Atypical electrophysiological and behavioral responses to diazepam in a leading mouse model of Down syndrome

Mounting evidence implicates dysfunctional GABAAR-mediated neurotransmission as one of the underlying causes of learning and memory deficits observed in the Ts65Dn mouse model of Down syndrome (DS). The specific origin and nature of such dysfunction is still under investigation, which is an issue with practical consequences to preclinical and clinical research, as well as to the care of individuals with DS and anxiety disorder or those experiencing seizures in emergency room settings. Here, we investigated the effects of GABAAR positive allosteric modulation (PAM) by diazepam on brain activity, synaptic plasticity, and behavior in Ts65Dn mice. We found Ts65Dn mice to be less sensitive to diazepam, as assessed by electroencephalography, long-term potentiation, and elevated plus-maze. Still, diazepam pre-treatment displayed typical effectiveness in reducing susceptibility and severity to picrotoxin-induced seizures in Ts65Dn mice. These findings fill an important gap in the understanding of GABAergic function in a key model of DS.

Before diazepam injections, mice were freely moving around their cages or standing along the cage sides. After injection, all diazepam doses produced relaxed wakefulness or drowsiness in both control and Ts65Dn mice. Diazepam-treated mice no longer showed a preference for the periphery of their cages and they frequently laid in a stretched-out posture (instead of curled-up) or sat quietly in any cage location. Mice eyelids were partially or fully open following diazepam administration. Visual inspection of the video-EEG recordings revealed that Ts65Dn and control mice showed no evidence of electrographic seizures or epilepsy-related phenotypes, neither spontaneously nor pharmacologically induced by any of the administered doses of diazepam.
To estimate the effects of diazepam on neuronal activity, we first calculated the power-spectrum distribution (PSD) over the 0.5-100-Hz frequency range in the video-EEG subset of animals. A normalized PSD was obtained for each experimental condition by the ratio of the PSDs for the EEG 30-min post-dosing segments and those for the EEG 30-min baseline-recording segments. Quantitatively similar results were observed for the EEG spectral analysis of cortical and hippocampal recordings. Thus, we will only report EEG data obtained from the cortical derivation in the main text (the EEG analysis dataset from the hippocampal derivation can be found in Supplementary Figs. S1 and S2 and Supplementary Tables S1-S5). Given that we did not observe major differences between control and Ts65Dn mice in EEG power across frequencies from 50 to 100 Hz, we will mostly focus on comparisons between EEG power across the frequency range of 0.5-50 Hz (results of spectral parameters within the high-gamma band can be found in Supplementary Tables S3 and S4).
We also evaluated similarities among groups by using hierarchical clustering analysis. A dendrogram representation of the Euclidian distances between group centroids (i.e., the group's mean), showed for example that the centroid of control mice group treated with 10 mg/kg diazepam was at the highest level of the hierarchical tree (Fig. 1g). The full set of observed inter-cluster differences in the grouped scatter plots suggest a genotypedependence for the effects of diazepam on EEG spectral power. This led us to compare the normalized PSDs obtained from baseline and post-dosing recordings for each dose in each genotype ( Fig. 1h-j, upper and middle portions) to determine the differences in diazepam effects between control and Ts65Dn mice for each experimental condition (Fig. 1h-j, lower portions). Diazepam induces significantly different changes on the cortical activity spectral profile of control and Ts65Dn mice. (a) Freely moving mice implanted with EEG electrodes were connected through a cable to the recording unit. The experiment consisted of a 30-min baseline recording, followed by a 90-min postdosing recording, in which mice were acutely treated with 1, 3, or 10 mg/kg diazepam (i.p.). (b-d) Grouped scatter plots of the first two canonical variables for all mice analyzed are shown in (b-d) for 1-, 3-, and 10-mg/ kg doses of diazepam, respectively. Colored points represent mice and are labelled by cluster group. (e) Can1 discriminated the control mice group from the Ts65Dn mice group at 1-, 3-, and 10-mg/kg doses. (f) Can2 differentiated control and Ts65Dn mice groups at 1-and 3-mg/kg doses. (g) The centroid of control mice group treated with 10 mg/kg diazepam was at the highest level of the dendrogram's hierarchical tree, whereas the centroids of Ts65Dn mice groups treated with 3-and 10-mg/kg doses had the shortest distance among all centroids. (h-j) PSD obtained from baseline and post-dosing recordings were compared for each dose in each genotype (black asterisks in upper and middle panels represent frequency bins in which significant differences in EEG spectral power were observed between baseline and treatment conditions; relative changes are expressed as the log 2 of the treatment/baseline ratio). Normalized PSD of control and Ts65Dn mice for each experimental condition was also compared (lower panels). Number of mice: Ct (N = 6) and Ts (N = 7) for all experimental conditions. In (e, f), data represent mean ± SEM and statistical significance is expressed as ** and *** for P < 0.01 and P < 0.001, respectively. In (h-j), solid lines represent means and dashed lines represent SEM and black asterisks indicate statistical significance (P < 0.05). We observed that the effects of diazepam on slower brain rhythms were mildly different between control and Ts65Dn mice ( Fig. 1h-j). Spectral power decreases in the 0.50-2.90 Hz and 4.50-5.70 Hz ranges were induced by 1 mg/kg diazepam in control mice (Fig. 1h, upper portion), whereas in Ts65Dn mice this decrease in power only occurred in the 0.50-2.70 Hz range (Fig. 1h, middle portion). Additionally, spectral power reductions were restricted to narrower frequency bands at the 3-mg/kg dose in both genotypes (0.50-1.10 Hz and 1.95-2.45 Hz in control and Ts65Dn mice, respectively; Fig. 1i, upper and middle portions). Finally, power reductions induced by 10-mg/kg doses occurred at the frequency ranges of 0.50-2.90 Hz (Fig. 1j, upper portion) in control mice and at 1.20-2.90 Hz and 8.80-9.50 Hz (Fig. 1j, middle portion) for Ts65Dn mice.
The presence of increased beta activity following PAM of GABA A Rs has been widely recognized as an EEG biomarker in both animal models 23,24 and humans 25,26 . As shown in Fig. 1h (upper portion), we detected a consistent power increase affecting the frequency range of 20.50-22.60 Hz in control mice treated with 1 mg/kg diazepam, which did not reach significance for the 3 and 10-mg/kg doses (Fig. 1h,j, upper portions). In contrast, we did not find evidence of power increase within beta-frequency range in Ts65Dn mice treated with any dose of diazepam ( Fig. 1h-j, middle portions). In addition, we detected significant differences in the spectral profile of beta activity between control and Ts65Dn mice at both 1 and 3 mg/kg diazepam (Fig. 1h,i, lower portion). Finally, significant differences in the spectral profile of gamma activity (34.15-37.35 Hz) were detected between control and Ts65Dn mice at 1-mg/kg dose (Fig. 1h, lower portion).
Given that 1-mg/kg was more effective than 3-and 10-mg/kg doses in eliciting a measurable physiological effect in cortical network oscillations, we will focus on the EEG effects of 1 mg/kg diazepam in the following analyses (results of the EEG parameters for 3 and 10 mg/kg diazepam can be found in Supplementary  Tables S3-S5). Furthermore, 1-mg/kg dose in mice can be considered equivalent to a ~ 5 mg dose in humans (using 60 kg as body weight reference 27 ), which is in the middle range of the common diazepam doses used in clinical practice (2-10 mg 28 ).
We observed that the effects of diazepam on EEG spectral power were dependent on frequency band. Diazepam at 1-mg/kg dose significantly (F (1, 11) = 9.472, P = 0.0105) increased beta-2 band power in control but not in Ts65Dn mice. A significant interaction was also found between genotype and treatment (F (1, 11) = 5.307, P = 0.0418). Post hoc tests confirmed that diazepam-treated control mice showed a significantly higher beta-2 band power compared with its baseline condition (P = 0.0037; Fig. 2a). In addition, we found significant treatment-and genotype-dependent effects (F (1, 11) = 12.268, P = 0.0049 and F (1, 11) = 6.323, P = 0.0288; respectively) on the beta-2 mean frequency power. A significant interaction between genotype and treatment was also detected (F (1, 11) = 6.537, P = 0.0267), with diazepam-treated control mice showing a significant increase in the power of the beta-2 mean frequency compared to its baseline condition (P = 0.0017) and to diazepam-treated Ts65Dn mice (P = 0.0017; Fig. 2b). Such findings support our previous observation that diazepam increases EEG activity within the beta-2 frequency range in a genotype-dependent manner.
We also found that diazepam at 1-mg/kg dose decreased beta-1 band power in control but not in Ts65Dn mice. Both genotype and treatment had significant effects (F (1, 11) = 5.350, P = 0.0411 and F (1, 11) = 5.496, P = 0.0389; respectively), with detected decreases in the beta-1 band power compared to its baseline condition (P = 0.0456) and to diazepam-treated Ts65Dn mice (P = 0.0237; Fig. 2c). These findings show that diazepam has opposite Data represent mean ± SEM. Statistical significance is expressed as * and ** for P < 0.05 and P < 0.01, respectively. Diazepam-induced changes in neural synchrony are significantly different between control and Ts65Dn mice. Phase synchronization plays an important role in supporting neural communication and plasticity either within the same or across brain regions 29 . To determine whether diazepam affects phase synchronization between neural oscillations, we calculated the instantaneous phase of cortical and hippocampal EEG recordings through Hilbert transform, and then we estimated the distribution of phase-difference between these pairs of signals across 0.5-50 Hz frequency range. We found that control and Ts65Dn mice showed similar phase-lag distributions during baseline recordings (P = 0.052, KS test with Bonferroni-adjusted P-value; Fig. 3a), with phase differences clustered around zero in both genotypes. In contrast, 1 mg/kg diazepam changed probability density functions of the phase difference between cortical and hippocampal EEG recordings in a genotype-dependent manner. Whereas the peak of phase-difference distribution became narrower as the number of phase-lag estimates close to zero increased in diazepam-treated control mice, phase lag distributions occurred with approximately equal probability in diazepam-treated Ts65Dn mice (P < 0.0001, KS test with Bonferroniadjusted P-value; Fig. 3b). We next estimated spectral coherence to measure the extent to which diazepam affects oscillatory coupling between cortical and hippocampal signals and to determine how coherent activity is distributed across frequency bins. Pairwise coherence in the beta-2 frequency range (19.40-23.20 Hz) was significantly increased by 1 mg/kg diazepam in control mice (Fig. 3c), whereas diazepam-treated Ts65Dn mice showed no significant changes in coherent activity compared to their pre-injection baseline (Fig. 3d). Significant differences in pairwise coherence at 20.60-23.40 Hz and 43.20-44.60 Hz frequency ranges were detected between diazepam-treated control and Ts65Dn mice (black asterisks; Fig. 3c). These findings show that cortical (and hippocampal) beta-2 frequency oscillations display a preferred phase-difference of coupling in diazepam-treated control mice, but not in Ts65Dn mice.
Ts65Dn and control mice show similar susceptibility to picrotoxin-induced seizures. Given that our previous findings did not produce any evidence of observable seizures or epileptiform activity induced by GABA A R PAM in Ts65Dn mice, we tested the effects of the inhibition of GABAergic transmission by the GABA A R antagonist picrotoxin on the seizure susceptibility of Ts65Dn mice. For this set of experiments, videoonly or video-EEG was recorded from control and Ts65Dn mice that were intraperitoneally (i.p.) injected with Histograms of phase differences between cortical and hippocampal EEG recordings, which were computed by dividing the phase-spectrum (0-360°) into equal-sized 100 bins (3.6°/bin). (a) Control and Ts65Dn mice showed similar phase-lag distributions during baseline recordings. (b) Diazepam at 1-mg/kg dose narrowed the peak of phase-difference distribution in control mice as the number of phase-lag estimates close to zero increased, whereas phase lag distribution of diazepam-treated Ts65Dn mice exhibited an almost uniform distribution. (c) Diazepam-treated control mice showed an increase in the pairwise coherence in the beta-2 frequency range (19.40-23.20 Hz; grey asterisks) compared to their pre-injection baseline. (d) Diazepam-treated Ts65Dn mice showed no significant changes in coherent activity compared to their pre-injection baseline. Then, seizure severity was scored using a modified Racine scale system 30 , in which higher values correspond to more severe seizures and a score of seven indicates death. Only fully induced convulsive (motor) seizures (classes 4-7) were considered to estimate the latency, which was defined as the time in minutes from the administration of picrotoxin to the onset of convulsive seizures. The administration of 2 mg/kg picrotoxin evoked convulsive seizures in approximately 64% and 47% of control and Ts65Dn mice, respectively (Fig. 4b). In contrast, all animals of both genotypes exhibited convulsive seizures when treated with 4 mg/kg picrotoxin (Fig. 4b). The administration of picrotoxin also produced a dosedependent effect on the mean values of seizure severity score as revealed by two-way ANOVA (F (1, 50) = 30.002, P < 0.0001; Supplementary Table S11). Post hoc tests identified a significant increase in the seizure severity score , Ts non-pretreated with 3 mg/kg diazepam (NPT; N = 15), and Ts pretreated with 3 mg/kg diazepam (PT; N = 15). Data represent mean ± SEM. Statistical significance is expressed as *, ** and *** for P < 0.05, P < 0.01, and P < 0.001, respectively. www.nature.com/scientificreports/ for both control (P = 0.0035) and Ts65Dn (P < 0.0001) mice treated with 4 mg/kg picrotoxin compared with their 2 mg/kg picrotoxin-treated peers ( Fig. 4c; Supplementary Table S12). Similarly, the mean values of latency were significantly affected by the picrotoxin dosage (F (1, 37) = 16.275, P = 0.0003; Supplementary Table S11), which was confirmed by post hoc comparisons for the latency to convulsive seizure onset in control (P = 0.0089) and Ts65Dn (P = 0.0056) mice treated with 4 mg/kg picrotoxin compared with their 2 mg/kg picrotoxin-treated peers ( Fig. 4d; Supplementary Table S12). We found that 71% of Ts65Dn mice treated with 4 mg/kg picrotoxin died during the course of the experiment, in contrast to 55% mortality in controls (Fig. 4e). Such difference, however, was not statistically significant (P = 0.2325, Mantel-Cox log-rank test).
Diazepam was efficacious in preventing picrotoxin-induced seizures in both Ts65Dn and control mice. To investigate a potential genotype dependence of diazepam's protective effect against picrotoxininduced seizures, we administered 2 mg/kg picrotoxin i.p. to control (N = 12) and Ts6Dn (N = 15) mice pretreated with 3 mg/kg of diazepam (30 min before the picrotoxin injections). Mouse behavior was monitored for 60 min following picrotoxin injection for seizure severity scoring (Fig. 4f). The anticonvulsant activity of diazepam was considered effective when it prevented the occurrence of convulsive seizures (classes 4-7). Treatment (F (1, 52) = 16.668, P = 0.0002), but not genotype (F (1, 52) = 0.393, P = 0.5333) had a significant effect on the mean values of seizure severity score (Supplementary Table S11). Post hoc analysis confirmed that the acute pretreatment with 3 mg/kg diazepam significantly decreased the severity of seizures induced by the administration of 2 mg/kg picrotoxin in both control (P = 0.0016) and Ts65Dn mice (P = 0.0189) compared to the animals that did not receive the pretreatment ( Fig. 4g; Supplementary Table S12). The incidence of picrotoxin-induced seizures with scores 4-7 in control and Ts65Dn mice was almost completely abolished by pretreatment with 3 mg/kg diazepam (only one pretreated mouse of each genotype developed convulsive seizures; Fig. 4h). These findings indicated that diazepam was similarly effective in reducing picrotoxin-induced seizure severity and susceptibility to convulsive seizures in both control and Ts65Dn mice.  Table S7). We then normalized data to untreated genotype-specific levels as an alternative way to investigate genotype-dependent effects of GABA or diazepam on TBS-induced LTP in Ts65Dn and control mice. Two-way ANOVA revealed that genotype (F (1, 88) = 10.987, P = 0.0013), but not treatment (F (3, 88) = 1.494, P = 0.2217), had significant effects on the mean levels of LTP. No significant interaction was found between genotype and treatment (F (3, 88) = 1.314, P = 0.2750; Supplementary Table S8). Post hoc analysis confirmed that control-derived hippocampal slices showed relatively greater inhibition to LTP by 200 μM GABA (P = 0.0211) and diazepam at 10 μM (P = 0.0145) compared with Ts65Dn-derived hippocampal slices ( Fig. 5e; Supplementary Table S9). These findings show that TBS-induced LTP is not sensitive to GABAergic modulation by either GABA or diazepam in Ts65Dn-derived hippocampal slices.

GABA and diazepam decrease hippocampal CA1 LTP induced by theta-burst stimulation (TBS) in control-but not in
Ts65Dn and control mice show differential sensitivity to diazepam modulation of anxiety-like behavioral responses. We used an EPM behavioral protocol to address how acute treatment with diazepam modulates anxiety-like responses in Ts65Dn mouse model. Control and Ts65Dn mice (8-12 weeks old) received either diazepam (control, N = 16; Ts65Dn, N = 14) at an anxiolytic but not sedative dose (1 mg/kg 33 ) or vehicle (saline; control, N = 16; Ts65Dn, N = 14) 30 min before testing in the EPM (Fig. 6a). Then, both standard parameters of spatiotemporal distribution patterns of exploratory activity across the maze (Fig. 6b) and ethological measures (which are used to assess a range of specific behaviors related to the murine defensive repertoire) were determined.
We found that treatment with diazepam (F (1, 56) = 6.123, P = 0.0164), but not genotype (F (1, 56) = 3.935, P = 0.0522), had significant effects on the mean values of open arm entries. No significant interaction was found between genotype and treatment (F (1, 56) = 1.065, P = 0.3066; Supplementary Table S13). Post hoc test revealed that diazepam-treated control mice entered open arms more frequently than vehicle-treated control mice (P = 0.0130; Fig. 6c; Supplementary Table S14). Similar results were observed for the analysis of percentage of time spent in open arms, which resulted in significant dependence on treatment (F (1, 56) = 10.124, P = 0.0024), but not on genotype (F (1, 56) = 2.337, P = 0.1319). Again, no significant interaction between genotype and treatment was found (F (1, 56) = 0.290, P = 0.5921; Supplementary Table S13). Post hoc tests showed that the percentage of time spent in open arms was significantly higher for diazepam-treated control mice than for vehicle-treated control mice (P = 0.0086; Fig. 6d; Supplementary Table S14). Finally, analysis of total arms entries showed a significant dependence on treatment (F (1, 56) = 5.947, P = 0.0179), but not on genotype (F (1, 56) = 1.412, P = 0.2397). No significant interaction was found between genotype and treatment (F (1, 56) = 0.553, P = 0.4600; Supplementary Table S13). Post hoc comparisons showed that total arms entries were almost two times as high for diazepam-treated control mice as for vehicle-treated control mice (P = 0.0235; Fig. 6e; Supplementary Table S14). Data on treatment effects on the mean values of closed arm entries can be found in Supplementary Fig. S5  We then scored ethological patterns of response as a means to assess behavioral strategies (i.e. risk assessment behaviors) expressed by mice when dealing with stressful situations. First, we assessed the frequency of stretched attend posture (SAP), which is characterized by a forward elongation of the animal's body followed by a retraction to its original position. SAP was differentiated as protected (i.e., occurring on/from the closed arms or central platform) or unprotected (occurring on/from the open arms). Two-way ANOVA showed that treatment (F (1, 56) = 6.955, P = 0.0108), but not genotype (F (1, 56) = 0.551, P = 0.4611), had significant effects on the percentage of protected SAP [(protected/total) × 100]. A significant interaction was found between genotype and treatment (F (1, 56) = 5.104, P = 0.0278; Supplementary Table S13). Post hoc tests revealed that vehicle-treated Ts65Dn mice had significantly less protected SAP than vehicle-treated control mice (P = 0.0383; Fig. 6e). Treatment with diazepam significantly decreased protected SAP in control mice when compared to vehicle-treated control mice (P = 0.0007; Fig. 6e; Supplementary Table S14). Regarding the percentage of unprotected SAP [(unprotected/total) × 100], analysis showed that treatment (F (1, 56) = 7.253, P = 0.0093), but not genotype (F (1, 56) = 0.071, P = 0.7903), had significant effects on this parameter. No significant interaction was found between genotype and treatment (F (1, 56) = 1.721, P = 0.1949; Supplementary Table S13). Post hoc tests showed that the percentage of unprotected SAP was significantly higher for diazepam-treated control mice than for vehicle-treated control mice (P = 0.0049; Fig. 6f; Supplementary Table S14).  Table S13). Post hoc tests confirmed that vehicle-treated control mice performed protected HD most frequently than vehicle-treated Ts65Dn mice (P = 0.0349) and diazepam-treated control mice (P = 0.0003; Fig. 6g). Vehicletreated Ts65Dn mice performed protected HD most frequently than diazepam-treated Ts65Dn mice (P = 0.0314; Fig. 6g; Supplementary Table S14). Regarding the percentage of unprotected HD [(unprotected/total) × 100], treatment (F (1, 56) = 17.866, P = 0.0001), but not genotype (F (1, 56) = 1.214, P = 0.2752), had significant effects on this parameter. No significant interaction was found between genotype and treatment (F (1, 56) = 1.045, P = 0.3111; Supplementary Table S13). Post hoc tests showed that the percentage of unprotected HD was significantly higher for diazepam-treated control (P = 0.0003) and Ts65Dn mice (P = 0.0324; Fig. 6h; Supplementary Table S14) than for vehicle-treated mice of the same genotype. These findings indicated that diazepam-treated control and Ts65Dn mice exhibited a similar pattern of exploratory behavior, but vehicle-treated Ts65Dn mice appeared "less fearful" or "distressed" than vehicle-treated control mice when exposed to EPM. (Analyses on frequency of closed arm returns and grooming time can be found in Supplementary Fig. S5 and Supplementary Tables S13 and S14). . Data represent mean ± SEM. Statistical significance is expressed as *, ** and *** for P < 0.05, P < 0.01, and P < 0.001, respectively.

Discussion
In this study, we investigated the effects of diazepam on brain activity, synaptic plasticity, and behavior in Ts65Dn and control euploid mice and found distinct, genotype-dependent drug responses. When compared to control animals, we found that diazepam neither increased beta-2 activity nor decreased the power of gamma and beta-1 oscillations in Ts65Dn mice. Although we observed slowing effects of diazepam on delta oscillations in both control and Ts65Dn mice, a similar effect on frequencies within the theta range was only found in control mice. Our findings also revealed that diazepam modulated the dynamics of phase relationships between brain areas in a genotype-dependent manner, with increased phase synchronization between oscillatory activities in control mice and opposite effects in Ts65Dn mice. In a separate set of experiments, we found that diazepam was effective in reducing seizure susceptibility and severity to picrotoxin-induced convulsive seizures in both control and Ts65Dn mice. Differences in diazepam effects were observed in the magnitude of CA1 TBS-induced LTP, which was depressed by diazepam in control-but not in Ts65Dn-derived slices. Finally, we observed genotypedependent differences in sensitivity to the modulation of anxiety-like responses by diazepam. Together, these findings highlight the complex interplay between the modulation of GABA A R-mediated signaling in electrophysiological and behavioral outcomes in Ts65Dn mice. The finding that control mice treated with 1 mg/kg diazepam showed a 30% increase in beta-2 activity compared to its EEG baseline recording is consistent with previous findings from several preclinical and clinical studies [23][24][25][26][34][35][36] . However, no evidence of diazepam-induced increase in EEG beta-band power was found in Ts65Dn mice. The specific reasons for this genotype-dependent difference are yet to be determined, but we know that the beta rhythm is closely tied to the coupling of excitatory and inhibitory cells in cortical tissue 37 . Given previous evidence on abnormalities in synaptic structure and function in Ts65Dn mice, we speculate that the beta-related electrophysiological phenotype displayed by diazepam-treated Ts65Dn mice may, at least in part, reflect abnormalities in reciprocally connected networks of inhibitory GABAergic interneurons in cortical areas of Ts65Dn mice.
The generation of gamma oscillations depends on neural synchronization either by mutual inhibition between interneurons or by the excitatory-inhibitory loop 38 . In both models, inhibition is mediated by GABA A R 38 . Previous studies showed that diazepam appears to modulate both the amplitude and frequency of gamma oscillations 39 , but not to disrupt their synchrony or rhythmicity 40 . Given that the modulation of GABA A R-mediated synaptic inhibition can control gamma-band rhythmogenesis 38,39 , we hypothesize that the lack of diazepam effect on the modulation of gamma/beta-1 activity may be associated with an impairment of such rhythmicity control mechanisms in Ts65Dn mice. This is consistent with recent findings reporting that spontaneous network oscillations (UP states) were significantly shorter and less frequent in the neocortex of Ts65Dn mice 41 . The dynamics of these spontaneous patterns of massive, persistent network activity reflect a complex, stereotypically organized sequential spread of activation through local cortical networks 42 in which fast-spiking GABAergic interneurons expressing parvalbumin (PV + cells) and deep pyramidal cells play a critical role [43][44][45] . The density of PV + cells across multiple cortical layers was found to be increased in Ts65Dn mice by some investigators 46,47 . However, Zorrilla de San Martin and collaborators 48 reported that perisomatic inhibition from PV + interneurons was unaltered in Ts65Dn mice. Instead, these investigators observed that these cells lost their classical fast-spiking phenotype and exhibited increased excitability. In either one of these scenarios, it is possible that an altered pattern of organization of excitatory and inhibitory cell assemblies in Ts65Dn vs. control euploid mice may account for the genotype-dependent differences in cortical network dynamics observed after diazepam treatment.
Disruption of excitation and inhibition balance (E/I balance), which is critically involved in the stability of network dynamics 38 , has been reported in Ts65Dn mice and inferred in persons with DS 49 . Recent findings showing a delayed GABA polarity switch in the hippocampus of neonatal Ts65Dn mice suggest developmental delays in maturation of neural circuits of these animals, which in turn may contribute to the disruption of E/I balance towards inhibition in adult Ts65Dn mice 50 . In addition, Zorrilla de San Martin et al. 48 reported that the dendritic inhibitory loop involving excitatory pyramidal neurons and Martinotti cells (which are somatostatinpositive interneurons that inhibit distal dendrites of excitatory pyramidal neurons) is enhanced in Ts65Dn mice 48 . Since α5-containing GABA A R mediate synaptic inhibition of pyramidal cell dendrites by Martinotti cells 51 , these findings may support previous evidence on the rescue of learning/memory deficits and synaptic plasticity deficits in Ts65Dn mice by negative allosteric modulators (NAMs) of these receptors [52][53][54][55] . Our findings also revealed that diazepam differently modulates phase relationships between brain areas in control and Ts65Dn mice. We found that diazepam-treated control mice showed a narrower distribution of phase differences when compared to diazepam-treated Ts65Dn mice, which suggests that diazepam increased phase synchronization between oscillatory activities of cortex and hippocampus in control mice, while it had an opposite effect in trisomic animals. We also observed a greater consistency of the phase difference (i.e., a higher coherence) between cortical and hippocampal oscillations in the beta-2 frequency range of diazepam-treated control mice compared to diazepam-treated Ts65Dn mice.
Brain oscillations at different frequencies can interact with each other via cross-frequency coupling 56,57 . Previous findings showed that, although diazepam changed the frequencies in which cross-frequency coupling occurs due to its slowing effect on frequencies, the strength of phase-amplitude coupling between slow and fast oscillations was unaffected by the drug 58 . Although we did not directly assess the properties of the phase-amplitude coupling between theta and gamma oscillations, we did see modulatory effects of 1 mg/kg diazepam on theta frequencies and gamma power in control mice, but not in Ts65Dn mice. We thus speculate that the GABA A R-mediated mechanisms underlying the coordination of oscillatory activity across cell assemblies are impaired in Ts65Dn mice brain, which is consistent with recent evidence suggesting abnormal prefrontal-hippocampal functional connectivity across different brain states in Ts65Dn mice 59  www.nature.com/scientificreports/ The administration of picrotoxin induced a dose-dependent increase in seizure severity and a decrease in latency to convulsive seizure onset in both control and Ts65Dn mice. Acute pretreatment with 3 mg/kg diazepam, however, significantly decreased seizure severity as well as the incidence of seizures with scores 4-7 similarly in mice of both genotypes, which demonstrates that diazepam was typically effective in reducing Ts65Dn mice susceptibility to picrotoxin-induced seizures.
The strength of GABAergic transmission has been shown to modulate plasticity at excitatory synapses on principal cells, which is a key mechanism underlying learning and memory encoding. We and others had previously shown that reduced levels of LTP in the hippocampus of Ts65Dn mice can be pharmacologically enhanced by the antagonism of GABA A -mediated synaptic inhibition with picrotoxin [11][12][13][14] . Here, we found that diazepam produced a significant depression in the mean level of TBS-induced LTP in control-but not in Ts65Dn-derived slices at 1 and 10 μM. Similar results were observed when control-and Ts65Dn-derived hippocampal slices were treated with 200 μM GABA. Together, these results extend previous findings on the role of GABA A -R in neuroplasticity dysfunction observed in Ts65Dn mice. In fact, they underscore the high-level of complexity of GABA A -mediated signaling in such animals since GABA A R appears sensitive to inhibition by picrotoxin, but less sensitive to diazepam-mediated modulation. Diazepam acts specifically through GABA A R containing α1, α2, α3 or α5, but not α4 or α6 subunit types, whereas picrotoxin exerts a noncompetitive inhibition via any combination of subunit subtypes. Therefore, the pattern of subunit arrangement of GABA A R, as well as their synaptic localization, might account for the pharmacological genotype-dependent differences in LTP levels displayed by control-and Ts65Dn-derived slices when treated with picrotoxin and diazepam. This is especially interesting in light of findings that NAMs of α5-containing GABA A R can enhance induction of LTP in Ts65Dn mice, as well as improve mouse behavioral performance without causing anxiogenic side effects or seizures [52][53][54][55] .
Given the crucial role of GABAergic function in modulating anxiety responses, an additional goal of this study was to investigate both anxiety-related behavioral differences between Ts65Dn and control mice in EPM exploration and the effects of diazepam treatment. To this end, both standard and ethological parameters were considered. In the first set of analyses, similar spatiotemporal patterns of exploratory activity were found between vehicle-treated groups. Both Ts65Dn and control mice preferred closed over open arms, which is consistent with the natural tendency of mice to avoid open areas in favor of enclosed spaces 60 . The ethological approach, however, allowed us to detect significant differences between defensive elements displayed by trisomic and euploid mice during EPM testing. For instance, Ts65Dn mice exhibited a relatively lower percentage of protected movements of SAP and HD. In addition, trisomic mice turned back into (rather than leaving) a closed arm less often than controls. Together, these results suggest that the differences observed in behavioral responses of Ts65Dn mice may stem from a lower level of anxiety when faced with approach-avoidance conflicts compared to the control mice. Alternatively, these behavioral responses of Ts65Dn mice may simply reflect a failure of this animal in recognizing, attending to, or interpreting environmental cues appropriately 61 . It is also possible that Ts65Dn mice have a decreased ability to recognize the aversiveness represented by the environmental context of the EPM, which may reflect deficits in executive function [62][63][64][65] .
Genotype-dependent differences in sensitivity to the anxiolytic properties of diazepam were also detected by EPM testing. Diazepam produced a significant increase in both time spent and number of entries into the open arms in control mice, while it had no major effects on spatiotemporal patterns of exploratory activity in Ts65Dn mice. We also observed that diazepam reduced risk assessment behaviors in both control and Ts65Dn mice. Considering the absence of a significant diazepam effect on percentage time and number of open arm entries in Ts65Dn mice, and the fact that risk assessment behaviors are very sensitive to anxiolytic drugs 66 , our findings suggest that diazepam may have induced a more subtle anxiolytic effect in Ts65Dn mice compared to control mice.
We cannot ignore the potential for genotype-dependent differences in pharmacokinetics and/or pharmacodynamics of diazepam effects. For instance, given diazepam's high lipophilicity 67 , genotype-dependent differences in body composition may also affect drug distribution and rate of blood-brain barrier penetration. It is also possible that regional differences in GABA A R subunit composition between Ts65Dn and control mice may contribute to the differential profile of responses displayed by animals of these two genotypes.
Many questions remain about the precise mechanisms behind the genotype-dependent differences in responses to diazepam documented in the present work, but it is clear that our findings do not support a global reversal of GABA A R responses from inhibitory to excitatory in the Ts65Dn mouse model of DS. Not only diazepam did not produce seizures in Ts65Dn mice, but, in agreement with clinical observations in patients with DS, it was similarly effective in reducing susceptibility and severity of picrotoxin-induced convulsive seizures in Ts65Dn and control euploid mice. Although it is still possible that Cl − homeostasis may be altered in limited brain regions of Ts65Dn mice, we should note that none the results of the present work, including in vitro work on hippocampal slices, support this notion.
The dysfunctional GABA A R-mediated neurotransmission phenotypes reported here should prove useful in enhancing our understanding of some of the neural substrates for the cognitive deficits observed in persons with DS. Furthermore, these findings also provide preclinical support for the use of EEG techniques for monitoring therapeutic responses in patients with DS. For example, the validation of specific EEG biomarkers of brain functioning may prove helpful in discriminating between potentially effective and dead-end pathways on the road toward improved clinical care of patients with DS with or without neurological and/or psychiatric co-morbidities. www.nature.com/scientificreports/ from The Jackson Laboratory, Bar Harbor, ME) with B6EiC3Sn.BLiAF1/J hybrid male mice (C57BL/6JEiJ females x C3Sn.BLiA-Pde6b + /Dn males, also purchased from The Jackson Laboratory). The colony was maintained in a controlled 12:12 light/dark cycle (lights off at 6:00 p.m.) in the Animal Resource Center at CWRU. Animals were provided with food and water ad libitum. Ts65Dn mice were genotyped using polymerase chain reaction (PCR) 68 . Euploid littermates of Ts65Dn mice were used as control mice.

Methods
Electrode preparation. Three bare silver wires (127 μm diameter; A-M Systems, Inc., Carlsborg, WA, USA) and two formvar-insulated nichrome wires (66 μm diameter; A-M Systems, Inc., Carlsborg, WA, USA) were soldered into separate channels of an eight-channel headstage (Plexon Inc., Dallas, TX, USA). Two silver wires were used as recording electrodes, while the other one was used as reference electrode. The bipolar electrode was made by twisting the nichrome wires around each other.
Surgical procedure for electrode implantation. Mouse survival surgeries were performed using aseptic techniques. Control (N = 6) and Ts65Dn (N = 7; 13-20 weeks old) male mice were anesthetized with ketamine and xylazine (95 mg/kg and 5 mg/kg, respectively, as a single intraperitoneal (i.p.) injection of 6.25 ml/kg), and then placed in the stereotaxic apparatus (Model 1900 Stereotaxic Alignment System, David Kopf Instruments, Tujunga, CA, USA). Body temperature was maintained at approximately 37 °C using a thermal pad. After fixing the mouse's head, the scalp was locally anesthetized with 2% lidocaine hydrochloride subcutaneous (s.c.) injection and the skull was exposed by a small midline incision. Hydrogen peroxide-soaked cotton swabs were used for removing any remaining connective tissue on the skull surface. Two small craniotomies (~ 0.5 mm diameter) were drilled on both sides of the frontal cortex (anterioposterior AP: + 1.5 mm, mediolateral ML: ± 1.5 mm) 69 for screw fixation. One more craniotomy was performed above the parietal cortex for positioning the bipolar electrode into the left hippocampal CA1 area (AP: − 2.3 mm, ML: − 1.7 mm; and dorsoventral DV: − 1.2 mm) 69 . The reference screw was placed above the contralateral parietal cortex. The stainless steel screws and the exposed skull were covered by Teets 'Cold Cure' dental cement (A-M Systems, Inc., Carlsborg, WA, USA). Carprofen (5 mg/kg, s.c. injection) was used for preemptive analgesia and within 24 h following surgery. Animals were singly housed after surgery and allowed to recover at least one week before being submitted to video-EEG monitoring.
Drug treatments. Diazepam was purchased from Hospira, Inc. (5 mg/ml; Lake Forest, IL, USA) and was diluted to 0.16, 0.48, and 1.6 mg/ml with 0.9% saline solution prior to administration. This diazepam formulation contains 40% propylene glycol, 10% alcohol, 5% sodium benzoate and benzoic acid added as buffers, and 1.5% benzyl alcohol added as a preservative. Diazepam was administered i.p. to mice in a volume of 6.25 ml/ kg body weight. The dose-response relationships for diazepam in Ts65Dn mice were evaluated following single administration of 1-, 3-, and 10-mg/kg doses. The rationale for choosing these doses was based on previous findings showing that 1 mg/kg diazepam was effective in reducing anxiety-like behaviors 33 , whereas 3 mg/kg diazepam was effective against seizures 70 . Finally, 10 mg/kg diazepam is a behaviorally disruptive dose, wherein treated-animals showed gait abnormalities, hypoactivity, and reduced muscle tone 71 . For these in vivo experiments, 0.9% saline solution was used as the placebo, which obviated the use of three separate vehicle-treated groups. There were three main reasons supporting this choice of vehicle in the placebo group. First, the v/v % of saline vs. the Hospira diazepam formulation was 97, 90, and 68 for 1-, 3-, and 10-mg/kg doses, respectively, which means that saline solution was the most representative constituent of the injections (mainly at the key doses of 1-and 3-mg/kg). Second, the corresponding dose of ethanol (perhaps the most potentially problematic vehicle component) in this formulation as used here were equivalent to 0.016, 0.048, and 0.158 g/kg for 1-, 3-, and 10-mg/kg diazepam doses, respectively. This means that, even at the maximum diazepam dose, the amount of ethanol in the Hospira formulation is well below any minimum ethanol dose in the literature necessary to produce detectable behavioral effects. Third, and most important, the Hospira formulation is a veterinarian-grade drug that is identical to the diazepam injections that have been used in human beings at emergency rooms across the globe for decades, and the effects of such clinical formulations have always been attributed to diazepam and not to any of its vehicle components. Still, one cannot totally discount the very small possibility that the lack of identical chemical contents in the vehicle-treated group could have a minor effect on the interpretation of diazepam effects vis-à-vis the placebo group in this subset of experiments. Picrotoxin was purchased from Sigma-Aldrich (St. Louis, MO, USA) and was diluted to 0.32 and 0.64 mg/ml with 0.9% saline solution prior to administration. Picrotoxin was administered i.p. to mice as individual single doses of 2 and 4 mg/kg in a volume of 6.25 ml/kg body weight. Individual mice were subjected to all experimental conditions with inter-injection intervals of at least 24 h.

Video and video-EEG recordings.
To determine the behavioral effects of diazepam on Ts65Dn (N = 15) and control (N = 14) mice (13-20 weeks old), video recordings were performed in freely moving mice acutely treated with 1, 3, or 10 mg/kg diazepam (i.p.). Subsets of Ts65Dn (N = 7) and control (N = 6) animals were previously chronically implanted with EEG electrodes to record brain electrical activity during these video recording sessions. A webcam (C920 HD Pro Webcam, Logitech, Newark, CA, USA) was placed above the recording chamber for continuous monitoring of mouse behavior. For the video-EEG subset of animals, on the day of the experiment, mouse head mount was connected to the recording apparatus (OmniPlex Neural Recording Data Acquisition System, Plexon Inc., Dallas, TX, USA) by an interface cable. Such cable was attached to a commutator, which allowed mice to move freely in the recording chamber. After 30 min of recording of baseline activity, mice were i.p. injected with one of the following: (1) 1 mg/kg diazepam; (2) 3 mg/kg diazepam; (3) 10 mg/kg diazepam; (4) 2 mg/kg picrotoxin; or (5) 4 mg/kg picrotoxin. Each injection was followed by data acquisition for EEG data analysis. Local field potentials were acquired at 2-kHz sampling rate and band-pass filtered (3-500 Hz, four-pole Butterworth filter). The adaptive power line noise removal was used to remove 60 Hz and its five first harmonics. Signal processing was carried out off-line by custom-written MATLAB codes (The MathWorks Inc., Natick, MA, USA). Multiple artifact-free EEG segments within the 30 min of baseline recording (prior to diazepam administration) and within the 30-min post-dosing period were manually selected for further analysis. We specifically focused on this period based on previous findings showing that diazepam and its active metabolites achieve maximum concentration in mouse brain within 30 min after dosing 33 . Power spectral density (PSD) was estimated by Welch's method (window-size: 5 s; window function: Hamming; overlap: 50%). The PSD was calculated for each animal of each experimental group and then a mean PSD was estimated per group. For the analysis of band power, mean frequency, power of the mean frequency, skewness, and kurtosis we partitioned the PSD into the wavebands delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), beta 1 (12-15 Hz), beta 2 (15-30 Hz), gamma , and high gamma (50-100 Hz) 72 . Because we found no significant differences between data from left and right cortical derivations, EEG recordings from both hemispheres of a given mouse in a given experimental condition were averaged to represent that animal in further analysis.
To determine phase-difference between hippocampal and cortical electrodes, raw time series data were bandpass filtered (0.5-50 Hz), and then the instantaneous phase was estimated from the complex-valued analytic signal via Hilbert transform in each time window. Phase-difference histograms were computed by dividing the phase-spectrum (0°-360°) into equal-sized 100 bins (3.6°/bin), and the number of observed events within each bin was determined in order to quantify its empirical probability density function.
To quantify the strength of the relationship between hippocampal and cortical signals within each frequency bin, spectral coherence was determined as a function of the PSD by dividing the square of the absolute value of the cross-spectral density function of these two signals by the product of the autospectral density function of each signal 73 . The coherence ranges from 0 to 1, wherein 1 denotes maximum coherence and 0 indicates no coherence between signals. The power of each waveband was calculated as the area under the power spectrum curve within a frequency range of interest and then it was expressed as a ratio of the total power of the signal.
Stock solutions of GABA 74 To reduce day-to-day variations, parallel recordings from untreated (aCSF only) and treated (aCSF containing GABA or diazepam) hippocampal slices of control and Ts65Dn mice were performed using a customized electrophysiology setup as described previously 12,13 . In all experiments, one hippocampal slice was used per genotype per condition. Field excitatory postsynaptic potentials were recorded with Ag/AgCl recording electrodes through thin-walled, 1.5 mm, WPI borosilicate glass micropipettes filled with aCSF (3-5 MΩ resistance), inserted into the CA1 region of the hippocampus. Two fine bipolar platinum/iridium electrodes (FHC, Bowdoinham, ME) were positioned on the CA1 Schaffer collateral fibers at opposite sides of the recording pipette. Stimulation intensity was adjusted to 40-50% of the amplitude that is required to produce population spikes. For all experiments, a stable baseline of synaptic transmission was established for 20 min prior to the induction of LTP. LTP was then induced by theta-burst stimulation (TBS, five bursts of four pulses at 100 Hz, 200 ms intertrain interval) in the CA1 region of Ts65Dn-and control-derived hippocampal slices and the fEPSPs were recorded for an additional 60 min. Signals from the recording electrode were amplified 1000 times (Brownlee Precision Electrophysiology Amplifier Model 440, San Jose, CA), low-pass filtered (8-pole Bessel) at 2 kHz, and digitized at 20 kHz by a Digidata digitizer (1322A, Axon Instruments) into a Microsoft Windows-based computer. PCLAMP software (PCLAMP 8.2, Axon Instruments) was used for data acquisition and offline analysis. Synaptic efficacy was determined by the slope of fEPSPs normalized to the mean value of fEPSP slopes recorded prior to the LTP induction. The open arms had a 5 mm high border to minimize animal falls. The height of the platform was 40 cm, and it was set at the center of an empty, circular tank to protect the mouse from falling or attempting to escape during the test. Home cages were transferred to the testing room at least 1 h prior to the beginning of the test. Each mouse was placed in the central area of the maze, facing one of the open arms, and tested for 5 min. The test was recorded using a digital camera mounted above the apparatus. The ToxTrac software 79 was used for analyzing spatiotemporal parameters, while ethological measures were scored by a trained observer blind to mouse genotype and treatment. However, regarding genotype, we know historically in our group that a trained observer can visually identify a Ts65Dn mouse from its euploid counterpart with approximately 70% accuracy. For drug treatment, one of the lab team members (typically the lab PI) prepared the study medication and labeled the vial with a randomization code. The experimenter then recorded and analyzed the video session without knowledge on whether she was analyzing behavior resulting from drug or saline injections.