Only male mice were used for all experiments described in this work. Fmr1 knockout (KO; DutchBelgianFXSConsortium, 1994), and wildtype (WT) littermate controls were generated by crossing heterozygote Fmr1 C57BL/6J females with WT C57BL/6J males. The females used for breeding originated from backcrossing on the C57BL/6J line (Charles River) for at least 10 generations. Experiments were carried following the European Communities Directive of 24th of November 1986 (86/609/EEC) and with approval of the local animal care and use committee of the Vrije Universiteit. For all experiments, prepubescent animals were between the ages of postnatal (P) day 14 to P21, and adolescent animals were between P42–P49, except the adolescent spontaneous IPSC group that ranges between P36–P42.

Following swift decapitation, brains were removed and placed on an ice-cold platform. Three frontal cortex slices were excised, and excess tissue surrounding the mPFC was cut away. The remaining tissue was first frozen in dry ice and subsequently stored at −80°C until further use. mPFC samples from three animals were pooled together for each age and genotype, resulting in eight samples per age per genotype. Synaptosomal fractions, enriched in pre- and postsynaptic proteins (Li et al., 2004), were isolated following sucrose gradient-assisted biochemical fractionation. In brief, samples were homogenized in ice-cold 0.32 M sucrose (5% of homogenate was collected as to total tissue lysate) and then centrifuged at 1,000× g for 10 min. The supernatant was loaded on top of a sucrose gradient consisting of 0.85 M and 1.2 M sucrose. After centrifugation at 100,000× g for 2 h, the synaptosomal fraction at the interface of 0.85 M/1.2 M sucrose was collected. Following the last centrifugation step (71,000× g for 30 min), the resulting pellet was dissolved in 70 μl 5 mM Hepes (pH 7.4) and stored at −80°C.

For each sample, protein concentration was determined with a Bradford assay (Bio-Rad Laboratories) and 5 μg protein was used for immunoblotting. Samples were lysed in Laemmli lysis buffer, separated by electrophoresis on gradient precast gels (4–20% Criterion TGX stain-free, Bio-Rad Laboratories), and blotted to PVDF membrane (Bio-Rad Laboratories). Samples were loaded in alternate order, so as each KO sample ran adjacent to a WT sample. The following primary antibodies were used: mouse anti-GABA_A α1 (1:1,000, Neuromab), rabbit anti-GABA_A α2 (1:500, Novus Biologicals), mouse anti-GABA_A β3 (1:500, Neuromab), rabbit anti-GABA_A γ2 (1:1, 000, Thermo Scientific -Pierce), mouse anti-GABA_B R1 (1:500, Neuromab), mouse anti-GABA_B R2 (1:1,000, Neuromab), rabbit anti-GAT1 (1:1,000, Chemicon/Millipore), goat anti-gephyrin (1:500, Santa Cruz). After incubation with horseradish peroxidase-conjugated secondary antibody (1:10,000 or 1:5,000; Dako, Glostrup) and visualization with Femto Chemiluminescent Substrate (Thermo Scientific) blots were scanned using the Li-Cor Odyssey Fc (Westburg) and analyzed with Image Studio (Li-Cor). Total protein was visualized using trichloro-ethanol staining, scanned using Gel Doc EZ imager (BioRad Laboratories) and analyzed with Image Lab (BioRad Laboratories) to correct for input differences per sample, as this is a reliable method not dependent on a single protein for normalization (Van den Oever et al., 2010). In brief, we corrected for input differences by taking the ratio between each band and the corresponding gel lane (Supplementary Figure S3). For quantification, intensity value was expressed as a ratio between that value and the WT group average, namely Sample1 = (Sample1)/(WT average), and so on. As such, the group average for WT ratios equaled to 1 (±SEM). Accordingly, fold change vs. WT was represented in the ratio of the KO group average. Differences between genotypes were assessed with either the Student’s t-test or with the Mann–Whitney U test. For both prepuberty and adolescence, one WT and one KO sample were excluded from statistical analysis as outliers (>2 × SD).

For spontaneous, miniature, evoked, and unitary recordings tissue was prepared as follows. Mice were swiftly decapitated and brains were extracted in ice-cold choline solution (110 mM choline chloride, 11.6 mM Na-ascorbate, 7 mM MgCl₂, 3.1 mM Na-pyruvate, 2.5 mM KCl, 1.3 mM NaH₂PO₄, 0.5 mM CaCl₂, 26 mM NaHCO₃, 10 mM glucose, at ~300 mOsm, pH 7.4) continuously gassed with carbogen mixture (95% O₂ and 5% CO₂). Subsequently, the brain was mounted and acute coronal slices 300–350 μm were obtained using a vibrating microtome (Microm) while submerged in ice-cold choline solution, continuously gassed with carbogen. Slices were left to recover for <5 min in room temperature choline solution before being transferred into a slice chamber containing continuously carbogen gassed ACSF (125 mM NaCl, 3 mM KCl, 1.2 mM NaH₂PO₄, 1 mM MgSO₄, 2 mM CaCl₂, 26 mM NaHCO₃, 10 mM glucose, ~ at 300 mOsm, pH 7.4) for at least 1 h before recordings.

For spontaneous, miniature, evoked, and unitary recordings, measurements were conducted in the presence of AMPA receptor blocker CNQX (10 μM, Abcam), NMDA receptor blocker DL-AP5 (50 μM, Abcam), and the GABA_B receptor blocker CGP55845 (4 μM, Tocris). For miniature IPSC recordings, tetrodotoxin (1 μM, Abcam) was additionally used to block Na⁺ gating channels. Slices were transferred to a submerged recording chamber and left to equilibrate for 10 min under continuous perfusion of ~2 ml/min of ACSF solution at 32°C. mPFC was selected under visual guidance from differential interference contrast microscopy and layer V pyramidal cells were identified based on their distance from the midline, morphology, and responses to current injections of 1,000 ms from −200 pA to +100 pA at 25 pA steps. Whole-cell recordings were conducted using borosilicate glass pipettes (2.5–5.5 MΩ) containing high Cl⁻ intracellular solution (70 mM K-Gluconate, 70 mM KCl, 10 mM Hepes, 4 mM Mg-ATP, 4 mM K2-phosphocreatine, 0.4 mM GTP, 0.2% biocytin, at 280–290 mOsm, 7.2–7.3 pH). Cells were held at −70 mV and recordings were terminated if series resistance changed by more than 20% during recordings; cells were rejected if access resistance was greater than 20–25 MΩ. Recordings were acquired with pClamp software (Molecular Devices), using a Multiclamp 700B amplifier (Molecular Devices), low-pass filtered at 3 kHz, and digitized with an Axon Digidata 1440A (Molecular Devices). Sampling frequency for spontaneous and miniature IPSCs was at 10 kHz, and for evoked and unitary STP at 50 kHz.

For sIPSC and mIPSC recordings, a total of 15 min were recorded per cell, and only the last 10 min were analyzed, to allow for equilibration and stabilization of the patch. Data reported are from the following number of animals (a) and cells (c): Pre-puberty [sIPSC WT(4a, 12c) KO(4a, 12c), mIPSC WT(6a, 9c) KO(7a, 10c)], Adolescence [sIPSC WT(7a, 14c) KO(9a, 11c)], mIPSC WT(5a, 13c) KO(5a, 14c). Traces were analyzed with mini Analysis software (Synaptosoft) to extract frequency, amplitude, rise time (10%–90% of peak amplitude), decay kinetics, and current waveforms. Biexponential fittings of spontaneous, miniature, end evoked IPSC decays were also conducted in mini Analysis, on a subset of events that were evaluated for the goodness of such fits, from an original set of ~400 randomly selected events, using the following equation: f(t) = Ifast*exp(–t/taufast) + Islow*exp(–t/tauslow). Weighted tau was calculated using the following equation: τweighted = (τfast*Ifast + τslow*Islow) / (Ifast + Islow). General amplitude histograms at 3 pA bin widths were generated from a random selection of 5,000 events per genotype/pharmacology/age from the pool of all events per condition, to avoid confounding effects due to differences in total numbers. Non-parametric probability density fits were performed using the statistics toolbox of Matlab (Mathworks).

Data reported are from the following number of animals (a) and cells (c): Pre-puberty [WT(7a, 10c) KO(7a, 11c)], Adolescence [WT(7a, 8c) KO(6a, 8c)]. Upon achieving a stable whole-cell patch with pyramidal cells, a unipolar stimulating electrode was positioned at a distance of ≤100 μm from the soma. Test pulses were delivered to assess the location, and recordings proceeded if stable and uniform (<2 ms rise time 10%–90%) responses could be observed. Injection current magnitude was set to yield ≥ of half-maximal synaptic responses, ranging from 20 μA to 50 μA. The short-term plasticity protocol was initiated, by delivering five pulses at frequencies of either 5 Hz, 20 Hz, 50 Hz, or 100 Hz, with a recovery pulse delivered 500 ms after the 5th pulse. The stimulation regime cycled between these frequencies, with an inter-sweep interval of 16 s, until a total of 15–30 sweeps were recorded for each frequency. Sweep duration was set to 2 s and the pulse duration was set at 300 μs. A small voltage step was included 80 ms from the start of each sweep to monitor the stability of the recording. Current injections were mediated through a Master-9 pulse stimulator and an ISO-flex stimulus isolator (A.M.P.I). For both ages and genotypes, 1–2 cells per group exhibited an averaged potentiation of responses and were excluded from the analysis. A low number of failures were observed under our eSTP protocol, and such sweeps were removed during analysis, leaving 20–25 sweeps per frequency. In a few cells, especially at 100 Hz, the number of failures was increased and a lower number of sweeps were used, typically 10–15.

Data reported are from the following number of animals (a) and cells (c): Pre-puberty [WT(4a, 4c(6c for 50 Hz) KO(7a, 11c)]. Upon achieving a stable whole-cell patch with pyramidal cells, fast-spiking interneurons in the vicinity were identified by the rounded morphology and patched. Action potential profiles of putative interneurons were generated and if they matched profiles of fast-spiking cells, current injections were initiated to probe for a unitary inhibitory connection between the interneuron and the pyramidal cell. If a unitary inhibitory to excitatory connection was present, the uSTP protocol was initiated, akin to the eSTP protocol. In uSTP however, each pulse was generated by supra-threshold current injections, for 2 ms at 1,800 pA per pulse, delivered via the patch pipette to the interneuron to elicit presynaptic action potentials. uSTP frequencies were prioritized, recorded as sets, and were not cycled. This was done to maximize complete data sets, and frequencies were collected in the order of 50 Hz, 20 Hz, 5 Hz, and 100 Hz. A maximum of 25 sweeps was averaged per frequency per cell, with a minimum of 10 sweeps for some cells at the highest frequency. Due to the high degree of failures for at least one of the five pulses, no sweeps were excluded based on that.

The quantal analysis relies on the hypothesis that the variation in post-synaptic responses occurs at multiples of the current elicited by elementary release (Bekkers, 1994). As such amplitude histograms can be described by multiple equidistant Gaussian distributions, reflecting discrete post-synaptic responses upon elementary release. For each cell events ranging from 5 pA to 150 pA amplitudes were binned at 2 pA increments, and best-fit functions were determined by the least-squares method along with visual confirmation of the fits (Edwards et al., 1990; Vislay et al., 2013).

In the absence of single-channel measurements, the use of PSnSN analysis on the decay current waveform can provide accurate estimates of single-channel currents, and the number of post-synaptic receptors open during peak current (De Koninck and Mody, 1994; Hartveit and Veruki, 2007). From each cell 50–500 events were selected, ensuring no overlapping events occurred during their decay phase. The mean response from all events per cell was calculated, scaled to the peak amplitude of each event, and subsequently subtracted from that event, generating a matrix of the remainders from each event per cell. The variance of this matrix was calculated across each index point. Next, the amplitude of the mean response of the cell was divided into 25 bins, and the time (index) ranges corresponding to each amplitude bin were used to calculate the mean-variance (σ²) from the remainders matrix, for each amplitude bin (I). The relationship between mean amplitude bin and corresponding variance was fitted with the following equation: σ²(I) = (i*I) - (I²/N) + σ_b², where (i) corresponds to the single-channel current, (N) corresponds to the number of receptors open during peak, and (σ_b²) is mean-variance of baseline. Unitary conductance was calculated by dividing the single-unit currents with the driving force corresponding to the intracellular and extracellular Cl⁻ concentrations in our preparations.

eSTP dynamics were further analyzed using the Tsodyks-Markram phenomenological synaptic transmission model as described before (Tsodyks and Markram, 1997; Loebel et al., 2009; Testa-Silva et al., 2012).

Analyses of Quantal distributions, PSnSN, Tsodyks-Markram modeling, eSTP, and uSTP, were performed with custom-built Matlab (Mathworks) scripts. Fittings of eSTP decay current waveforms were performed in GraphPad (Prism), using the same functions as with spontaneous and miniature IPSCs. Statistical tests were performed as described in each figure legend and the Supplementary Table S1. Normality distribution was assessed with either the D’Agostino-Pearson omnibus normality test or the Shapiro-Wilk normality test. Welch’s correction was applied in cases where parametric data exhibited the unequal distribution of variances.

The precise composition of GABAergic receptors and the degree of expression of GABAergic auxiliary proteins is both a cause and a therapeutic target for several neuropathological conditions including FXS (Fritschy et al., 1994; Lozano et al., 2014). Synaptosomal fractions were prepared from mPFC enriched protein lysates (see “Materials and Methods” section) and analyzed for the expression of several GABA subunits and auxiliary proteins (Figure 1, Supplementary Figure S3). No differences in protein expression for either GABA_A or GABA_B components tested were reported during prepubescence (Figures 1Ai,ii). However, during adolescence, an expression of GABA_A receptor α2 subunit was increased in the Fmr1-KO mPFC (Figure 1Bi, p = 0.04). The expression of the GABA_B receptor B1 subtype was found reduced compared to WT littermate controls (Figure 1Bii, p = 0.03). During either postnatal developmental times no difference in the expression of gephyrin (Figures 1Aiiii,Biii), or GABA transporter 1 (GAT1, Figures 1Aiv,Biv) was observed. An increase in α2 GABA_A subunit and a reduction in GABA_B receptor subtype B1 expression is suggestive of dysregulated inhibitory signaling in adolescent FXS mPFC.

To assess putative changes in functional inhibition onto layer V mPFC pyramidal neurons, inhibitory postsynaptic currents (IPSCs) were studied in the absence (spontaneous IPSC, Figure 2A) or presence (miniature IPSC, Figure 2B) of Na⁺ channel blocker tetrodotoxin, during prepuberty (Figure 2C) and adolescence (Figure 2D). Fmr1-KO sIPSC frequency was increased during prepubescence (Figure 2Ci, p = 0.04) with a shift toward shorter inter-event intervals (solid blue line, Figure 2Cii). During adolescence, however, Fmr1-KO sIPSC frequency was significantly reduced compared to WT littermate controls (Figure 2Di, p = 0.03). These changes were not observed in mIPSC recordings during either prepubescence (Figure 2Ci, p = 0.53) or adolescence (Figure 2Di, p = 0.12). Together with our data reveal dynamic and activity-dependent changes of prepubescent and adolescent GABA_A mediated inhibitory drive in Fmr1-KO mPFC.

Receptor kinetics orchestrate precise transitions between open and close channel states, and changes in stoichiometry, neurotransmitter levels, or general synaptic morphology, can impact gating currents. Prepubescent GABA_A synaptic rise time and weighted tau (tauW) of decay were equal between genotypes (Figure 3A, Supplementary Figure S2, Rise Time p = 0.12, tauW p = 0.12). By adolescence, Fmr1-KO synaptic kinetics slowed down substantially, exhibiting prolonged synaptic activation times (Figure 3Bi, sIPSC p = 0.04, mIPSC p < 0.01) and tauW of decay (Figure 3Bii, sIPSC p < 0.01, mIPSC p = 0.01), in an action potential independent manner (Supplementary Table S1). This slowdown coincides with the increase in α2 subunit expression, shown to prolong receptor decays (Lavoie et al., 1997), promoting extended charge transfer times.

During prepubescence, synaptic amplitudes in Fmr1-KO were larger for both spontaneous and miniature IPSCs (Figure 4Ai, sIPSC p < 0.01, mIPSC p < 0.01). We did not observe any differences in input resistance (WT vs. KO: sIPSC 122.49 ± 36.91 MΩ vs. 140.08 ± 48.62 MΩ, p = 0.35; mIPSC 170.05 ± 48.73 MΩ vs. 179.45 ± 60.83 MΩ, p = 0.74, data not shown), or capacitance (WT vs. KO: sIPSC 208.75 ± 59.49 pF vs. 185.54 ± 30.25 pF, p = 0.29; mIPSC 161.73 ± 23.13 pF vs. 149.95 ± 25.96 pF, p = 0.39, data not shown) between these cells. Prepubescent amplitude distributions from Fmr1-KO cells exhibited a reduction in smaller currents with a parallel shift towards larger amplitude bins in both recording conditions (Figures 4Aii,iii). During adolescence, sIPSC and mIPSC amplitudes were comparable between the two genotypes (Figure 4Bi, sIPSC p = 0.18, mIPSC p = 0.36), and no differences in amplitude distributions were observed (Figures 4Bii,iii). The nature of synaptic current amplitude distributions can reveal the number of discrete post-synaptic responses underlying the variations in amplitudes, deduced by the number of Gaussians best fitting amplitude distributions (Edwards et al., 1990). To that end, we performed quantal analysis on mIPSC amplitudes (Figure 5A), during prepuberty (Figures 5Ai,ii) and adolescence (Figures 5Aiv,v). In line with the increased IPSC amplitudes during prepubescence, a greater number of distributions better fitted Fmr1-KO amplitudes as compared to WT littermates (Figure 5Aiii, p < 0.05). During adolescence, no difference in the number of Gaussians that best fitted the data was observed (Figure 5Avi, p = 0.67). Our analysis of IPSCs received by layer V mPFC pyramidal cells demonstrates an early potentiation in amplitudes along with an expansion in discrete post-synaptic responses that normalizes during adolescence.

To better define the nature of the post-synaptic changes observed, we applied PSnSN analysis on synaptic current waveforms (Figure 5B), during prepuberty (Figure 5Bi) and adolescence (Figure 5Bii). Such analysis can reveal the single-channel current and the average number of receptors open at peak (Hartveit and Veruki, 2007). During prepubescence, PSnSN analysis revealed a significantly greater number of receptors open during peak activation in Fmr1-KO cells (Figure 5Biii, p < 0.05), without a change in unitary conductance (Figure 5Biv, p = 0.94). During adolescence, the number of receptors open at peak current (Figure 5Bv, p = 0.55) and unitary conductance (Figure 5Bvi, p = 0.16) was similar between genotypes. Therefore, the observed action potential independent shift towards larger prepubescent amplitudes can partly be due to increased receptor numbers open during peak activation. Whether this reflects a general increase in total receptor numbers per synapse or an increase in the number of receptors engaged, can not be deduced from this analysis. Nevertheless, heightened prepubescent GABA receptor activation leads to stronger post-synaptic inhibition onto layer V mPFC pyramidal cells.

Short-term plasticity (STP) reflects use-dependent dynamic changes in synaptic transmission during repeated presynaptic activity, and it is believed to participate in neuronal information processing (Zucker and Regehr, 2002). To study STP a stimulating electrode was used to deliver five pulses at frequencies of 5 Hz, 20 Hz, 50 Hz, and 100 Hz, and a recovery pulse was evoked at 500 ms after the 5th response (Figures 6i–iv). This protocol resulted in progressively depressing plasticity. No differences in the magnitude of the first response were observed between the two genotypes, during either prepubescence or adolescence over all frequencies (Supplementary Figures 1Ai–Bi, Supplementary Table S1). During prepubescence stimulation at intermediate frequencies resulted in reduced inhibitory synaptic depression in Fmr1-KO, both at 20 Hz [Figure 6Aii, Supplementary Table S1, two-way-RM-ANOVA; Genotype (F_(1,19) = 4.49, p = 0.04), Evoked Response (F_(3,57) = 50.14, p < 0.0001), Interaction (F_(3,57) = 0.82, p = 0.49), Subjects (F_(19,57) = 9.7, p < 0.0001)] and at 50 Hz [Figure 6Aiii, Supplementary Table S1, two-way-RM-ANOVA; Genotype (F_(1,19) = 4.75, p = 0.04), Evoked Response (F_(3,57) = 29.37, p < 0.0001), Interaction (F_(3,57) = 0.72, p = 0.55), Subjects (F_(19,57) = 12.38, p = 0.0001)]. By adolescence, the degree of inhibitory depression was comparable between genotypes (Figures 6Bii,iii, Supplementary Table S1). Therefore, prepubescent mPFC inhibitory STP was reduced, however, this reduction returned to WT levels by adolescence.

Furthermore, we extended the STP protocol, now between connected pairs of layer V mPFC pyramidal and fast-spiking interneurons (Figure 7). Unitary connections were subjected to a stimulation regime similar to the evoked IPSC protocol (Figures 6i–iv, see “Materials and Methods” section). No differences in unitary STP depression were observed between the two genotypes at 5 Hz, 50 Hz, and 100 Hz (Figures 7Bi,iii,iv, Supplementary Table S1). However, a significant attenuation in the depression of unitary STP responses was observed in Fmr1-KO connected pairs at 20 Hz [Figure 7Bii, Supplementary Table S1, two-way-RM-ANOVA; Genotype (F_(1,13) = 5.00, p = 0.04), Unitary Response (F_(3,39) = 9.71, p < 0.0001), Interaction (F_(3,39) = 0.58, p = 0.63), Subjects (F_(13,39) = 4.77, p = 0.0001)], akin to the one observed during the evoked STP protocol (Figure 6Aii). Our STP experiments reveal an activity-induced reduction in inhibitory synaptic depression, partially recapitulated in fast-spiking to pyramidal synaptic communication. Collectively, our data highlight dynamic GABAergic alternations at the molecular and functional levels, for the first time in the mFPC of FXS mice, during two crucial timepoints for postnatal prefrontal development.

Collectively, we observed enhanced inhibition during prepubescent prefrontal development in Fmr1-KOs. Increased inhibitory frequency and amplitudes, together with reduced inhibitory depression, promote stronger inhibition, possibly to counteract ongoing excitatory hyperconnectivity (Testa-Silva et al., 2012). Although the enhanced inhibition could restore network balance, it highlights significant inhibitory imbalances occurring in immature prefrontal circuits, that could fail to correctly establish and mature, thus negatively affecting prefrontal mediated cognition.

In contrast to enhanced prepubescent Fmr1-KO inhibition, during adolescence prefrontal inhibitory drive was reduced, and receptor kinetics slowed down. Although this could be a rebound response due to stronger prepubescent inhibition, it highlights further derailment of prefrontal inhibitory circuitry in FXS. As prefrontal cortices take longer to mature, a continued inhibitory network imbalance during ongoing circuit development and maturation can severely hinder prefrontal mediated cognition. Importantly, during adolescence the subunit expression changes we describe, can aid in devising more meaningful therapeutic strategies.