Acamprosate in a mouse model of fragile X syndrome: modulation of spontaneous cortical activity, ERK1/2 activation, locomotor behavior, and anxiety

For total protein determination, the hippocampus and striatum were homogenized in ice-cold RIPA buffer (500 and 100 μl, respectively) with the fresh addition of HALT phosphatase inhibitor cocktail (ThermoScientific) and protease inhibitor cocktail (Sigma) and assayed using the Pierce BCA Protein Assay Kit (ThermoScientific) according to manufacturer’s instructions. Samples were diluted to 50 μg/ml for phosphorylated ERK1/2 (pERK1/2) and 2.5 μg/ml for ERK1/2 total prior to analysis. pERK1/2 and ERK1/2 total were analyzed by semiquantitative SimpleStep ELISAs (enzyme-linked immunosorbent assay; ABCAM; phosphoERK1/2 pT202/Y204, ab176640 and ERK1/2 total, ab176641) according to manufacturer’s instructions. Briefly, supplied concentrated capture and detector antibody was diluted in supplied antibody dilution buffer. Standards were prepared as directed and 50 μl of samples and standards were added to each well and assayed in duplicate. The optical density (OD) was read at 450 nm. Data were verified to fall within the linear range of the standard curve. These ELISAs are semiquantitative with standards supplied at an unknown concentration of phosphorylated recombinant ERK protein and do not allow for the exact concentration of pERK1/2 or ERK1/2 total. Therefore, mean OD of duplicate samples was used for calculations. ERK1/2 total and the ratio of pERK1/2 over ERK1/2 total normalized to WT + SAL were analyzed by two-way ANOVA with genotype (WT or Fmr1 KO) and drug (SAL, 300 mg/kg acamprosate) as factors. For pERK/ERK total, a priori comparisons between the WT + SAL and KO + SAL groups, and the KO + SAL and KO + Acamp groups were performed with predictions of increased pERK/ERK total ratio in the KO + SAL group compared to the WT + SAL control, and decreased ratio in the treated KO mice compared to SAL-treated KO group in both the striatum and hippocampus. All pair-wise comparisons were corrected using FDR.

60 min following a final treatment dose (2 days following the completion of the adult behavior battery), the animals were deeply anesthetized with pentobarbital and transcardially perfused with 5-mL ice-cold 1× PBS followed by 4% PFA. Whole brains were sectioned coronally using a Leica SM2000R freezing, sliding microtome at 35 μm. Tissue sections were bleached in 3% H₂O₂ for 30 min. Sections were then blocked in 10% normal donkey serum (NDS) for 1 h. Sections were incubated in 1:400 rabbit, anti-pERK1/2 primary antibody (#4370; Cell Signaling) for 48 h followed by incubation in 1:200 swine, anti-rabbit, biotinylated secondary antibody (E0353; Dako) solution for 3 h. Following secondary, tissue was incubated for 1 h in ABC solution (VECTASTAIN Elite ABC HRP Kit; Vector) which was prepared 30 min prior to use. Tissue was then incubated in tyramide biotin solution prepared in 0.1-M Borate buffer, pH 8.0 with 0.003% H₂O₂ for 10 min. Tissue was then incubated with 1:200 Alexa 488 conjugated streptavidin (Jackson ImmunoResearch) for 2 h. Sections were then placed in 1:500 mouse, anti-NeuN primary antibody (MAB377; Milllipore) solution overnight. Sections were then incubated in 1:200 donkey anti-mouse Alexa 594 conjugated secondary antibody (Jackson ImmunoResearch) for 2 h. All steps were performed at room temperature. Sections were washed between incubations 3 times in 1× KPBS with 0.2% Triton X-100 for 10 min per wash. All antibody solutions were prepared in 1× KPBS with 0.2% Triton X-100 and 2% NDS. Images were acquired using a Nikon A1 inverted, single photon, confocal microscope, using a 4× objective with pixel size minimized to the Niquist limit. Images were taken from sections at −2.5 mm from Bregma, and pERK1/2 positive cells were identified using the General Analysis functionality in NIS-Elements. ROIs were then manually applied and pERK1/2 positive nuclei were automatically counted using NIS-Elements. Neuronal identity of cells was assessed by colocalization of pERK1/2 with NeuN.

For the groups of mice that were assessed in the adult behavior battery (and subsequent pERK1/2 immunostaining), male WT and Fmr1 KO littermates (5–7 months old) were randomly assigned to a treatment group and treated once daily with 0 (SAL vehicle), 300 mg/kg of acamprosate calcium (expressed as the free base), or 122.2 mg/kg calcium chloride USP (CaCl₂ × 2H₂O; Sigma-Aldrich) in a volume of 10 ml/kg via IP injection. Note that calcium salt and acamprosate calcium contained equivalent amounts of Ca²⁺ ions (0.8 mmol/kg/day). Dosing commenced 10 days prior to, and continued throughout behavior testing. Drug treatment occurred between 0900 and 1100 h with an interval of 60 min between drug treatment and the start of behavior assessment each day. Mice were treated for a total of 21 days (9–13 mice per treatment group were tested). Adult behavior analysis was completed in two separate cohorts with genotype and drug group combinations balanced across cohorts. Data are shown as single treatment groups since no differences between cohorts were apparent.

The dose used in the current study was based on previously published reports in rodents which demonstrated that > 100 mg/kg was needed to reduce alcohol craving and nicotine-seeking behavior, and 200 mg/kg was required to improve transient hemispheric ischemia-induced neurological deficits [69, 70]. The therapeutic dose of acamprosate for alcohol withdrawal and the current adult FXS treatment dose is ~2 g/day for an average 70 kg human subject (equivalent to 28.5 mg/kg). Using the human equivalent dose based on body surface area calculation for inter-species dose scaling, the daily mouse adult behavior battery dose (300 mg/kg; free base) is equivalent to 1.9 g/day in a 70 kg human ((333 mg/kg × 3/37 (mouse to human ratio) = 27) × 70 kg adult = 1.9 g dose).

Behavior was assessed during the light portion of the light/dark cycle, and food and water were available ad libitum except during behavior testing. Mice began testing on day 11 of treatment. To minimize the impact of stress during behavioral testing, mice were transported across the hallway to the Rodent Behavior Core and dosed with SAL, CaCl₂, or acamprosate and allowed at least 60 min in the testing room to acclimate before behavior assessment daily. Elevated zero maze was the only exception in which mice were brought into the testing room one at a time just prior to being placed on the maze in order to get an accurate anxiety assessment. Animals were tested in only one paradigm per day and were given at least 1 day of rest in between each test (drug treatment continued even on resting days). Behavior was evaluated in the following order so that tests easily influenced by stress were completed early during the behavior battery: elevated zero maze, locomotor activity, novel object recognition, acoustic startle habituation, and prepulse inhibition. Apparatus surfaces were cleaned with Process NPD (Steris) before and between animals.

The EZM was used to assess anxiety-like behavior as previously described with modification of the maze size [71]. Briefly, mice were transported from the housing room to the testing room individually and placed on the apparatus. The experimenter exited the room immediately after placing the mouse in one of the closed quadrants of the apparatus. A camera mounted above the maze connected to a computer located outside the room was used to observed and score, in real-time, time in open quadrants, number of head dips, number of open arm entries, and latency to first enter an open quadrant during a single 5 min trial (ODLog, Macropod Software). The test room was dimly lit (30 lux (lx)) to encourage exploration of the test environment. Two mice were removed from the EZM analysis after falling from the maze.

Activity analysis in an open field, an overall indication of an animal’s activity level, is sensitive to sedative drugs or those inducing stereotypy or catatonia, and is especially useful in better interpreting other tasks that depend on the overall activity of the animal. Locomotor activity was measured in infrared photocell activity chambers (41 × 41 cm; PAS Open Field, San Diego Instruments, San Diego, CA) for 1 h. Number of beam breaks was recorded during 5 min intervals for a total of 12 intervals and analyzed by three-way ANOVA with repeated measures. Room lights were at full level (1200 lx).

A solid black enclosure with dimensions 19.5 cm L×40 cm W×35 cm H was used to assess NOR. During the familiarization phase, mice were presented with two identical objects for a total of 5 min. Mice were returned to their cage and left undisturbed for 30 min. Next, mice were placed back in the enclosure with a novel object and one identical copy of the familiarization phase objects. Pilot mice were previously shown to have no inherent preference for familiar or novel objects used in this test (data not shown). The amount of time each mouse spent paying attention to the familiar and novel objects during the familiarization and test phases was recorded using OD Log (Macropod Software) for the 5 min duration of each phase. Time spent paying attention was recorded when the mouse was oriented toward the object with snout within 1 cm of the object or when forepaws were up against the object. Mice in these cohorts did not climb on top of the objects used for this test. The discrimination index (DI; novel object time—familiar object time/novel object time + familiar object time) was used to determine the degree of object memory. Dim lighting conditions (20 lx) were used to reduce anxiety and encourage object exploration during both phases. Six mice were removed from the NOR analyses due to accumulating less than 6 s of total time paying attention to the objects during the test phase. Total exploration time and DI during the test phase were analyzed separately by two-way ANOVA.

Acoustic startle habituation and PPI were assessed in a sound-attenuating test chamber (SR-LAB apparatus; San Diego Instruments, San Diego, CA) as previously described with modifications [72]. Mice were placed in an acrylic cylindrical holder that was mounted on a platform with a piezoelectric force transducer attached to the underside of the platform. For both habituation and PPI, a 5 min acclimation period preceded test trials. For habituation, each animal received 50 repeated 20 ms 120 dB SPL mixed frequency sound bursts (1.5 ms rise time). Maximum velocity for each trial (V_max; measured in arbitrary units; a.u.) was analyzed by repeated measures three-way ANOVA. For PPI, each animal received a 5×5 Latin square sequence of trials that were of five types: startle stimulus (SS) with no prepulse (PPI0), no SS with no prepulse, 73 dB prepulse + SS, 77 dB prepulse + SS, or 82 dB prepulse + SS. The startle signal was a 20 ms 120 dB SPL mixed frequency sound burst (1.5 ms rise time). Prepulses preceded the startle-eliciting stimulus by 70 ms (onset to onset). The startle recording window was 100 ms. Background noise level was 70 dB. Each set of 25 trials was repeated 4 times for a total of 100 trials. The inter-trial interval averaged 14 s and varied randomly from 8–20 s. Percent PPI was calculated as (100*(V_max at PPIxx/max velocity PPI0) for the PPI trials. Percent PPI at each prepulse level was analyzed by three-way mixed factor ANOVA with gene and drug as between factors and PPI Trial Type as a within factor (Table 1). Two mice were removed from the startle habituation analysis and one removed from the PPI analysis due to equipment errors in data recording (i.e., no data recorded by software).

Juvenile Fmr1 KO neocortical circuits are hyperexcitable as revealed by the long duration of spontaneous persistent, activity, or UP states of neuron networks [25]. Here we measured UP states with extracellular, multiunit recordings in layer IV of acute slices of somatosensory, or barrel, neocortex from WT or Fmr1 KO mice littermates with bath application of acamprosate or vehicle (Fig. 1a). Duration and amplitude for each UP state as well as the number of UP states during the five-minute time-period were analyzed by two-way ANOVA (Additional file 1: Table S3) with pairwise differences corrected using FDR (two-tailed; Fig. 1). For duration of UP states (Fig. 1a), there was a significant main effect of gene (ANOVA, F(1, 78) = 4.71, p = 0.0001) and drug (ANOVA, F(1, 78) = 15.74, p = 0.0002). As previously reported [25], UP state duration was greater in the KO + VEH group compared to the WT + VEH group (p = 0.0002). Acamprosate treatment in the KO mice reduced this increase compared to the KO + VEH (p = 0.0002), although this was still slightly elevated compared to the WT + VEH mice (p = 0.049; see Fig. 1d for representative traces). Acamprosate treatment in the WT mice produced a trend towards a decrease in duration compared to the WT + VEH group (p = 0.071) and a significant decrease compared to the KO + VEH (p = 0.0002) and the KO + Acamp groups (p = 0.0002). No significant effects were found for amplitude normalized to detection threshold (Fig. 1b). For number of events in 5 min (Fig. 1c), there was a main effect of gene (ANOVA, F(1, 78) = 5.14, p = 0.026) although pairwise differences were not evident in pertinent group comparisons (WT + Acamp vs. KO + VEH group (p = 0.035)). These data indicate that hyperexcitability of neocortical circuits in the developing Fmr1 KO mice, as measured by prolonged UP states, is improved by acamprosate treatment.

Juvenile Fmr1 KO mice are susceptible to audiogenic-induced seizures although WT mice (B6 background) of all ages and adult KO mice are resistant. A pilot experiment using 300 mg/kg acamprosate failed to attenuate seizure susceptibility (data not shown) and therefore the higher dose of 500 mg/kg was chosen for this experiment. In the current study, seizure severity score was analyzed in P21 Fmr1 KO and WT littermates following 5 days of SAL or acamprosate (500 mg/kg) treatment using the Wilcoxon statistic, S = 175.5, and demonstrated a significant effect of treatment group (p = 0.0004) (Fig. 2). Exact probabilities were computed to determine pairwise comparisons corrected using FDR (two-tailed) and revealed significant increases in seizure severity score in both FXS groups compared to each WT group (p = 0.003 for each comparison). No within-genotype differences were detected indicating acamprosate treatment did not alter seizure severity in either the WT or KO mice, although a baseline difference was detected between control-treated KO and WT mice as expected.

A three-way mixed factor ANOVA with gene and drug as between factors and segment as a within factor (Additional file 1: Table S4) was used to analyze spine number along the first 150 μm length of apical dendrites divided into six 25 μm segments from layer V pyramidal neurons located in the somatosensory cortex in adult mice (n = 25 cells/group). There was a significant main effect of segment (ANOVA, F(5, 460) = 87.36, p = 0.0001) in which the number of spines in all groups increased as a function of distance from the soma (Fig. 3c). Gene × drug (Fig. 3b) and drug x segment interactions were only approaching significance and therefore additional post hoc analyses were not completed. These data indicate that there were no observable spine differences detected between the control-treated KO and WT mice and therefore no deficit for acamprosate to modulate.

Separate two-way ANOVAs (Additional file 1: Table S3) were used to determine the effects of gene and drug and the interaction of gene × drug in the hippocampus and striatum on pERK/ERK total ratio and ERK1/2 total (each region was normalized to WT + VEH; n = 6 per group and brain region). All pairwise group comparisons were corrected using FDR. For ERK1/2 total absorbance, no significant main effects or interactions were identified in the hippocampus (Fig. 4b) or striatum (Fig. 4d), demonstrating that neither genotype nor drug altered ERK1/2 total protein expression. Therefore, group differences in ERK1/2 activation/phosphorylation are not influenced by baseline changes in total ERK1/2 expression and can be attributed to changes in ERK activation. For pERK/ERK total ratios, there was a significant main effect of gene in the hippocampus (ANOVA, F(1, 20) = 6.06, p = 0.023) (Fig. 4a) and a main effect of drug in the striatum (ANOVA, F(1, 20) = 5.89, p = 0.02) (Fig. 4c). We predicted baseline increases in pERK/ERK total ratios in the KO + SAL group compared to the WT + SAL group based on previous reports in which ERK1/2 activation has been shown to be elevated in the brains of Fmr1 KO mice compared to WT mice [39, 74]. Furthermore, we predicted acamprosate treatment would decrease pERK/ERK total ratios based on data showing drugs with similar anti-glutamatergic actions to acamprosate significantly decreased aberrant ERK1/2 activation in Fmr1 KO mice and decreased ERK1/2 activation kinetics in FXS patient blood samples [7, 75]. Because our a priori predictions were directional for these specific comparisons (WT + SAL vs. KO + SAL; KO + SAL vs. KO + Acamp), one-tailed tests were used for these specific ERK1/2 preplanned tests. Baseline comparisons showed a significant increase in pERK/ERK total ratio in the KO + SAL group compared to the WT + SAL group in both the hippocampus (p = 0.008) and striatum (p = 0.035) which is in line with previous reports. Preplanned comparisons between the KO + SAL and the KO + Acamp mice showed a reduction in pERK/ERK total ratio in both the hippocampus (p = 0.026) and striatum (p = 0.03) with acamprosate treatment as predicted. When comparing the KO + SAL-treated mice to the WT + Acamp-treated mice, there was a trend toward a pERK/ERK total increase in the hippocampus (p = 0.05) and a significant increase in the striatum (p = 0.04). No differences were noted in pERK/ERK total ratio in the hippocampus or striatum between the two WT groups (p = 0.71 and p = 0.43, respectively).

To determine if acamprosate modulated ERK1/2 activity in a region/cell type specific manner, we immunostained brain sections from mice that completed the adult behavior battery. Data were first analyzed to determine if there were any within genotype differences in pERK1/2+ cell counts in mice treated with either SAL or CaCl₂ and found no differences in the dentate gyrus (DG), auditory cortex, or visual cortex (Additional file 1: Figure S2). Since there were no effects of CaCl₂ treatment on either WT or KO mice compared to SAL-treated mice, these groups were combined to create a single control group. A two-way ANOVA for cell counts revealed a main effect of drug (ANOVA, F(1, 30) = 7.59, p = 0.01) in the DG (Fig. 5a, e, f), but no effects in cortical regions (Fig. 5b, c). In the DG, baseline differences between genotypes in pERK1/2+ cell counts demonstrated a trend showing an increase in pERK1/2+ nuclei in the KO_Controls compared to the WT_Controls (p = 0.09). This finding is consistent with our above data in hippocampal lysates. Likewise, acamprosate treatment reduced the number of pERK1/2+ cells in KO mice compared to KO_Controls in the DG (p = 0.024). This change was driven by decreases in the number pERK1/2+ neurons in the granule cell layer as evidenced by nuclear co-localization of NeuN in all pERK1/2+ cells in the DG. This suggests that acamprosate can affect neuronal ERK1/2 activation in a manner likely to alter neuronal signal transduction.

An initial analysis was completed for all behavior paradigms and dependent measures assessed in the adult behavior battery comparing only the two control groups (i.e., SAL- vs. CaCl₂-treated mice). Complete F statistics are presented in Additional file 1: Tables S1 and S2. No main effects of drug or drug interactions were observed, indicating that CaCl₂ treatment did not alter the behavior of WT or KO mice compared to those treated with SAL in any test (see Additional file 1: Figure S1). There were significant effects of Genotype, which are further detailed below. Four groups were compared in the final analysis of the behavior battery: (1) WT_Controls (WT + SAL and WT + CaCl₂ combined), (2) KO_Controls (KO + SAL and KO + CaCl₂ combined), (3) WT + Acamp, (4) KO + Acamp.

Elevated zero maze was used to assess anxiety behavior in control (SAL- and CaCl₂-treated) and Acamp-treated Fmr1 KO and WT mice during a 5-min test. Separate two-way ANOVAs were used to analyze time in open (primary anxiolytic measure), latency to first open arm entry, number of head dips, and number of open arm entries in the EZM (Fig. 6). Pairwise comparison testing using FDR correction (two-tailed) was performed for significant main effects. For time in open, there was a significant main effect of gene (ANOVA, F(1, 60) = 12.41, p = 0.001) and drug (ANOVA, F(1, 60) = 6.32, p = 0.015; Fig. 6a). Pairwise comparisons showed a significant increase in time in open observed in the open quadrants for the KO_Controls group compared to the WT_Controls group (p = 0.031) indicating an observable baseline difference between the two genotypes. In the KO mice, acamprosate treatment further increased time spent in the open quadrants compared to the control-treated KO mice (p = 0.049). This increase in the KO + Acamp group was also increased compared to both WT groups (vs. WT_Controls p = 0.001; vs. WT + Acamp p = 0.031). For head dip frequency (ANOVA, F(1, 60) = 10.39, p = 0.002; Fig. 6c) and number of transitions from dark to light quadrants (ANOVA, F(1, 60) = 5.88, p = 0.018; Fig. 6d), there was also a main effect of gene. For number of head dips, the gene main effect was driven by an increase in head dips in both the KO_Controls (p = 0.039) and KO + Acamp (p = 0.035) groups compared to the WT_Controls. The number of open arm entries was increased in the KO + Acamp mice compared to the WT_Controls (p = 0.038) which is consistent with the increase in time spent in the open that was observed for the KO + Acamp group. No significant effects were observed for latency to first open arm entry (Fig. 6b), indicating all mice began exploring the maze at similar times. No other main effects or interactions were noted (see Additional file 1: Table S3 for complete F statistics). Taken together, these data indicate that there was a baseline difference between the KO and WT mice and that acamprosate treatment resulted in an observable behavioral change that is consistent with an anxiolytic effect in only the KO mice.

A three-way repeated measures ANOVA (auto regressive (AR) (1)) for number of beam breaks revealed main effects of interval (ANOVA, F(11, 646) =2.41, p = 0.006) and a significant gene×drug interaction (ANOVA, F(1, 114) = 7.06, p = 0.009) during the 60-min test (Additional file 1: Table S4). Since there were no interactions with interval (Fig. 7a), FDR-corrected pairwise comparisons (two-tailed) were performed on data collapsed across time (Fig. 7b). There was a significant baseline increase in beam breaks in the KO_Controls group compared to the WT_Controls group (p = 0.003). Acamprosate treatment in the KO mice reduced this increase compared to KO_Control mice (p = 0.023) such that there was no difference between WT_Controls and KO + Acamp mice (p = 0.84). These data indicate that there was a significant baseline difference between the KO and WT mice and that acamprosate treatment normalized open field behavior in the KO mice.

Separate two-way ANOVAs (Additional file 1: Table S3) were used to analyze test phase total object attention time and test phase discrimination index (DI) in a short-term object recognition test [76]. During the test phase of NOR, there were no group differences between the total time the mice paid attention to the two objects with the average time being 46.46 ± 3.4 s for WT_Controls, 46.7 ± 3.5 s for KO_Controls, 33.68 ± 5.5 s for WT + Acamp, and 47.43 ± 4.7 s for KO + Acamp (data not shown). There were no main effects or interactions noted for DI (time with the novel object–time with familiar object/time with the novel object + time with the familiar object), nor were there any significant differences between any individual groups (DI LSmean ± SEM, n): WT_Controls = 0.29 ± 0.04, n = 21, WT + Acamp = 0.31 ± 0.07, n = 8; KO_Controls = 0.27 ± 0.04, n = 20; KO + Acamp = 0.26 ± 0.06, n = 11; data not shown. All groups spent more time with the novel object (indicated by a DI greater than zero) which suggests that both KO and WT mice were able to remember the familiar object. These data indicate that there was no observable difference in object recognition memory between the control-treated KO and WT mice in this experiment and therefore no deficit to be corrected by acamprosate treatment.

An acoustic startle habituation protocol was utilized to determine if there were differences between WT and KO mice in startle habituation and to acclimate the mice to the chamber and tones for the PPI test assessed 2 days later. A three-way repeated measures ANOVA (Additional file 1: Table S4; auto regressive (AR) (1)) for V_max revealed a main effect of drug (ANOVA, F(1,60) = 4.37, p = 0.041). However, pairwise comparisons failed to reach significance with FDR correction, indicating little effects of gene or drug on startle habituation in 5–7-month-old mice (Fig. 7c). These data indicate that there was no difference between control-treated WT or KO mice in this acoustic startle habituation test and therefore no deficit that required correction.

PPI has been shown to be impaired in young males with FXS, but enhanced in adult male mice [77]. Although the reasons for these discrepancies are unknown, it is clear that both mice and people lacking FMRP exhibit aberrant sensorimotor gating [77, 78]. PPI is a test of startle reactivity and sensorimotor gating and was the final behavior test assessed in the adult behavior battery. PPI was calculated for each animal at each of the prepulse trial types, and a three-way mixed factor ANOVA with gene and drug as between factors and trial type (PPI73, PPI77, PPI82: PPIxx) as a within factor was used. The omnibus ANOVA did not reveal any main effects or interactions for % PPI (Additional file 1: Table S4). There was a trend for a drug×trial type interaction although not significant. Data are shown collapsed across trial type since no interaction of prepulse was detected (Fig. 7d). No differences were detected between control-treated KO and WT mice or in the groups that received acamprosate, suggesting that all groups were similarly able to inhibit the startle response when a prepulse preceded the startle stimulus.

Spanagel et al. has suggested that the anti-relapse properties of acamprosate (the calcium salt of N-acetylhomotaurinate) and neuroactivity of the molecule are solely due to calcium rather than N-acetylhomotaurinate since an equimolar concentration of a corresponding sodium salt of N-acetylhomotaurinate produced no reductions in alcohol consumption while calcium chloride at equimolar calcium concentrations produced effects similar to acamprosate [53]. It was also suggested that alcohol dependent patients with high plasma calcium levels following treatment with acamprosate had better treatment responses. Although plasma calcium levels in FXS have not been reported to date, FMRP has been shown to regulate several calcium-binding proteins involved in activity-dependent calcium signaling and has been shown to regulate calcium signaling dynamics during development in the dfmr1 null mutant Drosophila FXS disease model [79‐82]. As such, the implications that the effects of acamprosate may be reliant on calcium rather than N-acetylhomotaurinate would have significant implications for the future drug development of acamprosate for the treatment of FXS. In the current study, we found that an equimolar concentration of calcium salt, alone, did not produce any effects that were significantly different from saline-treated mice in any behavior paradigm or in any brain regions assessed for pERK1/2 immunostaining in either WT or KO mice. Furthermore, when a treatment effect of acamprosate was observed in the KO mice (EZM, open field, pERK1/2 immunostaining) we did not observe any acamprosate-like effects in the CaCl₂ group suggesting that the treatment effects of acamprosate in FXS are not due to calcium. Mann et al. recently conducted a study on calcium plasma levels from alcohol dependent patients and showed that there were no differences between acamprosate and placebo-treated patients and that the effect of calcium plasma concentrations on severe relapse was always non-significant. These results also fail to support the hypothesis that calcium is the active moiety of acamprosate [62]. In the current experiments, it is unlikely that differences in calcium bioavailability or elimination rates are likely to affect our results since Chabernat et al. demonstrated that salts of the N-actylhomotaurinate molecule become totally dissociated in hydrophilic media. Since CaCl₂ is also a hydrophilic molecule, this suggests that the similar amount of Ca2+ ions in both the acamprosate and CaCl₂ doses used in our current experiments should result in similar Ca2+ bioavailability and elimination rates [83].

With our behavior data demonstrating no differences between SAL and CaCl₂ treatment, it is unclear why CaCl₂ had effects on alcohol-seeking behavior as previously reported; however, it is possible that a CaCl₂ injection may cause some physical discomfort over and above saline treatment due to stinging or burning at the injection site [84]. The mice in our study were treated once-a-day for 10 days prior to behavior testing whereas the rats in the Spanagel et al. paper were injected only twice within 12 h before ethanol intake was assessed. The pain/discomfort from the CaCl₂ injection may have been sufficient to prevent alcohol seeking whereas in our study, mice may have acclimated to the CaCl₂ injection, or alternatively, the behavior assessments we conducted were less severely influenced by pain. Although our studies are not able to explain the outcomes of the Spanagel et al. paper, they do suggest that acamprosate rather than calcium may have treatment utility in FXS.

UP states are a spontaneous, oscillatory (0.5–1 Hz), synchronized firing of neocortical neuron networks driven by recurrent excitatory and inhibitory synaptic circuits and provide a readout of the intact functioning of neocortical circuits [85, 86]. The examination of spontaneous cortical UP states in the current experiment found prolonged UP state duration in control-treated KO mice compared to control-treated WT mice as expected. Importantly, acamprosate treatment in the KO mice reduced this exaggerated UP state duration. It is thought that the increase in Fmr1 KO UP state duration is indicative of altered recurrent excitatory signaling or response to signaling through mGluR5 receptor stimulation, as the increased duration remains in the presence of GABA receptor antagonists and is restored to normal by genetic reduction of mGluR5 in Fmr1 ^-/y mice and by the mGluR5 receptor antagonist, MPEP (2-methyl-6-(2-phenylethynyl)pyridine) [25, 87]. Furthermore, Hays et al. demonstrated that depletion of Fmr1 in glutamatergic neurons but not GABAergic neurons was sufficient to detect increased UP state duration. Acamprosate is suggested to reduce neuronal hyperexcitability, by potentially acting on both glutamate and GABA systems [55, 88‐91]. Future work may clarify the mechanism by which acamprosate improves excessive spontaneous cortical activity in Fmr1 KO mice and to determine if systemic drug treatment has similar effects in vivo.

It has been suggested that Fmr1 KO-associated increased duration of UP states may contribute to the increased audiogenic seizure susceptibility of juvenile Fmr1 KO mice, although this has yet to be directly studied. In the current study, we did not observe any reduction in seizure severity score following 5 days of acamprosate treatment in P21 Fmr1 KO mice. This effect could indicate that spontaneous UP state duration does not directly contribute to seizure susceptibility following intense auditory stimulation. Many non-cortical brain regions are involved in auditory processing, auditory induced seizure behavior, and have been shown to be altered in the Fmr1 KO mice. Altered spontaneous UP states may not be a critical determinant in AGS susceptibility in these mice, but may contribute to other aberrant behavior in KO mice [92‐94]. It is also possible that systemic administration of acamprosate may not have a similar effect on UP state duration as observed in slice application of the drug. It is also possible that a systemic dose of 500 mg/kg of acamprosate may not result in drug concentrations nearing 200 μM in the brain as was bath applied in the UP state study. Additionally, attenuation of UP state duration in these mice may not be sufficient to abrogate increased seizure susceptibility in the AGS test. More work is needed to fully understand any possible connections between FXS-related UP state dysfunction and seizure susceptibility.

The ERK1/2 signaling cascade plays critical roles in brain development and behavior [28]. In neurons, the ERK1/2 cascade is activated by synaptic activity. In turn, ERK1/2 phosphorylates numerous proteins involved in a diverse number of cellular processes including translational and transcriptional regulation, long-term potentiation and depression, and synaptogenesis [30, 95]. In the brain, critical control over temporal and spatial ERK1/2 regulation (nuclear and cytoplasmic), both activation and deactivation, are required for appropriate behavior, and can contribute to maladaptive behavior and central nervous system (CNS) disorders [96‐99]. In the first ERK1/2 study (tissue lysates), we observed a ~20% increase in hippocampal and striatal ERK1/2 activation from SAL-treated Fmr1 KO mice compared to SAL-treated WT mice. This effect has been observed by others using similar techniques [7, 36, 37]. Chronic acamprosate treatment significantly reduced ERK1/2 activation in lysates from both brain regions assessed in acamprosate-treated KO mice compared to control-treated KO mice, indicating a treatment effect. The hippocampus and striatum data characterize ERK1/2 activity in a variety of cell types and throughout the cells (including cytosolic and nuclear ERK1/2) of the regions dissected. Once ERK1/2 is activated in the cytoplasm, it travels to the nucleus where it can then phosphorylate other target proteins and inhibit or activate transcription of many genes [100]. In the second ERK1/2 experiment, the number of cells expressing activated nuclear ERK1/2 immunoreactivity was found to be reduced by acamprosate treatment in the DG, although a difference between control-treated KO and WT mice was only approaching significance with a corrected one-tailed test. The pERK1/2 positively stained cells in these brains were relatively sparse (with no staining in the striatum) and likely represent only those cells with the highest level of nuclear ERK1/2 activity. Nonetheless, we found that in the DG (where we saw a pERK1/2+ cell reduction in acamprosate treated mice), all pERK1/2+ cells were also NeuN+, suggesting that systemic acamprosate treatment modulated neuronal ERK1/2 activity in a cell type- and region-specific manner. Furthermore, CaCl₂ treatment did not mimic this effect and was indistinguishable from the KO + SAL mice. To our knowledge, these are the first data to suggest that acamprosate modulates central ERK1/2 signaling in vivo and that this change occurs to some degree in the nucleus.

These data are particularly interesting due to the suspected contribution of altered ERK1/2 signaling in FXS and autism pathophysiology. In human study, ERK1/2 activation kinetics following stimulation with phorbol ester have been demonstrated to be delayed in persons with FXS compared to controls [101]. Excessive basal levels of ERK1/2 activation have been reported in FXS mice and in human FXS post-mortem study [39]. In ASD, ERK1/2 dysregulation has been noted in animal model study [102], genetic study [103‐105], and in human post-mortem brain study where enhanced ERK1/2 activation has been reported [106]. We and others have shown increases in basal ERK1/2 activation and rescue with various treatments including other GABA and glutamate modulators. Normalization of delayed ERK1/2 activation kinetics with riluzole treatment (glutamate and GABA modulator) was observed in adults with FXS [75]. Both upstream modulators driving increased ERK1/2 activation and the mechanisms by which acamprosate alters ERK1/2 activity in FXS are unknown. However, we have previously shown that acamprosate reduced plasma APP total and secreted APPα levels (sAPPα) in human subjects with FXS [40]. Since APP can induce ERK1/2 activation in vitro [42], there may be a link between the observed effects of acamprosate on APP and ERK1/2 activation in FXS. Furthermore, ERK1/2 activation is thought to be overactive during alcohol withdrawal and suggested to contribute to alcohol dependence and neuronal hyperexcitability associated with chronic alcohol exposure [107]. These data suggest that overactive ERK1/2 signaling associated with other conditions may be attenuated by acamprosate treatment and that one mechanism of acamprosate treatment for alcohol dependence may involve changes in ERK1/2 activation.

Our data and others suggest that central and peripheral ERK1/2 activity in the blood and brain are responsive to neuroactive compounds (including acamprosate). However, more work is needed to determine if these changes impact behavior in a significant way and to what extent ERK1/2 activity can or should be used as a biomarker in FXS. Currently, ERK1/2 activation alterations are being piloted as a biomarker for treatment response and may help identify certain individuals who may respond better to an ERK1/2-modifying drug. Although reduced ERK1/2 activation is typically viewed as the goal of pharmacological treatment in FXS, ERK1/2 signaling abnormalities in FXS are likely much more complicated. Kim et al. demonstrated that in response to synaptic mGluR stimulation, ERK1/2 phosphorylation is rapidly decreased due to over-activated protein phosphatase 2A activity in Fmr1 KO synaptoneurosomes, whereas in WT samples the opposite occurs resulting in increased phosphorylation/activation [108]. As such, future work is needed to better understand aberrant ERK1/2 signaling abnormalities in FXS, specifically related to cell type, intracellular location, and circuit dysfunction in both drug naïve mice and following pharmacological treatment. It is also critical to determine to what degree any CNS changes in ERK1/2 activity manifest in the type of blood-biomarker samples used in clinical trials.

In the adult behavior battery, we studied the baseline differences between Fmr1 KO and WT mice in several behavior paradigms and identified genotype differences (WT_Controls vs. KO_Controls) in the EZM and locomotor activity tests. KO mice spent an increased amount of time in the open quadrants of the EZM, suggesting reduced anxiety (opposite of the human phenotype) and were more active in the open field test (hyperactivity and ADHD symptoms are common in individuals with FXS) [109, 110]. Interpretation of rodent EZM or related elevated plus maze data must take locomotor behavior into consideration, since mice that are hyperactive will tend to spend more time in the open quadrants due to increased locomotion. It is possible that the observed increased time in open that is routinely observed in Fmr1 KO mice, here and by others, is the result of increased locomotor behavior rather than the result of anxiety or risk-taking behavior, although this finding is difficult to reconcile with the human condition [111]. Although we show that treatment with acamprosate further increased time in the open while also reducing open field locomotor behavior in the KO mice, we are unable to determine if treatment reduced anxiety or exacerbated a preexisting abnormality. Interestingly, acamprosate treatment in rodents has been previously associated with anxiolytic properties. In an amphetamine withdrawal-evoked anxiety rodent model, acamprosate treatment increased time in open in the elevated plus maze without a change in locomotor behavior. Another group found that acamprosate reduced social anxiety in a combination stress/ethanol withdrawal rodent model, further supporting the drug’s utility at alleviating anxiety in a manner pertinent to humans with FXS [112, 113]. Koltunowska et al. suggested that this anxiolytic effect of acamprosate may be due to its effects at mGluR receptors which is thought to be a key player in FXS pathophysiology [6]. In human study, open-label treatment with acamprosate in persons with chronic anxiety resulted in reduced anxiety ratings suggesting that acamprosate may modify anxiety behavior although blinded, controlled studies are required to make an accurate determination in this regard [114]. Although the current Fmr1 KO mouse anxiety data are difficult to interpret, taken together with previous reports in other rodent models and humans with FXS, acamprosate may have utility as an anxiolytic agent in FXS.

Locomotor behavior is not only useful for ensuring proper interpretation of other rodent behavior tests reliant on the movement of the animal but it can also be used to gage baseline levels of hyperactivity. The increased baseline locomotor behavior in Fmr1 KO mice observed in the current study is consistent with previous data in KO mice as well as well in persons with FXS [115‐117]. The attenuation of increased locomotor activity in KO mice with acamprosate treatment is also consistent with our study of acamprosate treatment in person with FXS in which hyperactivity/ADHD symptoms were improved [41]. However, an important distinction must be made between our mouse data and the data that is gathered in many FXS treatment studies related to ADHD symptoms. Open field behavior does not assess ADHD symptoms, but rather the physical activity and movement of mice in a novel environment. One cannot assume that attentional deficiencies in persons with FXS will be improved simply based on reductions in locomotor behavior in rodents. For future clinical trials, the use of wearable activity trackers may improve the translational value of rodent locomotor behavior improvements in FXS studies.

Several experiments did not reveal differences between the control-treated KO and WT mice and subsequently conclusions about the treatment effects of acamprosate could not be made in these instances. These tests included object recognition memory, acoustic startle reactivity, prepulse inhibition of the acoustic startle response, and assessment of dendritic spine morphology. Deficits/differences in Fmr1 KO mice have been observed in these types of experiments previously, but can be difficult to replicate. The experimental parameters are critical determinants in identifying phenotypic deficits in all rodent models, not just Fmr1 KO mice [118, 119]. For behavior studies, these can include details such as the age of mice at testing, background strain, maternal genotype, loudness/duration of tones, behavior test order, degree of animal handling, inclusion of a pharmacological treatment, injection/treatment exposure route (gavage, IP, food additive), duration of treatment, age at treatment, environmental enrichment, and housing conditions (barrier vs. conventional housing). Cellular and molecular experiments can also be influenced by many experimental parameters including cellular sub-region analyzed (apical vs. basal dendrite/primary vs. secondary branches), methodology of quantification, antibody used, dissection procedure, previous exposure to behavior testing (can function as environmental enrichment condition), staining/imaging techniques, ex vivo vs. culture systems, method of tissue collection/processing (sacrifice method: anesthesia vs. no anesthesia, delay between disruption of the mice and actual time of tissue collection), age at sacrifice. This list is not meant to be exhaustive but meant to highlight the many details that play a role in types of tests commonly used to decipher positive drug effects in FXS translation drug development. Some parameters are at the discretion of the investigator while others are imposed by equipment available or vivarium constraints. In many instances, it is unclear which parameters specifically lead to a significant difference between WT and Fmr1 KO mice making it difficult to guarantee a particular method will lead to genotype differences at the outset of a preclinical treatment study. In the current study, it is unclear if the age of the mice at testing had any significant effect on a lack of phenotype in NOR or in the acoustic startle tests between the WT and KO mice. Furthermore, a broader characterization of dendritic spine differences may have revealed genotype differences or drug effects. Nonetheless, extrapolation pertaining to the effects acamprosate may have on cognition, sensory reactivity, and gating in humans can not be made from the current results.

The dose used for the adult behavior battery (300 mg/kg) closely matches the clinical dose based on body surface area calculations (see methods for additional information) however, the half-life of acamprosate has been shown to be species dependent. The half-life of acamprosate in humans is approximately 18–32 h following oral administration with 5–7 days of treatment required to reach steady-state plasma concentrations. In rodent plasma, acamprosate has an elimination half-life of 132 ± 56 min, and in brain this can be as short at 43.33 ± 9.55 min [120]. Therefore, the timing of the behavioral tests (1 h following treatment) was chosen to allow mice to recover from the treatment injection while still assessing behavior prior to drug elimination. Furthermore, chronic administration of acamprosate in rodents has been shown to result in increased extracellular brain concentrations of the drug relative to a single treatment suggesting that repeated administration may be needed to achieve clinical efficacy and supports the chronic treatment paradigm used in the current in vivo tests [121].