Research ArticleFragile X Syndrome

Reversal of Disease-Related Pathologies in the Fragile X Mouse Model by Selective Activation of GABAB Receptors with Arbaclofen

See allHide authors and affiliations

Science Translational Medicine  19 Sep 2012:
Vol. 4, Issue 152, pp. 152ra128
DOI: 10.1126/scitranslmed.3004218


Fragile X syndrome (FXS), the most common inherited cause of intellectual disability and autism, results from the transcriptional silencing of FMR1 and loss of the mRNA translational repressor protein fragile X mental retardation protein (FMRP). Patients with FXS exhibit changes in neuronal dendritic spine morphology, a pathology associated with altered synaptic function. Studies in the mouse model of fragile X have shown that loss of FMRP causes excessive synaptic protein synthesis, which results in synaptic dysfunction and altered spine morphology. We tested whether the pharmacologic activation of the γ-aminobutyric acid type B (GABAB) receptor could correct or reverse these phenotypes in Fmr1-knockout mice. Basal protein synthesis, which is elevated in the hippocampus of Fmr1-knockout mice, was corrected by the in vitro application of the selective GABAB receptor agonist STX209 (arbaclofen, R-baclofen). STX209 also reduced to wild-type values the elevated AMPA receptor internalization in Fmr1-knockout cultured neurons, a known functional consequence of increased protein synthesis. Acute administration of STX209 in vivo, at doses that modify behavior, decreased mRNA translation in the cortex of Fmr1-knockout mice. Finally, the chronic administration of STX209 in juvenile mice corrected the increased spine density in Fmr1-knockout mice without affecting spine density in wild-type mice. Thus, activation of the GABAB receptor with STX209 corrected synaptic abnormalities considered central to fragile X pathophysiology, a finding that suggests that STX209 may be a potentially effective therapy to treat the core symptoms of FXS.


Fragile X syndrome (FXS) is the leading heritable cause of intellectual disability and the single most prevalent known cause of autism (1, 2). Children with FXS exhibit cognitive and behavioral deficits including social withdrawal, epilepsy, attention deficits and hyperactivity, and stereotyped behaviors (35). The phenotypes present at different developmental stages, and most persist throughout adulthood. For example, cognitive deficits are evident from infancy and the social deficits emerge within the first 2 to 3 years. Epilepsy emerges in childhood; however, seizures are often outgrown before adulthood (6).

FXS typically results from transcriptional silencing of FMR1, which encodes the polyribosome-associated RNA binding protein fragile X mental retardation protein (FMRP) (710). FMRP inhibits ribosome translocation by stalling ribosomes on their target mRNA, thereby inhibiting protein synthesis (11). The temporal and spatial regulation of protein synthesis is a key mechanism of neuronal differentiation during development and contributes to the synaptic plasticity that underlies learning and memory. FMRP is enriched at excitatory synapses where the local activity-dependent control of protein synthesis is essential for functional plasticity (12, 13). Deregulated synaptic protein synthesis has been proposed to underlie many of the cognitive and behavioral deficits associated with FXS (14, 15).

Signaling through the group 1 metabotropic glutamate receptors mGluR1 and mGluR5 stimulates mRNA translation at synapses (16, 17). The finding that some functional consequences of mGluR-stimulated protein synthesis are abnormal in the Fmr1-knockout mouse model of FXS (15, 1823) supports the theory that exaggerated responses to mGluR1/5 activation may contribute to the pathology and symptoms of FXS (14, 24). This theory has now been validated in several animal models of the disease (25, 26). For example, Fmr1-knockout mice display multiple disease-related phenotypes including audiogenic seizures, altered dendritic spine density, increased AMPA receptor (AMPAR) trafficking, and excessive protein synthesis that can all be rescued by reducing mGluR5 signaling (20, 2730).

Fmr1-knockout mice also show deficiencies in γ-aminobutyric acid (GABA)–mediated inhibitory neurotransmission, suggesting that there is an overall imbalance between excitatory and inhibitory neurotransmission in FXS. For example, Fmr1-knockout mice show reductions in GABA type A (GABAA) receptor subunit expression and GABA synthetic enzyme levels compared with wild-type mice (3133), as well as defects in inhibitory synaptic transmission in the cortex (34, 35). Fmr1-knockout mice also display reduced inhibition in the amygdala (36), which is paralleled by observations in FXS patients who show excessive activation of the amygdala during social cognitive tasks (37). These observations suggest that targeting GABA receptors may be therapeutic by directly restoring inhibitory tone in the FXS brain. Of particular interest is the metabotropic GABA type B (GABAB) receptor, which regulates cell excitability through pre- and postsynaptic mechanisms (38, 39). Presynaptic GABAB receptors on glutamatergic terminals inhibit glutamate release (38) and therefore have the potential to indirectly reduce postsynaptic mGluR5 activation.

The Fmr1-knockout mouse has expanded the understanding of FXS pathophysiology and has enabled development of targeted treatments designed to correct core disease pathologies instead of merely masking the symptoms. The hope is that treatments that correct the underlying pathologies will produce meaningful and multifaceted improvement in behavioral symptoms of humans affected by FXS. We initiated the present investigation to determine whether GABAB receptor activation can correct specific molecular pathophysiologies in the Fmr1-knockout mouse model of FXS. The first goal was to test whether STX209 could correct the excess protein synthesis that is believed to lie at the core of FXS pathophysiology and to contribute to synaptic phenotypes such as increased AMPAR turnover. The second goal was to test whether treatment initiated in juvenile animals can prevent or correct the dendritic spine phenotype that is characteristic of Fmr1-knockout mice and FXS patients (12, 27, 40). To guide our treatment strategy, we performed several behavioral tests—audiogenic seizure (a convulsive seizure model), marble burying (perseverative behavior), open-field activity (locomotor activity), and rotarod (motor coordination)—and determined STX209 blood and brain exposure at doses that alter behavior.


STX209 reduces protein synthesis in the hippocampus of Fmr1-knockout mice

Several methods have been developed to measure the excessive protein synthesis present in Fmr1-knockout mice (17, 27, 4143). Here, we used biosynthetic labeling to directly measure protein synthesis in hippocampal brain slices and synaptoneurosomes. In acute hippocampal slices, the effect of STX209 (10 μM) was genotype-specific [genotype × drug interaction was significant, F(1,82) = 4.16, P < 0.05]. Protein synthesis was elevated in hippocampal slices from Fmr1-knockout mice compared with wild-type mice (data expressed as percent of wild type ± SEM: wild type, 100 ± 5%; knockout, 129 ± 8%; t = 3.03, P < 0.01; Fig. 1). STX209 significantly reduced protein synthesis in Fmr1-knockout mice (knockout, 129 ± 8%; knockout + STX209, 94 ± 7%; t = 3.67, P < 0.001) but did not have a significant effect in wild-type mice (wild type, 100 ± 5%; wild type + STX209, 93 ± 6%; t = 0.73, P > 0.05). The protein synthesis inhibitor cycloheximide (60 μM) inhibited protein synthesis by >85% in both wild-type and knockout genotypes, indicating that the assay was measuring newly synthesized proteins. These results confirm previous findings in Fmr1-knockout mice (27, 29) and demonstrate that STX209 reduces excessive protein synthesis in hippocampal slices from Fmr1-knockout mice.

Fig. 1

STX209 inhibits basal protein synthesis in the Fmr1-knockout hippocampus. Protein synthesis was measured by radioactive amino acid incorporation in hippocampi from Fmr1-knockout (KO) and wild-type (WT) mice with or without STX209 (10 μM) treatment. Fmr1-KO mice (four slices per mouse from 21 WT mice and 22 Fmr1-KO mice) show elevated basal protein synthesis in comparison with WT mice (**P < 0.01), and STX209 inhibits protein synthesis in Fmr1-KO mice (***P < 0.001) but not in WT mice (P > 0.05). Differences were measured by Bonferroni t test.

To confirm that STX209 was affecting protein synthesis specifically at synapses, we measured protein synthesis in synaptoneurosomes prepared from mouse hippocampus (representative image shown in Fig. 2A). As observed in the slices, there was a significant effect of STX209 only in Fmr1-knockout samples [the drug × genotype interaction was significant: F(1,8) = 25.35, P = 0.001]. Pretreatment of Fmr1-knockout synaptoneurosomes in vitro with STX209 (10 μM) significantly reduced protein synthesis to wild-type levels (knockout + STX209, 78 ± 2%; t = 8.19, P < 0.0001; Fig. 2B). STX209 did not affect protein synthesis in wild-type synaptoneurosomes (wild type + STX209, 90 ± 9%; t = 1.07, P > 0.05).

Fig. 2

STX209 inhibits synaptic protein synthesis in Fmr1-KO, but not in WT, hippocampal synaptoneurosomes. (A) Representative electron micrograph of synaptoneurosome preparation. Arrows, presynaptic vesicles; arrowhead, postsynaptic density (PSD). Scale bar, 100 nm. (B) Fmr1-KO synaptoneurosomes (SNS) (n = 3) show elevated protein synthesis in comparison with littermate WT synaptoneurosomes (***P < 0.001). STX209 (10 μM) inhibits protein synthesis in Fmr1-KO synaptoneurosomes but not in WT synaptoneurosomes (***P < 0.001). Two-way analysis of variance (ANOVA) followed by Bonferroni t test.

STX209 corrects AMPAR trafficking in Fmr1-knockout neurons

One known functional consequence of the excessive synaptic protein synthesis in Fmr1 deficiency is increased AMPAR turnover (28). Therefore, we tested the effects of STX209 on the rate of AMPAR endocytosis in Fmr1-knockout mouse neurons in culture. Primary hippocampal neurons were analyzed. We have previously determined that inhibiting translation does not affect spontaneous AMPAR trafficking in wild-type neurons (28); therefore, only neurons from Fmr1-knockout embryonic mice were treated with or without STX209 (100 μM for 1 hour or 10 μM for 5 hours). As expected, Fmr1-knockout neurons showed increased endocytosis of AMPARs compared with wild-type neurons (ratio of internalized to total AMPAR expressed as percent; wild type, 42.5 ± 6.7%; knockout, 56.6 ± 5.4%; P < 0.001; Fig. 3); STX209 treatment significantly reduced AMPAR internalization to wild-type levels [knockout + STX209 (10 μM), 42.7 ± 5.0%; knockout + STX209 (100 μM), 41.8 ± 5.8%; P < 0.001; Fig. 3]. The persistent AMPAR internalization in the absence of FMRP has been proposed to negatively affect synaptic plasticity and thus contribute to impaired cognition. These results suggest that GABAB receptor activation may correct this molecular defect of Fmr1 deficiency.

Fig. 3

STX209 rescues AMPAR trafficking in Fmr1-KO neurons in culture. Constitutive endocytosis of AMPARs was measured in Fmr1-KO and WT hippocampal neurons (ratio of internalized to total AMPAR expressed as a percent; ***P < 0.001). The elevated constitutive endocytosis of AMPARs in Fmr1-KO neurons was inhibited by application of STX209 at either 100 μM for 1 hour or 10 μM for 5 hours (***P < 0.001). Differences were measured by one-way ANOVA followed by Bonferroni t test. For each condition measured, n = 30 individual dendrites.

Acute administration of STX209 inhibits mouse behavioral and convulsant phenotypes

To determine whether STX209 corrects synaptic abnormalities central to fragile X pathophysiology in vivo, we first established the effective dose range by testing STX209 in several behavioral assays.

Audiogenic seizures are a robust and reproducible phenotype in the Fmr1-knockout mouse (4446), and previous work has shown that R-baclofen suppresses audiogenic seizures in the Fmr1-knockout mouse on the seizure-resistant C57BL/6 background with an effective dose of 1.0 mg/kg (47). We extended these findings to Fmr1-knockout mice on the FVB background, a strain that is relevant to our studies, because they show a robust seizure phenotype and have defects in GABA-mediated inhibitory neurotransmission (33). We found that acute intraperitoneal administration of STX209 significantly reduced the percent of mice displaying seizures (seizure incidence), with a minimum effective dose (MED) of 1.5 mg/kg (Fig. 4A, P < 0.0001). Racemic baclofen also significantly reduced seizure incidence, but with a higher MED of 6.0 mg/kg (P < 0.0001), suggesting that STX209 may be more potent than racemic baclofen in reducing audiogenic seizures in Fmr1-knockout mice. All subsequent studies were therefore performed using only STX209.

Fig. 4

STX209 inhibits mouse behavior. (A) Effect of STX209 and racemic baclofen on seizure frequency in Fmr1-KO mice. STX209 reduces seizures at 1.5 mg/kg; ****P < 0.0001. Racemic baclofen reduces seizures at 6.0 mg/kg; ****P < 0.0001. n = 10 to 14 mice per group from two independent experiments. These differences were tested for significance with two-tailed Fisher’s exact test. (B) STX209 inhibits marble burying in WT and Fmr1-KO mice at 6 mg/kg (***P < 0.001). n = 12 mice per group. (C) STX209 inhibits total distance traveled in the open-field arena in WT and Fmr1-KO mice at 3 mg/kg (**P < 0.01, ****P < 0.0001). Results shown are from three independent experiments; n = 9 to 12 mice per group. (D) Effect of STX209 on time spent on the rotarod. WT mouse behavior is altered with STX209 at 6 mg/kg (*P < 0.05); mice from both genotypes are affected at 10 mg/kg (***P < 0.001). n = 6 to 18 mice per group. Group differences were tested for significance with Bonferroni t test. Bars represent SEM.

Repetitive behavior is a characteristic phenotype in FXS and a defining symptom of autism spectrum disorders (48, 49). Marble burying is an assay of perseverative/repetitive behavior in mice (50), and the effects of a selective GABAB receptor agonist on marble burying behavior in the FXS mouse have not been reported. Therefore, we tested the effects of STX209 on marble burying behavior in wild-type and Fmr1-knockout mice. We found that acute intraperitoneal administration of STX209 significantly reduced marble burying [Fig. 4B; F(3,87) = 31.09, P < 0.0001] in both wild-type and Fmr1-knockout mice without a significant genotype × drug interaction [F(3,87) = 0.20, P > 0.05] and with an MED of 6 mg/kg for both genotypes (Fig. 4B; P < 0.001). However, STX209 also significantly reduced total distance traveled in an open field [Fig. 4C, KO; F(1,36) = 45.52, P < 0.0001] with an MED at 3 mg/kg, which is lower than the effective dose in the marble burying assay. These data suggest that the STX209 effect on marble burying may be due to reduced locomotor activity and not a specific effect of STX209 on repetitive behavior.

We next used the rotarod assay, which measures motor coordination in rodents and is commonly used to assess the sedative-like properties of compounds, including racemic baclofen, in mice (51, 52). STX209 significantly reduces the latency to fall [Fig. 4D; F(4,102) = 23.96, P < 0.0001] in wild-type mice (Fig. 4D; MED = 6 mg/kg, P < 0.05) and Fmr1-knockout mice (Fig. 4D; MED = 10 mg/kg, P < 0.001), demonstrating impaired motor coordination at slightly higher doses (~10 mg/kg). On the basis of the finding that Fmr1-knockout mice were not significantly impaired on the rotarod at 6 mg/kg, we chose this dose to produce maximal efficacy in Fmr1-knockout mice while minimizing secondary effects.

In vivo administration of STX209 reduces protein translation in Fmr1-knockout mice

To test whether protein synthesis is altered in the Fmr1-knockout brain after acute STX209 administration, we used a surrogate measure for protein synthesis, polysome profiling. FMRP is a polyribosome-associated RNA binding protein that co-sediments with polyribosomes (53) and large mRNA granules during sucrose density gradient centrifugation (41, 54, 55). The fractionated polysome profile is altered in Fmr1-knockout mice but can be corrected by in vivo administration of the mGluR5 inhibitor MPEP [2-methyl-6-(phenylethynyl)pyridine hydrochloride] (41). We tested whether the in vivo administration of STX209 (6 mg/kg) could correct the abnormal polysome profile observed in Fmr1-knockout mice. In assays that resolved the polyribosome fraction, we showed an increase in the ratio of 80S monosomes to polysomes in the samples from STX209-treated mice (P = 0.03, paired t test; Fig. 5B). The data were expressed as the slope of a line drawn from the 80S monosome peak to the trisome peak (Fig. 5A). The increase in this index reflects a relative reduction in polysome formation, and because the polysome fraction is translationally active (55), these data indicate that there is reduced translation.

Fig. 5

STX209 alters the polysome profile in the Fmr1-KO mouse brain. (A) Sample traces showing A260 absorbance profiles through sucrose gradients used to resolve the polysome profile in saline-treated (red) and STX209-treated (blue) Fmr1-KO mouse brains. Dashed lines show differences in the monosome-trisome slope. (B) Box plots summarizing the monosome-polysome index in saline- and STX209-treated Fmr1-KO mice. n = 6 mice; paired t test, P = 0.03.

STX209 administered in drinking water provides therapeutic exposure levels

To test the effects of STX209 on dendritic spine density, we sought to mimic the effects of continuous exposure in patients over a defined period of brain development. Recent results indicate that environmental enrichment itself can modify the spine phenotype (56), and therefore, the daily handling required to administer multiple drug injections might alter the phenotype we sought to measure. We therefore administered STX209 in the drinking water. To ensure that this method resulted in plasma and brain concentrations similar to those that were effective in the behavior and mRNA translation assays, we performed a pharmacokinetic study. Brain and plasma samples were collected after 9 days of administration of STX209 by either twice-daily intraperitoneal injection (6 mg/kg) or oral administration in drinking water (0.5 mg/ml). For each study, we took samples for about 24 hours. When administered in the drinking water, STX209 produced similar systemic exposure as STX209 (6 mg/kg) given twice daily intraperitoneally (Table 1). Although STX209 (0.5 mg/ml) in the drinking water results in lower peak plasma concentrations relative to twice-daily intraperitoneal administration, with a Cmax of 2251 ng/ml, the associated area under the curve (AUClast) was 4502 ng hour/ml, a systemic exposure nearly equivalent to that observed with STX209 (6 mg/kg) administered by intraperitoneal injection. Brain concentrations were somewhat higher in the drinking water group (0.5 mg/ml) (Cmax of 0.181 ng/mg and AUClast of 0.467 ng hour/mg) when compared with those observed with 6 mg/kg given intraperitoneally (Cmax of 0.133 ng/mg and AUClast of 0.223 ng hour/mg). These results indicate that STX209 (0.5 mg/ml) added to the drinking water provides plasma and brain exposures similar to those achieved by intraperitoneal administration of STX209 at a dose that reduces mRNA translation in brain.

Table 1

STX209 concentrations in mouse plasma and brain. The pharmacokinetics of STX209 in plasma and brain after inraperitoneal injection or administration in the drinking water.

View this table:

We assessed the in vitro plasma protein binding and brain homogenate binding of STX209. Across species, the STX209-free fraction in both matrices is about 70% (over an STX209 concentration range of 0.5 to 10 μM, the average plasma protein binding is 34, 26, and 21% for human, rat, and mouse, respectively; the average brain homogenate binding is 37 and 34% for rat and mouse, respectively). The Ki (inhibition constant) for STX209 in rat cortical brain homogenates is 620 nM and the EC50 (concentration producing a half-maximal specific response) in recombinant Chinese hamster ovary (CHO) cells is 1.2 μM. Owing to low protein binding, a dose of 6 mg/kg results in a maximum unbound STX209 brain concentration of 594 nM after administration in the drinking water, which is sufficient to bind and activate the GABAB receptor.

STX209 corrects the increased spine density in Fmr1-knockout mice

A feature of FXS both in the mouse model and in patients is altered structure and density of dendritic spines, which is associated with impaired synaptic connectivity and ultimately neuronal network dysfunction (5759). The fragile X dendritic spine has been described as immature because it has a thin or filopodial appearance, in contrast to mature spines that are mushroom-shaped (58). In FXS patients, the immature spine morphology can be present both with and without changes in spine density (58, 6062). In the mouse model of FXS, spine density and morphology are altered in an age-, region-, and cell type–specific manner (63, 64).

We focused our study on pyramidal neurons from the binocular visual cortex, an area that has been well characterized during experience-dependent plasticity and that shows increased spine density in FXS (57, 58). In the Fmr1-knockout mouse, a reduction in mGluR5 activity by germline knockout (27) or postnatal inhibition (65) can rescue this spine phenotype. We therefore asked whether treatment with chronic STX209 could do the same.

Wild-type and Fmr1-knockout mice received STX209 in the drinking water from weaning at postnatal day (P) 23 (±1) until P35 (±1). At P35 (±1), the spine density of apical oblique segments was analyzed with Golgi staining (representative images in Fig. 6A). The effects of STX209 were significant and genotype-specific [genotype × drug interaction: F(1,25) = 7.44, P = 0.01]. Fmr1-knockout mice exhibited increased spine density on apical oblique branches of layer 2/3 pyramidal neurons compared with wild-type mice (mean spine density ± SEM per 10 μm: wild type, 11.47 ± 0.16; knockout, 13.18 ± 0.40; t = 3.34, P < 0.01; Fig. 6B). STX209 treatment in the Fmr1-knockout significantly reduced spine density (mean spine density ± SEM: knockout, 13.18 ± 0.40; knockout + STX209, 11.12 ± 0.50; t = 3.38, P < 0.01) but had no effect in wild-type cortex (mean spine density ± SEM: wild type, 11.47 ± 0.16; wild type + STX209, 11.54 ± 0.53; t = 0.14, P > 0.05). In both wild-type and knockout genotypes, the spines were distributed along the dendrite in the characteristic hyperbolic pattern between 40 and 70 μm from the dendrite’s origin on the apical shaft (Fig. 6C).

Fig. 6

STX209 corrects the excess dendritic spine phenotype of Fmr1 deficiency. Fmr1-KO and WT mice were treated with STX209 in drinking water, and spine density on apical oblique dendritic segments of layer 2/3 pyramidal neurons was assessed. (A) Experimental design and representative images of spine morphology from each experimental group. The images are from apical oblique dendrites branching within layer 3. Scale bar, 2.5 μm. (B) The mean spine density ± SEM per 10-μm segment of apical oblique dendrites for the entire apical oblique dendritic branch. Group differences were measured by Bonferroni t test; **P < 0.01. (C) Spine density represented as the number of spines per 10 μm along the apical oblique branch from its point of origin. WT mice, n = 11; Fmr1-KO mice, n = 6; WT mice treated with STX209 (WT + STX209), n = 7; Fmr1-KO mice treated with STX209 (KO + STX209), n = 5. Bars, SEM.


Current pharmacologic treatment of FXS and other neurodevelopmental disorders is limited to the management of symptoms with drugs such as antipsychotics and psychostimulants. Now, with a better understanding of FXS disease pathology, treatments can be developed that restore normal brain function. Here, we present evidence that the GABAB receptor agonist STX209 can modify aspects of the core pathophysiology in FXS. In the Fmr1-knockout mouse, STX209 restored protein synthesis to wild-type values, reversed the excessive AMPAR trafficking, and corrected the increased dendritic spine density. Moreover, because the correction of the spine phenotype occurred in mice after weaning and after the peak of spinogenesis in the neocortex, our results suggest that STX209 can correct an already established disease phenotype.

Learning and memory require lasting changes to neurons, and these occur through the strengthening and weakening of synaptic connections. Dendritic spine morphology is associated with this synaptic plasticity: mature, large spines are associated with strengthened neuronal connections, whereas immature or long, thin filamentous spines are associated with weakened connections (66). During development, large numbers of spines are formed and pruned so that the density of dendritic spines is higher during early stages of development than in adulthood. Abnormal spine density and morphology occur in patients with FXS, and spine abnormalities are also present in other neurodevelopmental disorders including Down syndrome and schizophrenia (67, 68). The increased density and immature morphology of dendritic spines in the Fmr1-knockout mouse has led to the hypothesis that aspects of the disease may lie in the pruning or shaping of the neuronal network, the results of which persist throughout adulthood (69).

Dendritic spine density of layer 2/3 pyramidal cells in the rodent visual cortex increases rapidly between P11 and P18 and reaches a plateau by P30 (70, 71). STX209 was administered at P23, after the rapid increase in spine density, but before the plateau. Recent work reveals that the fragile X spine phenotype has emerged by this age as many cell types from the primary somatosensory cortex show altered functional and structural synaptic development over the first 2 postnatal weeks (59, 7274). At P14, both layer 4 spiny stellate cells (74) and layer 5 pyramidal cells (59) of Fmr1-knockout mice show altered spine morphology. Spine stabilization is also abnormal in developing Fmr1-knockout mice; it is delayed in dendrites from layer 2/3 pyramidal cells (72) and more unstable spines are produced throughout development in layer 5 pyramidal cells (75). The alterations in spine density, structure, and function all occur at defined points in development. In humans, spine density in the visual cortex peaks during early infancy between 4 and 12 months of age, decreases rapidly between 2 and 10 years, and reaches a plateau after 10 years of age (76). The average age for diagnosis of FXS is 36 months (77), and FXS phenotypes are thought to arise from the observed deficits in synaptic connectivity. Therefore, phenotype reversal is an important goal in the development of targeted treatment.

Of note is our finding that STX209 fully restores normal dendritic spine density in Fmr1-knockout mice without significantly reducing spine density in wild-type littermates. This result is similar to the demonstration that the long-acting mGluR5 inhibitor CTEP [2-chloro-4-((2,5-dimethyl-1-(4- (trifluoromethoxy)phenyl)-1H-imidazol-4-yl)ethynyl)pyridine] can reverse increased spine density in young adult mice (65). These studies suggest that pharmacologic treatment can successfully correct an already established spine morphology phenotype. In mice, excess spine density has been corrected at P35 (this study) and P65 (65), ages roughly equivalent to juvenile and young adult humans. As shown in the STX209 clinical trial, the effects of STX209 on social behavior in human FXS can be seen in a somewhat wider age range, from 6 to 39 years (78).

De-repression of synaptic protein synthesis by the loss of FMRP is believed to contribute to the pathogenesis of FXS (79). We show here, using several independent assays, that excessive cerebral protein synthesis in the Fmr1-knockout mouse can be corrected selectively by STX209. In the in vitro assays, the effect was selective, correcting the excess protein synthesis in the knockout without affecting the wild type. Because we did not measure mRNA translation after STX209 treatment in vivo in wild-type mice, we cannot rule out the possibility that the observed effect is genotype-independent under these conditions. However, the finding that protein synthesis in vitro is selectively decreased only in Fmr1-knockout, and not wild-type, mice suggests that STX209 treatment affects translation of the FMRP targets. One consequence of excess protein synthesis in Fmr1-deficient hippocampal neurons is altered regulation of AMPAR trafficking (28). The dynamic regulation of AMPAR function is a central mechanism of synaptic plasticity, with long-term potentiation (LTP) accompanied by synaptic delivery of AMPARs and long-term depression (LTD) caused by AMPAR internalization (80). The elevated AMPAR endocytosis seen in neurons from Fmr1-knockout mice (28) is consistent with the finding of exaggerated mGluR5-dependent LTD in Fmr1-knockout hippocampal slices (15).

Whereas behavioral analyses informed our dose selection for the mechanistic studies, it offers little insight into the possible clinical use of STX209. The FXS mouse model exhibits few robust and reproducible disease-specific behaviors. Genotype-specific readouts for anxiety, perseverative behavior, and hyperactivity all show varying results (8191), and the Fmr1-knockout mouse has not shown a robust social phenotype. Despite this, some studies reported Fmr1-knockout behavioral phenotypes that are reversed with pharmacological treatment, such as with mGluR5 inhibitors. Of note is a recent report that CTEP, which corrects the Fmr1-knockout spine phenotype, also improves abnormal behaviors in the Fmr1-knockout mouse, including hyperactivity, reduced inhibitory avoidance, and hypersensitivity to auditory startle and audiogenic seizures (65). Similarly, we found here that acute administration of STX209 rescues audiogenic seizures in Fmr1-knockout mice, but the interpretation of these findings is complicated by the nonspecific sedative properties observed. Both the anticonvulsant (47) and the sedative (92) effects of R-baclofen tolerate with repeated administration in mice. In contrast, here we found that daily administration of STX209 rescued the dendritic spine phenotype of Fmr1-knockout mice, without affecting wild-type mice. Together, these observations suggest that the acute effects of STX209 on mouse behavior and the disease-modifying effects of STX209 after repeated administration occur through separate mechanisms.

Key findings in the clinical trial testing STX209 include improved social function in patients with more severe impairments and a relatively low frequency of sedation (8% in subjects receiving STX209 and 2% in subjects receiving placebo) (78). This response in a subset of the FXS patients could not have been predicted by behavioral tests in Fmr1-knockout mice, because they are an isogenic model. It will be of interest in future studies to assess whether alternative behavioral paradigms for Fmr1-knockout mice can detect phenocopies as they relate to treatment response. It is also of interest to examine whether chronic pharmacological treatments that correct structural and functional Fmr1-knockout phenotypes, such as STX209, eventually lead to improved cognition. For example, Fmr1-knockout mice are impaired in a five-choice cognitive task governed by the prefrontal cortex (93). If cognitive deficits of FXS arise from abnormalities in synaptic and dendritic structure and function during development, a key question is whether pharmacologic reversal of these phenotypes can affect cognition later in life.

In a recent clinical trial, a subset of patients treated with the mGluR5 antagonist AFQ056 showed improvements in the ABC-C score, a measure of problem behaviors (94). If STX209 solely functioned by reducing mGluR5 signaling, then some overlap in the treatment responses would be expected. The clinical trial results to date do not support this hypothesis (78, 94); however, longer treatment durations may reveal similarities. Several GABA-mediated mechanisms, acting individually or in parallel, may additionally contribute to the disease-modifying effects of STX209. Postsynaptic GABAB receptors on dendritic spines can dampen cell excitability by gating potassium channels (95), potentially reversing the imbalance between excitatory and inhibitory neurotransmission present in FXS. Presynaptic GABAB receptors inhibit glutamate release (38) and may therefore limit not only the activation of mGluR5 but also that of other glutamate receptors. Although delineation of the precise molecular mechanism(s) requires further examination, the magnitude of the rescue by GABAB receptor activation on the phenotypes examined here is comparable to that observed with inhibition of mGluR5 signaling (2730, 41).

Both STX209 and the mGluR5 inhibitors correct the increased protein synthesis, AMPAR turnover, and dendritic spine density of fragile X mice—phenotypes that are believed to be central to disease pathogenesis. GABAB receptor agonists may therefore provide a new avenue for disease-modifying treatment of FXS. Unlike the mGluR5 inhibitors, which will require extensive assessment in adults and adolescents before progressing to use in children, baclofen has been used safely in clinical practice for more than 30 years. This knowledge should accelerate development of STX209 for children with FXS.

Materials and Methods

Biosynthetic labeling

Age-matched male wild-type and Fmr1-knockout littermates between P25 and P34 were used to prepare hippocampal slices. Slices were prepared for radiolabeling and analyzed as described with minor modifications (29). Briefly, slices were incubated with [35S]methionine/cysteine (10 μCi/ml; EasyTag Express 35S Protein Labeling Mix, PerkinElmer) for 30 min in the absence or presence of STX209 (10 μM) or vehicle. Hippocampal slices were snap-frozen and homogenized, and radiolabeled proteins were precipitated with 10% trichloroacetic acid and quantified by liquid scintillation counting (Beckman Coulter, LS6500). For each experiment, hippocampi were dissected from five or more mice per genotype and four dorsal slices were analyzed for each animal treatment group. The combined results from four independent experiments (total n = 21 wild-type mice and n = 22 Fmr1-knockout mice) are reported. For statistical analyses, an n is an animal, and the data were analyzed with two-way ANOVA with post hoc Bonferroni test.

We used age-matched male wild-type and Fmr1-knockout littermates between P21 and P30 to prepare hippocampal synaptoneurosomes as described (96, 97). Because of inter-assay variability, we assessed the effect of drug treatment only in those experiments showing a significant genotype effect (determined by the Student’s t test). Each experiment was conducted on six pooled hippocampi from three mice per genotype. Synaptoneurosome samples from each genotype were divided equally and treated with vehicle or STX209; measurements were performed in triplicate. The combined results from three independent experiments (total n = 9 mice per genotype) are reported. Data were analyzed by two-way ANOVA with post hoc Bonferroni test.

AMPAR trafficking in primary neurons

Constitutive AMPAR internalization was measured in primary hippocampal cultures isolated from wild-type and Fmr1-knockout mice as described (28, 98). Briefly, primary hippocampal neurons from wild-type or Fmr1-knockout embryonic mice were isolated and grown in culture for 28 days as described (99). Surface AMPARs in live neurons were labeled with a rabbit polyclonal antibody against the N terminus of the GluR1 subunit for 15 min at 37°C, 0.5% CO2. Cells were fixed in 4% formaldehyde, processed, and analyzed as described (28). For each experimental condition, 30 individual dendrites were analyzed blind to treatment. The effect of STX209 on Fmr1-knockout neurons was tested by one-way ANOVA with post hoc Bonferroni test.

Behavioral studies

All animal work was conducted in compliance with the Animal Welfare Act, and the study protocols were approved by the Institutional Animal Care and Use Committee for each relevant institution. For behavioral studies, 8- to 12-week-old male mice were used and housed in a temperature- and humidity-controlled environment with standard alternating 12-hour light/dark cycles; food and water were provided ad libitum. The Fmr1-knockout mouse line congenic on the C57BL/6J background has been described (82), and wild-type and Fmr1-knockout mice were bred as littermates in a single colony. The Fmr1-knockout mouse line on the FVB background is congenic (FVB.129P2-Fmr1tm1Cgr/J, The Jackson Laboratory); wild-type and Fmr1-knockout mice were maintained as separate colonies and this line was studied only in the audiogenic seizure assay.

Audiogenic seizure susceptibility was examined in 10 to 14 mice per group as described (30), with minor modifications. Briefly, the 130-dB auditory stimulation was applied for two 2-min trials separated by a 1-min inter-trial interval. Seizure behavior was scored as follows: 0, no response; 1, wild running; 2, clonic seizure; 3, tonic seizure; 4, respiratory arrest/death. Seizure was defined as a seizure score of 2 or greater, and data are expressed as the percentage of mice displaying seizure (seizure incidence, %). Wild-type animals were included in each study (n = 36 in total) and consistently exhibited a seizure score of zero. Data were analyzed with two-tailed Fisher’s exact test.

For the marble burying assay, 12 mice per group were individually placed in standard mouse cages containing 20 marbles arranged in a 4 × 5 grid on aspen chip mouse bedding (NEPCO). The test time was 30 min. Digital images of marbles after the test were scored blind to genotype and treatment. Data from the marble burying assay were analyzed using a two-way ANOVA with post hoc Bonferroni test.

Locomotor activity in an open field was examined in 9 to 12 mice per group over 20 min with the AccuScan photobeam monitor and Superflex software (AccuScan Instruments). Open-field arenas (40 × 40 × 30 cm) were housed in environmental control chambers (AccuScan Instruments) with an ~50-lux illumination and 60-dB white noise. Data were analyzed by a one- or two-way ANOVA with post hoc Bonferroni test.

For the rotarod assay, 6 to 18 mice per group were trained to walk on a mouse rotarod (Stoelting) that accelerated linearly from 4 to 40 rpm over 5 min for three trials per day (≥20 min inter-trial interval) on 2 consecutive days. For each trial the mice were placed on the top of the rotarod, the accelerating mode was started, and the total time spent walking on the rod was recorded (maximum of 300 s). Data were analyzed with a two-way ANOVA with post hoc Bonferroni test.

STX209 [R-baclofen, arbaclofen, R-4-amino-3-(4-chlorophenyl)butanoic acid] was dissolved in 0.9% saline and administered via an intraperitoneal injection in an injection volume of 10 ml/kg. STX209 was administered 45 min before testing in the rotarod and audiogenic seizure assays. All statistical measures were performed with Prism (GraphPad Software).

Pharmacokinetics of STX209

Brain and plasma levels of STX209 were determined in C57BL/6J male mice obtained from The Jackson Laboratory. Beginning at P21 (±3), mice received STX209 for 9 consecutive days. In the first study, mice received 1, 3, or 6 mg/kg in twice-daily intraperitoneal injections. Brain and plasma samples were collected at 0.25, 0.5, 1, 2, 8, and 24 hours (three mice per dose per time point). A high-performance liquid chromatography–tandem mass spectrometry (HPLC-MS/MS) (Applied Biosystems API 4000 mass spectrometer with electrospray ionization in positive ion mode, Life Technologies Corp.) bioanalytical method was developed to quantify the levels of STX209 within each matrix. Noncompartmental analysis of plasma and brain concentrations was performed with Phoenix WinNonlin (Pharsight Corporation). Concentrations of STX209 to administer in the drinking water were calculated with a continuous infusion model (where dosing rate = target plasma concentration × clearance/bioavailability). In this second study, STX209 was administered in the drinking water at 0.05, 0.1, and 0.5 mg/ml, and water bottles (both vehicle and STX209 groups) were changed daily. Mice were weighed every 2 to 3 days. Plasma was collected 12 hours before drug removal on the last day of treatment and 0.25, 0.5, 1, 2, and 8 hours (five mice per dose per time point) after drug removal. Plasma was analyzed for STX209, and a noncompartmental analysis was performed as described above to characterize the accuracy of the dosing model. To accommodate comparison between studies, we integrated the plasma-time curve (AUClast) after assigning the −12-hour plasma level to time zero.

STX209 binding to mouse plasma and brain homogenates was performed with rapid equilibrium dialysis (Thermo Scientific). Briefly, the dialyzed samples were extracted by protein precipitation with acetonitrile and reconstituted for analysis by HPLC-MS. A bioanalytical method using electrospray ionization in positive mode with an LTQ Orbitrap XL (Thermo Fisher) in high-resolution mode was used for sample analysis. For receptor interactions, the Ki for STX209 was measured in the rat cortex (Caliper). The EC50 values were determined in CHO cells expressing recombinant GABAB1b receptor (Cerep).

Preparation of polysomes from mouse brain

Fmr1-knockout littermates at 4 to 8 weeks of age were used for each experiment. STX209 (6 mg/kg) or saline vehicle was administered intraperitoneally. Forty-five minutes after injection, mice were anesthetized with isoflurane, and brains were rapidly dissected in an icy slurry of Hanks’ balanced salt solution supplemented with cycloheximide (100 μg/ml), 100 μM leupeptin, and 2 mM vanadyl ribonucleoside complex. Sub-dissected cortices and hippocampi were then Dounce-homogenized (~12 strokes, on ice) in 1 ml of buffer A per brain: 20 mM tris-HCl (pH 7.4), 0.3% Triton X-100, 100 mM KCl, 10 mM MgCl2, cycloheximide (100 μg/ml), 1:400 RNaseOUT (Invitrogen), 0.5 mM dithiothreitol, and EDTA-free protease inhibitor cocktail (Roche Applied Science). The homogenate was centrifuged at 10,000g for 10 min, and 750 μl of the resulting supernatant was combined with an equal volume of buffer A containing 0.7% Triton X-100. Samples were lightly vortexed, placed on ice for 15 min, and then centrifuged at 20,000g for 15 min. The resulting supernatant was split between two 10-ml linear sucrose gradients (10 to 50%) by layering 650 μl onto the top of each gradient in a 4°C cold room. For each experiment, two gradients each were run for the saline- and STX209-treated mice, along with two mock gradients to calibrate baseline for the ultraviolet (UV) detector. The gradients were centrifuged in an SW40 Ti swinging bucket rotor at 18,200 rpm (41,829g relative centrifugal force) for 16 hours, with no braking on deceleration. The distribution of translation machinery within the gradient was measured by taking continuous absorbance readings at 254 nm during upward displacement of the sucrose with Perdenz (Accurate Chemical and Scientific Corporation) through an ISCO UA-6 UV monitor. UV absorbance profiles were scanned and traced at high resolution in Canvas (Deneba, ACD Systems International).

Golgi staining

Male wild-type and Fmr1-knockout littermate mice were given STX209 (0.5 mg/ml) in drinking water or vehicle (drinking water) beginning at P23 (±1). Drinking water, +/−STX209, was changed daily. At P35 (±1), mice were euthanized before cardiac perfusion (peristaltic pump, Thermo Fisher F30) with phosphate-buffered saline, followed by 4% formaldehyde in 0.1 M phosphate buffer (pH 7.4). Dissected brains were stored in 4% formaldehyde at 4°C.

Brains were prepared for single-section rapid Golgi staining and analyzed as described (100). Briefly, hemispheres were weighed and sectioned coronally at 100-μm thickness on a vibratome (Leica, VT1000S) in 0.1 M phosphate buffer (pH 7.4). All analyses of Golgi-impregnated neurons were performed blind to genotype. Apical oblique dendrites with lengths greater than 60 μm from layer 2/3 pyramidal cells in the primary binocular visual cortex were traced using the Neurolucida software (version 8.0; MBF Bioscience). One dendrite was analyzed per cell, and three independent cells were analyzed for each animal (number of mice: wild-type vehicle, 11; Fmr1-knockout vehicle, 6; wild type + STX209, 7; and knockout + STX209, 5). Mean spine density was analyzed by two-way ANOVA with post hoc Bonferroni test.

References and Notes

  1. Acknowledgments: We thank E. Osterweil for advice on the protein synthesis assays, M. Corlew for performing the plasma and brain binding assays, and T. Ocain for critical review of the manuscript. Funding: This work was supported in part by Seaside Therapeutics, NIH grants HD020521 and HD24064 to S.T.W., and Medical Research Council grant G0601584 to P.C.K. Author contributions: C.B., R.S.H., R.P., R.R., S.T.W., P.W.V., P.C.K., R.L.C., M.F.B., and A.M.H. designed the research. C.H., L.W., M.N.K., M.S., F.R.P., C.B., A.T., R.P., and P.W.V. performed the research. C.H., L.W., M.N.K., M.S., F.R.P., C.B., R.S.H., A.T., R.P., R.R., S.T.W., P.W.V., P.C.K., R.L.C., M.F.B., and A.M.H. analyzed the data. F.R.P., C.B., R.S.H., S.T.W., P.W.V., P.C.K., R.L.C., M.F.B., and A.M.H. wrote and revised the paper. Competing interests: R.P. and P.C.K. are or have been paid consultants for Seaside Therapeutics. R.P., S.T.W., R.L.C., and M.F.B. have financial interest in Seaside Therapeutics. C.H., M.S., F.R.P., C.B., R.S.H., R.R., R.L.C., and A.M.H. are employees of Seaside Therapeutics. Seaside Therapeutics is testing STX209 in clinical trials for the treatment of FXS.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article