Lovastatin, not Simvastatin, Corrects Core Phenotypes in the Fragile X Mouse Model

Visual Abstract

Previous work shows that the statin drug lovastatin, currently used for the treatment of high cholesterol in adults and children, resolves neuropathology in the Fmr1 -/y mouse model (Osterweil et al., 2013). Lovastatin normalizes protein synthesis by reducing the farnesylation and subsequent activation of the GTPase Ras, which lies upstream of the ERK1/2 signaling pathway (Schafer et al., 1989;Mendola and Backer, 1990). By this mechanism, lovastatin has also been shown to successfully correct electrophysiological and behavioral phenotypes in the mouse model of neurofibromatosis type 1 (NF1), a neurodevelopmental disorder of excess Ras (Li et al., 2005). In contrast to ERK1/2, the mTOR-p70S6K pathway activated by the GTPase Rheb is not altered by lovastatin suggesting the impact on farnesylation does not extend to all targets (Osterweil et al., 2013).
In the Fmr1 -/y mouse, the reduction of Ras-ERK1/2 by lovastatin ameliorates hippocampal epileptogenesis and neocortical hyperexcitability and significantly reduces the incidence of AGS (Osterweil et al., 2013). The AGS phenotype is one of the most robust behavioral phenotypes seen in the Fmr1 -/y mouse, and it models the epilepsy observed in FX patients (Musumeci et al., 2000;Berry-Kravis, 2002). Several previous studies have used AGS as a benchmark for determining the efficacy of potential treatment strategies, consistently finding a positive correlation between treatment efficacy at reducing seizure incidence and correction of other pathologies (Yan et al., 2005;Dölen et al., 2007;Osterweil et al., 2010Osterweil et al., , 2013Busquets-Garcia et al., 2013;King and Jope, 2013). Based on the positive outcome with lovastatin in Fmr1 -/y animal models, two open-label clinical trials tested the viability of lovastatin for the treatment of FX (Çaku et al., 2014;Pellerin et al., 2016). Both studies revealed a sig-nificant improvement with lovastatin treatment, and a double-blind placebo-controlled trial is ongoing (Berry-Kravis et al., 2017).
Interestingly, the availability of lovastatin is not widespread in Europe and is not licensed for use in the United Kingdom. Instead, the drug simvastatin has been proposed as an alternative therapeutic. Simvastatin is a structurally similar derivative of lovastatin that is twice as potent, with a daily dose of only 10 mg reducing cholesterol by 25-30% compared to 20 mg of lovastatin (Jones et al., 1998;Schaefer et al., 2004;Neuvonen et al., 2008). Simvastatin is also more brain penetrant than lovastatin, suggesting it may be a better option for neurologic indications (Tsuji et al., 1993). However, simvastatin has not been investigated in the Fmr1 -/y model, and the impact on Ras-ERK1/2 signaling in the brain is not well established. This information is critical, as clinical trials in NF1 have recently shown that lovastatin has a beneficial impact on cognitive function whereas simvastatin does not (Krab et al., 2008;Alabama-Birmingham andNCI, 2009, 2010;van der Vaart et al., 2013;Bearden et al., 2016;Payne et al., 2016).
In this study, we performed a side-by-side comparison of lovastatin and simvastatin to answer the simple but important question of whether there is a similar rescue of pathology in the Fmr1 -/y mouse. We focused on two core phenotypes in the Fmr1 -/y model: excessive protein synthesis and enhanced susceptibility to AGS. Importantly, our results clearly show that lovastatin, but not simvastatin, is effective in reducing ERK1/2 activity and normalizing protein synthesis in the Fmr1 -/y hippocampus. This suggests that simvastatin acts via a different mechanism from lovastatin with respect to ERK1/2-driven protein synthesis in the brain. To examine whether there was a similar impact on pathology, we performed a thorough AGS analysis using multiple doses of simvastatin. The results of these experiments show that simvastatin does not reduce the incidence or severity of AGS in the Fmr1 -/y mouse under conditions where lovastatin is significantly effective. Together, this evidence suggests simvastatin may not be a suitable replacement for lovastatin with respect to the treatment of FX.

Mice
All mice tested were male and were naive to drug and behavioral testing before experimentation. Mice were group housed with unrestricted food and water access and a 12/12 h light/dark cycle. Room temperature was maintained at 21 Ϯ 2°C. All animal procedures were performed in accordance with the University of Edinburgh animal care committee's regulations and the United Kingdom Animals Act. Fmr1 -/y mice (The Jackson Laboratory 003025, RRID:IMSR_JAX: 003025) were maintained on either a C57BL/6J (Charles River) or a mixed C57BL/6J x FVB background (C57BL/6J backcrossed to FVB by two generations).
After labeling, slices were homogenized in ice-cold buffer (10 mM HEPES, pH 7.4, 2 mM EDTA, 2 mM EGTA, 1% Triton X-100, protease inhibitors, and phosphatase inhibitors) and incubated in trichloroacetic acid (TCA; 10% final) for 10 min on ice before being centrifuged at 16,000 rpm for 10 min. The pellet was washed in ice-cold ddH 2 O and resuspended in 1 N NaOH until dissolved, and the pH was readjusted to neutral using 0.33 N HCl. Triplicates of each sample were subjected to scintillation counting and protein concentration assay kit (Bio-Rad). Counts per minute (CPM) were divided by protein concentration, and this was normalized to the CPM from the ACSF used for incubation. For display purposes, example slice homogenates were resolved on SDS-PAGE gels, transferred to nitrocellulose and exposed to a phosphorimaging screen (GE Healthcare). Phosphorimages were acquired using a Typhoon scanner (GE Healthcare) and compared to total protein staining of the same membrane.
To compare phopho to total for each target in the same lane, membranes developed for phospho [i.e., phosphorylated (p-)ERK1/2] were stripped and reprobed for total (i.e., ERK1/2). Phosphorylation of target proteins was calculated as a ratio of phospho to total. To correct for blot-to-blot variance, each signal was normalized to the average signal of all lanes on the same blot. Values are shown as a percentage of average WT vehicle for graphical purposes. All membranes were analyzed with experimenter blind to genotype and treatment.

AGSs
Test cohorts were counterbalanced for genotype and treatment. Naive WT and Fmr1 -/y male P18 -P29 mice bred on a mixed C57BL/6J x FVB background were weighed and injected intraperitoneally with 3 mg/kg simvastatin prodrug (CAS 79902-63-9), 50 mg/kg simvastatin active form (CAS 101314-97-0), or 100 mg/kg lovastatin active form (CAS 75225-50-2) or respective vehicle (3%, 20%, or 50% DMSO ϩ 10% Tween 80 in PBS). Animals were then transferred to a quiet (Ͻ60-dB ambient sound) room for 1 h. For testing, animals were moved to a transparent test chamber equipped with speakers and a webcam and allowed to habituate for 1 min. Audiogenic stimulation (recorded sampling of a modified personal alarm) was passed through an amplifier and 2 ϫ 50-W speakers (KRK Rokit RP5 G3 Active Studio Monitor) to produce a stimulus of Ͼ130 dB for 2 min. A decibel meter was placed at a standard distance from the speakers to ensure a stable emission of sound throughout each session. Incidence and severity of seizures was scored and video files for each session were saved. Latency was measured as the number of seconds between onset of the AGS stimulus and appearance of the first seizure. Stages of AGS severity were assigned according to previous work as follows: (1) wild running (WR; pronounced, undirected running and thrashing), (2) clonic seizure (violent spasms accompanied by loss of balance), or (3) tonic seizure (loss of movement and postural rigidity in limbs and tail). Any animal that reached tonic seizure was immediately humanely killed. All injections, testing and scoring was performed with the experimenter blind to genotype and treatment.

Statistics
Statistical testing was performed using GraphPad Prism 6 software, RRID: SCR_002798. For biochemistry experiments, outliers Ͼ2 SD from the mean were removed and significance determined by repeated measures twoway ANOVA and post hoc Sidak's multiple comparisons test. Significance for AGS incidence was determined using Fisher's exact test. AGS severity score distributions were tested for normality and found to be non-normal by Shapiro-Wilk test. These score distributions were then statistically compared using a Mann-Whitney U test for analysis of ordinal datasets with non-normal distributions.
Significant differences in latency to first seizure were determined using unpaired two-tailed Student's t test. Results of all statistical analyses are reported in detail in the statistical table (Table 1) and figure legends.

Results
Lovastatin, but not simvastatin, normalizes excessive protein synthesis in the Fmr1 -/y hippocampus Previous work shows that lovastatin normalizes excessive protein synthesis in the Fmr1 -/y hippocampus through reduction of Ras-ERK1/2 activation, which corrects epileptogenic phenotypes (Osterweil et al., 2013). To examine whether the same effect is seen with simvastatin, we used a metabolic labeling assay in hippocampal slices designed to assess protein synthesis in an intact preparation under physiologic conditions. Hippocampal slices were prepared from juvenile WT and Fmr1 -/y littermates, blind to genotype, and allowed to recover in oxygenating ACSF. Following this, slices were preincubated with Actinomycin D to block transcription, and new protein synthesis was labeled through incorporation of 35 S-labeled methionine/cysteine mix (Fig. 1A).
This puzzling increase in protein synthesis led us to wonder whether a reduced concentration of simvastatin might be more appropriate. To test this, we exposed slices to vehicle or simvastatin at concentrations of 0.1-0.5 M. Surprisingly, we find that even at these lower concentrations simvastatin causes a dose-dependent increase in protein synthesis, worsening the Fmr1 -/y phenotype (WT veh ϭ 100 Ϯ 2.21%, WT 0.  Fig. 1D). These results show that unlike lovastatin, simvastatin does not correct excessive protein synthesis in the Fmr1 -/y hippocampus.

Lovastatin, but not simvastatin, reduces ERK1/2 activation
Our metabolic labeling experiments show that 50 M lovastatin reduces protein synthesis in the Fmr1 -/y hippocampus by 15-20% (Fig. 1B). Conversely, 0.5 M simvastatin causes a 15-20% increase in protein synthesis in the Fmr1 -/y hippocampus (Fig. 1D). Given the opposite  effect of lovastatin and simvastatin on protein synthesis, we wondered whether these compounds acted differently on the ERK1/2 and mTOR translation control signaling pathways ( Fig. 2A). To confirm the same lovastatin treatment that reduces excess protein synthesis in the Fmr1 -/y also reduces ERK1/2 activation, we incubated slices in vehicle or 50 M lovastatin and performed quantitative immunoblotting for p-ERK1/2 ( Fig. 2B Next, to test whether simvastatin had a differential impact on ERK1/2 signaling at the same concentration that causes a 15-20% increase in protein synthesis, we repeated our immunoblotting analysis on slices exposed to vehicle or 0.1-0.5 M simvastatin. In contrast to lovastatin, our results show that simvastatin has no significant impact on p-ERK1/2 in either WT or Fmr1 -/y slices at any dose tested (WT veh ϭ 100 Ϯ 4.51%, WT 0.  Fig. 2-1). This suggests that simvastatin neither activates nor inhibits the ERK1/2 pathway under conditions where it increases protein synthesis.

Lovastatin, but not simvastatin, corrects the AGS phenotype in the Fmr1 -/y mouse
Our work in vitro shows that simvastatin does not correct the ERK1/2-stimulated excess in protein synthesis in the Fmr1 -/y hippocampus, suggesting that it may not have the same efficacy as lovastatin in ameliorating pathologic phenotypes. To directly test this, we performed a sideby-side analysis of the effect of lovastatin versus simvastatin on the incidence of AGS in the Fmr1 -/y mouse. Although the AGS phenotype is seen in Fmr1 -/y mice bred on multiple mouse background strains, a more robust  New Research phenotype is observed in mice bred on the FVB strain or a C57Bl6/J x FVB hybrid strain (Yan et al., 2004(Yan et al., , 2005. Therefore, we used Fmr1 -/y and littermate WT mice bred on a C57Bl6/J x FVB hybrid strain for our AGS study. Importantly, lovastatin corrects the AGS phenotype in Fmr1 -/y bred on both C57BL/6J and FVB strains, suggesting the rescue is not dictated by background genetics (Osterweil et al., 2013).
To test whether simvastatin could similarly correct the AGS phenotype, we injected Fmr1 -/y and littermate WT mice with 3 mg/kg simvastatin as described in Materials and Methods. We used the lactone prodrug version of simvastatin administered to human patients, which is hydrolyzed into the active hydroxy acid compound by the liver (Schachter, 2005). The initial dose of simvastatin was chosen based on previous work showing 1 mg/kg simvastatin reduces epileptogenic activity and neurotoxicity in a kainic acid (KA) rat model of epilepsy (Xie et al., 2011). Additionally, according to a conversion factor of 0.081 for mouse to human dosing recommended by the Food and Drug Administration (FDA), 3 mg/kg simvastatin in mouse would be equivalent to the 20 mg dose used in humans (Nair and Jacob, 2016).
Animals were injected with vehicle or simvastatin with the experimenter blind to genotype and treatment, and then left in a quiet environment for 1 h before AGS testing. A 1-h incubation time was chosen based on previous experiments using lovastatin, and on previous pharmacokinetic studies in mice and rats showing that simvastatin peaks in blood at 30 min to 1 h after administration (van de Steeg et al., 2013;Higgins et al., 2014;Xu et al., 2014), and peaks in brain 1 h after administration (Johnson-Anuna et al., 2005). To induce AGS, animals were transferred to a test chamber and exposed to a 2-min digitized sampling of a personal alarm passed through 50-W speakers at a level of Ͼ130 dB. Seizures were recorded at increasing levels of severity as: 1, wild running (uncontrolled and undirected running); 2, clonic seizure (loss of balance with violent spasms on all limbs); and 3, tonic seizure (loss of balance with postural rigidity in limbs and tail; Fig. 3A). Latency between the onset of the AGS stimulus and seizure was also used as a metric of seizure severity and measured as the number of seconds between the start of the alarm to the first appearance of wild running.
Although 3 mg/kg is consistent with a simvastatin dose used in previous studies of KA-induced seizure, higher doses of up to 50 mg/kg have also been investigated with respect to neurologic phenotypes in rodents (Ramirez et al., 2011;Ling and Tejada-Simon, 2016). Indeed, intraperitoneal injection of 50 mg/kg active simvastatin 24 h and 30 min before seizure induction protects against KA-induced seizures in mice (Ramirez et al., 2011), and increases learning in a mouse model of Alzheimer's disease (Li et al., 2006). To ensure that simvastatin is not effective in correcting the AGS phenotype in Fmr1 -/y mice, we repeated our experiments using a high dose of 50 mg/kg. To remove the potential confound of prodrug metabolism, we injected active simvastatin hydroxy acid rather than inactive lactone. In a comparison group, we tested an equipotent 100 mg/kg dose of active lovastatin hydroxy acid that was previously shown to correct AGS in adult Fmr1 -/y FVB mice (Osterweil et al., 2013). Separate groups of Fmr1 -/y and WT littermates were injected with 50 mg/kg simvastatin or 100 mg/kg lovastatin (with corresponding vehicle) and AGS testing performed as previously.

Discussion
The promising results using lovastatin in FX have led to the suggestion that simvastatin may be similarly effective. In this study, we investigated two core phenotypes in the Fmr1 -/y mouse model to test the prediction that simvastatin can be used in place of lovastatin. Our results show that simvastatin not only fails to correct excessive protein synthesis in the Fmr1 -/y hippocampus, it worsens this   (Fig. 1). We do not see a reduction of ERK1/2 activation at the concentrations of simvastatin tested (Fig.  2). Moreover, simvastatin does not reduce the incidence or severity of AGS in the Fmr1 -/y mouse even at a high dose of 50 mg/kg (Fig. 3). These results suggest that simvastatin should not be assumed to be an effective replacement for lovastatin with respect to correction of Fmr1 -/y pathology.
Although we propose the beneficial effect of lovastatin stems from the inhibition of ERK1/2-driven protein synthesis, it is important to note that statins are capable of affecting several biochemical pathways. Beyond the canonical impact on cholesterol biosynthesis, statins also decrease isoprenoid intermediates including farnesyl and geranylgeranyl pyrophosphates that regulate membrane association for many proteins including the small GT-Pases Ras, Rho, and Rac (Schafer et al., 1989;Liao and Laufs, 2005;Nürenberg and Volmer, 2012;Ling and Tejada-Simon, 2016). The increase in protein synthesis seen with simvastatin could be linked to altered posttranslational modification of these or other proteins. Indeed, although we see no change in mTORC1-p70S6K signaling, other studies have shown an activation of the PI3 kinase pathway that could be contributing to this effect (Mans et al., 2010). However, our comparison of lovastatin and simvastatin shows that there is a clear difference in the correction of pathology in the Fmr1 -/y model, suggesting that the impact on ERK1/2 is an important factor in terms of pharmacological treatment for FX.
There are many reasons why statins would be an attractive option for treating neurodevelopmental disorders such as FX. They are prescribed worldwide for the treatment of hypercholesterolemia and coronary heart disease (Istvan, 2003), and safely used for long-term treatment in children and adults (Ling and Tejada-Simon, 2016). However, our study suggests that care should be taken when considering which statin should be trialed for the treatment of FX and other disorders of excess Ras. Although the effect of different statins on cholesterol synthesis has been well documented, the differential impact on Ras-ERK1/2 signaling is not well established. We show here that, contrary to lovastatin, simvastatin fails to inhibit the Ras-ERK1/2 pathway in the Fmr1 -/y hippocampus, exacerbates the already elevated protein synthesis phenotype, and does not correct the AGS phenotype. These results are significant for considering future studies with lovastatin or simvastatin in FX or other disorders of excess Ras. Indeed, clinical trials using simvastatin for the treatment of NF1 have shown little promise, while trials with lovastatin show an improvement in cognitive deficits (van der Vaart et al., 2013;Bearden et al., 2016;Payne et al., 2016). Although further studies testing a broader dose range of simvastatin on additional Fmr1 -/y brain phenotypes will ultimately determine the feasibility of this strat-egy for FX, our study suggests caution should be used when assuming simvastatin is a suitable substitute for lovastatin.