by stimulating the Schaffer collateral/commissural pathway with a bipolar twisted platinum/iridium electrode (FHC, Bowdoinham, ME). Baseline stimulation (100-μsec duration) was adjusted to 30-40% (LTP) or 50% (LTD) of the maximum fEPSP amplitude. LTP was induced using different LTP protocols including: high frequency stimulation (100 Hz, 1 s) and theta burst stimulation (TBS) protocols. The TBS protocol consists of four trains at 5 Hz, each composed of four pulses at 100 Hz. Chemical mGluR-dependent LTD was evoked by application of (RS)-3,5-dihydroxyphenylglycine (DHPG, 100 μM) for 5 min. For LTD rescue experiments, either MEK-inhibitor (U0126, Tocris Bioscience, 1 μM) or DMSO was added to the recording solution before starting the protocol. For NMDA-dependent LTD, the average LTP or LTD over the last 10 minutes of each measurement was used for analysis.

Data and statistical analysis. Spontaneous synaptic currents (IPSCs and EPSCs) data were analyzed with
Mini Analysis (Synaptosoft Inc., Decatur, GA). All events were detected with a threshold >3× root-mean-square (RMS) of baseline noise. The detected currents were manually inspected to exclude false events. Passive and active membrane properties were analyzed offline using Clampfit 10 (Axon Instruments). Changes in synaptic strength were quantified for statistical comparisons by normalizing and averaging fEPSP slopes during the last 10 min of experiments relative to baseline.
All data are reported as mean ± SEM. Cumulative miniature and spontaneous synaptic current amplitude and inter-event interval (IEI) distributions were analyzed for statistical significance using the Kolmogorov-Smirnov (K-S) test (Statistica), and a conservative critical probability level of P < 0.01. All other statistical tests, including t-tests and one-or two-way (M)ANOVAs ([multivariate] analysis of variance), were performed using a critical probability of P < 0.05 and two-sided testing (SPSS version 20.0). Post-hoc Bonferroni correction analysis was performed only when an ANOVA yielded a significant (P < 0.05) main effect. Variance within each genotype or treatment was visualized in graphs. There were no overt differences in the variances between groups. All distributions were normal, either with 95% or 99% of confidence, and fulfilled the criteria for parametric testing. To detect or rule out meaningful, statistically significant differences between genotypes, we intended to include 10-15 animals per genotype for behavioral experiments, 5-10 animals per genotype for biochemical experiments, and 10-20 cells per genotype for electrophysiological measures. In general, this allowed us to reliably detect effect sizes of >15%. Genotypes were unblinded only after analysis was completed. Randomization of animals was exclusively implemented by the only person who had access to the genotypes (M.E.), and was typically based on the date of birth and the random distribution of mutant and wild-type mice in each litter. P < 0.05 is indicated with one asterisk (*), P < 0.01 with two asterisks (**), and P < 0.001 with three asterisks (***). P values ≥ 0.05 were considered as not rejecting the null hypothesis, and reported as not significant (N.S.).