Disrupted inhibitory plasticity and homeostasis in Fragile X syndrome.

Fragile X Syndrome (FXS) is a neurodevelopmental disorder instigated by the absence of a key translation regulating protein, Fragile X Mental Retardation Protein (FMRP). The loss of FMRP in the CNS leads to abnormal synaptic development, disruption of critical periods of plasticity, and an overall deficiency in proper sensory circuit coding leading to hyperexcitable sensory networks. However, little is known about how this hyperexcitable environment affects inhibitory synaptic plasticity. Here, we show that in vivo layer 2/3 of the primary somatosensory cortex of the Fmr1 KO mouse exhibits basal hyperexcitability and an increase in neuronal firing rate suppression during whisker activation. This aligns with our in vitro data that indicate an increase in GABAergic spontaneous activity, a faulty mGluR-mediated inhibitory input and impaired inhibitory plasticity processes. Specifically, we find that mGluR activation sensitivity is overall diminished in the Fmr1 KO mouse leading to both a decreased spontaneous inhibitory postsynaptic input to principal cells and a disrupted form of inhibitory long-term depression (I-LTD). These data suggest an adaptive mechanism that acts to homeostatically counterbalance the cortical hyperexcitability observed in FXS.

Introduction

FMRP is implicated in the transport of approximately 4–8% of all synaptic mRNAs and regulates the translation of numerous proteins involved in synaptic transmission and receptor systems (Brown et al., 2001). Defects underlying FXS and autism spectrum disorders are widely believed to lie at the level of the synapse (Zoghbi, 2003; Ebert and Greenberg, 2013) affecting both excitatory and inhibitory neurotransmission across multiple brain regions leading to hyperexcitable sensory circuits (Huber et al., 2002; Olmos-Serrano et al., 2010; Paluszkiewicz et al., 2011). Clinically, children with FXS reflect heightened response to somatosensory stimuli, and this manifests itself behaviorally as sensory “defensiveness”, characterized as retreating or pulling away when touched (Miller et al., 1999; Hagerman and Stafstrom, 2009). This hypersensitive phenotype has also been described in the mouse model of FXS (Fmr1 KO mice) whom have deficits in whisker-tactile learning tasks due to a hyperactive response to sensory activity (Arnett et al., 2014; He et al., 2017). Overall, this altered cortical response may support the hypothesis that a highly active network fails to adequately decipher sensory inputs due to inadequate excitatory/inhibitory (E/I) balance and deficiencies in proper plasticity mechanics within the neocortex.

Synaptically, the best documented consequence of FMRP loss is the lack of translation repression of mGluRs (Bagni and Greenough, 2005). The “mGluR theory” of FXS [reviewed in (Bear et al., 2004)] accounts for the diverse neurological phenotypes associated with uncontrolled mGluR-mediated protein-synthesis-dependent functions. Yet, while mGluR-mediated excitatory synaptic function and plasticity has been previously studied in FXS, little is known about how the mGluR theory plays into inhibitory synaptic mechanisms (El Idrissi et al., 2005; D’Hulst et al., 2006; Gibson et al., 2008; Olmos-Serrano et al., 2010; Martin et al., 2014). Therefore, considering that sensory input computation and interpretation are heavily dependent on inhibitory network function [reviewed in (Maffei, 2017)], it is important to understand how the aberrant glutamatergic and mGluR modulation seen in FXS alters inhibitory circuits of the primary somatosensory cortex (S1).

In this study, we show that in vivo Fmr1 KO mice have alterations in sensory processing in layer 2/3 (L2/3) of S1 and that neurons within this region are hyperexcitable and are differentially modulated by mGluR activation. This is further emphasized in in vitro recordings which demonstrate that mGluR activation has a diminished impact on inhibitory synaptic activity and fails to drive I-LTD in Fmr1 KO mice. This altered response seems to primarily be due to a decrease in the mGluR sensitivity rather than mGluR-mediated activation via endocannabinoid signaling, that could also be accounted for by a ceiling effect of the cortical network activity. We find that the disturbance in mGluR synaptic and neuromodulatory function, has a significant impact on inhibitory plasticity. Overall, we believe that this hyperexcitable phenotype observed in the Fmr1 KO mouse induces changes in the inhibitory drive of the network in a mGluR-dependent manner, possibly as a homeostatic counterbalance mechanism that can directly affect the computational processes occurring in L2/3 of S1 thereby disrupting sensory perception capabilities.

Materials and methods

Mice

All experiments were performed under protocols approved by the Ethics and Animal Care Committee of the University of Colorado|Anschutz Medical Campus (Protocol#00039). Both WT and Fmr1 KO mice were acquired from Jackson Laboratories (Bar Harbor, ME, USA) and bred onsite. The Fmr1 KO mice obtained from these crosses were tested via genotyping protocols. Male mice utilized in this study possess the same congenic FVB background and only X chromo-some Fmr1 gene hemizygotes were utilized for the experiments in this research. Neonatal mice were housed with their respective mothers before weaning.

In vivo recordings

Extracellular recordings in anesthetized and awake behaving mice were performed as previously described (Doucette et al., 2011; Gire et al., 2013; Li et al., 2014; Li et al., 2015a). Briefly, four tetrodes consisting of four polyimide-coated nichrome wires (diameter 12.5 μm, Sandvik) were connected to a 16-channel interface board (EIB-16, Neuralynx) and fed through a housing glued to the board. Immediately before implantation the tetrodes were gold-plated to an impedance of 200–350 MΩ. Adult mice (P60–90) were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) and were then implanted with the tetrodes in layer 2/3 at co-ordinates AP:−1.46 mm, ML: 3 mm. On the day of surgery, the electrodes were implanted 200 μm above the final location and every day it was lowered 50 μm until reaching a final depth of DV: 1 mm, respectively. A screw was also implanted in the skull in the opposite hemi-sphere (1 mm right and 2 mm posterior of bregma) to serve as a ground reference. The mice were allowed to recover at least one week before experiments were performed. On the day of the experiment, mice were placed in an 18×12×12 cm anesthesia plexiglass chamber, customized to deliver air puffs through a port while a 1% isoflurane in 2:1 oxygen/nitrogen inhalant was delivered using a SurgiVet model 100 vaporizer. Animal reflex responses were checked throughout to test for anesthetization. Before recording, mice were placed next to the air puff port and positioned to get the maximal contralateral neuronal response. Whiskers were then stimulated with a ~ 3 L/min air puff. The output of the tetrodes was connected to a 16-channel amplifier (A-M Systems 3500) through a 1× gain headstage (Tucker-Davis Technologies). The signal was amplified 1000× and was recorded digitally at 24 kHz with a Data Translation DT3010 A/D card in a PC computer controlled with a custom MATLAB (Mathworks) program. Spike clustering analysis was performed as is explained in detail in (Li et al., 2015b). Briefly, data was filtered digitally between 300 and 3000 Hz. Then, using custom written MATLAB programs, each of the 16 channels were put at a threshold that was three times the standard deviation of the mean. Every spike with an amplitude greater than the threshold was imported into a second program (1 ms record per spike) which performed superparamagnetic clustering and wavelet decomposition of the spikes using 13 different wavelets and three principal components (Quiroga et al., 2004).

To determine the responsiveness of the units to the different events, we aligned all trials to the starting point of the event and calculated the average firing rate (FR, in Hz). FR was calculated for 500 ms before the air puff (baseline) and for 1 s after the air puff (air) was delivered. A paired samples t-test was performed to compare before and after the event and the p-value was corrected for multiple comparisons using the false discovery rate (Curran-Everett, 2000). To display the results, the FR was calculated in 0.1 s bins and normalized per unit to the mean FR 1 s before the beginning of the event.

For the (S)-3, 5-Dihydroxyphenylglycine (DHPG) infusion experiments, a cannula (Plastics One Inc.) was attached to the implant housing and the same surgical protocol was performed. Once the mice recovered from surgery, twenty 9 s trials were recorded in the awake freely moving mouse. Then, 3 μL of DHPG (100 μM) were infused under anesthesia (1% isoflurane) at a rate of 1 μL/min. 30 min after the infusion (when the mouse was fully awake), twenty 9 s trials were recorded. The FR of the twenty trials before and after DHPG infusion was averaged. Student’s t-test, KeS tests, and Chi-Square test were applied to test for statistical significance.

Slice preparation

Postnatal 19 to 25-day old WT and Fmr1 KO mice were anesthetized by CO2 inhalation and decapitated. This age range was used, because this is the developmental time point when inhibitory neurotransmission and inhibitory interneurons have reached maturity (Huntsman and Huguenard, 2000; Doischer et al., 2008; Goldberg et al., 2011). Brains were removed and initially placed in a 4 °C oxygenated sucrose slicing solution for 2 min and was composed of (in mM): 234 sucrose, 11 glucose, 26 NaHCO3, 2.5 KCl, 1.25 NaH2PO4 10, MgSO4, and 0.5 CaCl2 (equilibrated with 95% O2 and 5% CO2, pH 7.4). Coronal slices were obtained at a 300-μm-thickness using a Vibratome (Leica VT1200S, Leica Biosystems, Buffalo Grove, IL, USA). Then slices were hemisected and incubated in pre-warmed (36 °C), oxygenated artificial cerebrospinal fluid (ACSF; in mM): 126 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1.25 NaH2PO4, 2 MgCl2, and 2 CaCl2) for 45–60 min before being transferred to the recording chamber, where they were continuously perfused with oxygenated ACSF at 32–36 °C.

Electrophysiology

Recordings were obtained from pyramidal cells in primary somatosensory cortex L2/3 in WT and Fmr1 KO mice. Slices were visually identified using differential interference contrast (DIC) on a modified Olympus upright microscope (Scientifica, East Sussex, United Kingdom). Whole-cell recordings were performed with a Multiclamp 700B amplifier (Molecular Devices Corp., Sunnyvale, CA, USA), using recording pipettes (3–5 MΩ) pulled on a PC10 vertical puller (Narishige International, Amityville, NY, USA). Individual spontaneous inhibitory postsynaptic currents (sIPSCs) were isolated by bath applying NMDA and AMPA receptor antagonists (50 μM D-APV and 10 μM DNQX) while holding L2/3 pyramidal cells at −60 mV in voltage-clamp mode. Spontaneous excitatory postsynaptic currents (sEPSCs) were isolated by blocking GABAAR-mediated currents (5 μM gabazine) while holding L2/3 pyramidal cells at −60 mV in voltage-clamp mode. Recording pipettes were filled with an intracellular solution containing (in mM): 90 CsCH3SO4, 1 MgCl2, 50 CsCl, 2 MgATP, 0.2 Cs4-BAPTA, 10 HEPES, 0.3 Tris GTP, and 5 QX314 for sIPSCs, I-LTD, and depolarization-induced suppression of the inhibition (DSI) recordings. For sEPSC recordings, pipettes were filled with an intracellular solution containing (in mM): 135 potassium gluconate, 20 KCl, 10 Hepes, 0.1 EGTA, 2 MgATP, and 0.3 Na2GTP. For all recordings, series resistance (Rs) was monitored throughout each voltage-clamp recording with 50 ms, −10 mV steps every 4 s for drug application and DSI experiments, and every 15 s for the I-LTD experiments. Recordings where Rs changed more than 20% were discarded. Recordings were low-pass filtered at 4 kHz (Bessel filter) and digitized at 10 kHz (Digidata 1440) using pClamp 10.3 software (Molecular Devices Corp., Sunnyvale, CA, USA).

Spontaneous postsynaptic experiments

For sIPSC/sEPSC recordings, we recorded 5 min of basal activity followed by 5 min of drug bath application (DHPG or carbachol (10 μM)). Five different DHPG concentrations were tested (1, 2, 5, 10, 100 μM) in this study, and only one concentration per pyramidal cell was recorded per slice. For analysis, postsynaptic events were visually identified using pre-written custom code routines in Axograph-X. For each recording, three 30 s time frames (at 30 s, 2.5 min, and 4.5 min) were analyzed from the 5 min. For drug recording conditions, time frames were taken after 2 min of bath application (which is the turnout point for the whole volume bath). These events were analyzed by comparing amplitude and frequency within and between WT and Fmr1 KO mice, and paired and unpaired t-tests were applied respectively for statistical comparison. Differences in the magnitude of the DHPG responses between WT and Fmr1 KO mice were further analyzed by one-way anova test. A wash-out of the respective drug for 10 min was performed before the next slice was added to the recording chamber.

I-LTD experiments

For electrically induced I-LTD experiments, evoked IPSCs (eIPSCs) were elicited by 1-ms-long extracellular stimuli using a concentric microelectrode (FHC) placed in layer 4 of the somatosensory cortex every 15 s. Stimulation amplitude was calculated by the 50% of the maximum amplitude response of the IeO curve (for consistency we maintained response amplitude and not stimulation amplitude for our recordings). After 5 min of stable baseline, electrical I-LTD was induced by high-frequency stimulation (HFS), which consisted of 2 trains (20s apart), each containing 100 pulses at 100 Hz. For chemically induced I-LTD, 5 min of baseline was recorded followed by 10 min of 10 μM of DHPG or 10 μM of muscarine and then washed-out for the rest of the experiment time. For drug bath application experiments, AM251 (4 μM), a combination of MPEP (4 μM) (2-Methy-6-(phenylethynyl)pyridinehy-drochloride) and LY367385 (100 μM) ((S)-(+)-α-Amino-4-carboxy-2-methylbenzeneaceticacid), and H89 (10 μM) (N-[2-[[3-(4-Bromophenyl-2-propenyl]amino]ethyl]-5-iso-qunolinesulfonamidedihydrochloride) were pre-incubated for 10 min before the I-LTD experiment and bath applied throughout the whole experiment. For all I-LTD experiments, recordings were 40 min long and performed in the continuous presence of NMDA and AMPA receptor antagonists (50 μM D-APV and 10 μM DNQX). The magnitude of I-LTD was estimated by comparing the average amplitude of responses for each minute from 30 to 40 min of recording to baseline-averaged responses one minute before induction. We also compared averaged amplitude responses from 4 to 7 min of recordings after electrical or chemically-induced protocols to baseline-averaged responses one minute before induction to estimate short term depression of the inhibition (I-STD). For statistical analyses purposes, we performed a paired t-test to determine the expression of I-LTD and we reported p-values for 5 min and 35 min after induction. Additionally, a one-way anova test was used for statistical comparisons to determine differences in the I-LTD response between genotypes.

DSI experiments

For DSI experiments, sIPSCs were recorded in the presence of carbachol (20 μM) for 4 min before DSI was evoked. DSI was evoked by a 1 s voltage step from −60 to 0 mV. DSI magnitude was measured as the percentage of change between the mean of the ten consecutive IPSCs preceding the depolarization and the mean of three IPSCs immediately following depolarization (acquired 3–12 s after the pulse). In a subset of experiments AM251 (4 μM) was added with carbachol for related experiments. To determine whether DSI was evoked a paired t-test was performed for statistical comparisons.

For Win55-212-2 (Win55) experiments, evoked IPSCs (eIPSCs) were elicited by 1-ms-long extracellular stimuli using a concentric microelectrode (FHC) placed in layer 4 of the somatosensory cortex every 15 s, as described above as electrically induced I-LTD. After two minutes of baseline recording, Win55 (5 μM) was bath-applied for 5 min and then washed-out for 15 min while recording. Statistical comparisons were made between minute 1 (baseline) and minute 10 (Win55 effective steady state) and between genotypes using unpaired and paired t-tests, respectively.

Statistical analysis

Data were acquired with pClamp10 and Analyzed with Axograph-X and MatLab. All data was tested for normality using the Shapiro-Wilk test. Accepted scale of p-values for the entire text were used (* = P ≤ .05, ** = P ≤ .01, *** = P ≤ .001, **** = P ≤ .0001). All error bars represent standard deviation unless otherwise noted.

Results

Cortical response to sensory stimulation recorded in vivo

Cortical hyperexcitability is prevalent in FXS and in the Fmr1 KO mouse model [reviewed in (Contractor et al., 2015)]. Therefore, we initially wanted to quantify this hyperexcitability by in vivo tetrode recordings of cellular responses at baseline and to whisker stimulation in L2/3 in anesthetized and awake behaving WT and Fmr1 KO mice (P60–90). The setup for tetrode recording neuronal activity in L2/3 in the anesthetized mouse is shown (Fig. 1A). Offline spike analyses of tetrode recordings revealed that the baseline rate of the recorded units (before stimulation of the whisker) was higher in Fmr1 KO mice, (Fig. 1B, t-test p < .02, KeS test p < .005; n = 29 units from 3 WT and n = 24 units from 3 Fmr1 KO) which is consistent with network hyperexcitability in these mice. Next, we stimulated the contralateral whiskers with an air puff in anesthetized mice and calculated the delta firing rate (FR) before and after the stimulus. We found that a subset of cells responded differentially (determined with a t-test corrected for multiple comparison of 1 s before the puff and 500 ms after, p < .05) with either increases or decreases in firing rate (FR) in both groups (Fig. 1C and D). Importantly, the units in the Fmr1 KO mice responded differently than those in the WT mice, with more units exhibiting a decrease in FR after the whiskers were activated (Chi-square test, units decreasing FR: 3 out of 29 in the WT mice vs 8 out of 24 in the Fmr1 KO, p = .04; units increasing FR: 4 out of 29 vs 2 out of 24 in the WT and Fmr1 KO, respectively, p = .5). The delta change of FR of the units that responded to the stimulus, however, is not significantly different between WT and the Fmr1 KO. These data suggest that there is a potential alteration in the excitatory/inhibitory balance. Specifically, increased basal FR could indicate a tonic (or static) hyperexcitable environment, and an increase in the number of units that decrease their FR after air puff might indicate that inhibition may also be hyperexcitable in Fmr1 KO mice.

Fig. 1.

S1 L2/3 neurons exhibit basal hyperexcitability and an increase in neuronal firing rate suppression during whisker activation in Fmr1 KOs.

(A) Upper panels: Diagram of the experimental set-up of in vivo electrophysiological recording in anesthetized mice. Mice implanted in L2/3 with a movable device with 4 tetrodes were anesthetized and contralateral whiskers were stimulated with an air puff (~3 L/min). Bottom panel: representative electrophysiological recording of one electrode. (B) Comparison of the basal firing rate of all the neurons recorded in the WT (grey) and Fmr1 KO (blue; n = 29 units from 3 WT and n = 24 units from 3 Fmr1 KO; mean ± S.E.M; **p = .02). Inset: cumulative probability function of the data (K–S test p < .005) (C) Left: representative excitatory response (10 trials, spikes and peristimulus time histogram (PSTH) shown, bar 10 spikes) of two units (top panel WT, bottom panel, Fmr1 KO) in response to the stimulus. The air puff was delivered for 1 s. Right: population graph depicting the number and percentage of units that exhibited a statistically significant increase, decrease or no change in FR after stimulus delivery. (D) Left, bar graph showing units separated by their statistical change in firing rate (FR) before and after stimulus presentation in the WT and KO response to the air puff (determined by a t-test corrected for multiple comparison, *p = .04) and population graph depicting the number and percentage of units that exhibited a statistically significant increase, decrease or no change in FR after stimulus delivery. Right, summary of the neuronal response to air puff of the units that exhibited a statistical change in FR after the air puff. (E) Top, diagram of experimental design of DHPG infusion. Twenty trials of baseline activity were recorded in the awake and freely moving mouse. The mouse was then anesthetized and DHPG (100 μM) was infused directly into L2/3. 30 min after infusion, twenty trials of extracellular activity were recorded. Bottom, representative raster plots of a unit during baseline condition and after DHPG infusion in the WT and Fmr1 KO. (F) Left, cumulative probability analysis of the mean baseline FR of awake and freely moving WT and Fmr1 KO mice (n = 16 units and n = 25 units from 3 WT and 3 Fmr1 KO, respectively; mean ± S.E.M; **p = .008). Left inset, bar graph describing the mean basal firing rate with scatter plot overlaid on top (t-test, P = .046). Right, bar graph depicting the delta FR of all the units recorded in basal conditions and after DHPG infusion.

mGluR activation has been widely studied in FXS models and has been shown to be one of the primary pathways affecting neuronal excitability in FXS (Bear et al., 2004). Therefore, we decided to study the effect of the activation of this receptor in vivo. Since the activation of mGluRs varies depending on brain state, specifically modulating neuronal synchronization and excitability, (Kuhn et al., 2008; Song et al., 2018), we decided to infuse the group 1 mGluR agonist, DHPG, in L2/3 of the somatosensory cortex in awake and freely moving mice. We re-designed our implantable recording device by attaching a cannula that allowed us to record electrical activity before and after local DHPG infusion (Fig. 1E). Only units that could be identified in both conditions were included in our analysis. In agreement with our anesthetized mouse data, awake Fmr1 KO mice exhibited a higher firing rate frequency compared to WT in basal conditions (Fig. 1F, left panel, K–S test p = .008; n = 16 units and n = 25 units from 3 WT and 3 Fmr1 KO, respectively), suggesting again that Fmr1 KO neurons are hyperexcited in different brain states. Next, we calculated the delta FR of each of the units before and after 30 min of DHPG (100 μM) infusion (FR after DHPG infusion minus FR before DHPG infusion). We found that some units increased their FR after the drug was applied while others decreased FRs (Fig. 1F, right; 8 units decreased and 8 units increased their FR in the WT while 6 units decreased and 19 increased their FR in the Fmr1 KO). Interestingly, we found that DHPG infusion did not cause a statistically significant difference in neuronal activity for WT or Fmr1 KO mice when compared to the control recordings without the drug, but Fmr1 KO mice did show a tendency toward an overall diminished sensitivity to mGluR activation (Fig. 1F, WT: 97 ± 29.3 Hz and − 13.8 ± 5.1 Hz, KO: 54.6 ± 12.9 Hz and − 18 ± 10.5 Hz, t-test p = .7 and p = .1 WT vs Fmr1 KO for units decreasing and increasing their FR, respectively). Taken together, our data indicates in vivo Fmr1 KO S1 L2/3 neurons are hyperexcited, respond differently to sensory stimuli, and possess mGluR sensitivity that resembles that of the WT mouse.

Decreased neuromodulatory role of mGluRs on inhibitory activity in the somatosensory cortex

In order to determine if E/I balance is also perturbed in vitro, we first recorded basal sIPSCs and sEPSCs in L2/3 of the somatosensory cortex in slice. Compared to our previous study, where recordings were made at room temperature, (Paluszkiewicz et al., 2011), we found that baseline sIPSC frequency was increased in Fmr1 KOs (KO: 7.12 ± 0.46 Hz, n = 38) compared to WT mice (WT: 5.48 ± 0.46 Hz, n = 32, p = .014), suggesting an augmented basal inhibitory activity in L2/3 of Fmr1 KO mice (Fig. 2A and B, upper panels and Fig. 2J). Previous reporting by us and others suggests that DHPG increases the frequency of sIPSCs in P19–30 WT mice cortical layer 2/3 pyramidal cells and that DHPG-induced increase in frequency is attenuated in Fmr1 KOs when higher concentrations of DHPG are applied (100 μM) (Fanselow et al., 2008; Paluszkiewicz et al., 2011). Based on these results, we wanted to assess if mGluR activation is altered in response to physiological levels of glutamatergic activity in FXS. To do this, we examined the effect of DHPG on sIPSCs recorded from pyramidal cells in L2/3 of somatosensory cortex of WT and Fmr1 KO mice in a range of concentrations from 1 to 100 μM (1, 2, 5, 10, and 100 μM; Fig. 2). When DHPG was bath applied on WT L2/3 pyramidal cells, changes in sIPSCs were noticed as low as 10 μM. Notably, WT sIPSC frequency (from 6.30 ± 0.90 Hz to 11.64 ± 1.20 Hz, n = 12, p = 4.5e-5; Fig. 2A, G and K) and amplitude (from 14.75 ± 1.16pA to 21.90 ± 2.47pA, n = 12, p = .005; Fig. 2A, G and L) were significantly increased. However, when DHPG (10 μM) was bath applied on L2/3 pyramidal cells from Fmr1 KO mice, there were significantly smaller changes for sIPSC frequency (from 7.55 ± 0.55 Hz to 10.16 ± 0.91 Hz, p = .025, n = 11; Fig. 2B, G and K) and no changes were observed in amplitude (from 14.08 ± 1.80pA to 22.04 ± 4.79pA, p = .1; n = 11; Fig. 2B, G and L).

Fig. 2.

In vitro recordings show augmented basal inhibition and a diminished inhibitory drive mediated by mGluR activation in Fmr1 KO L2/3 of the somatosensory cortex.

(A-D) Representative voltage clamp traces of sIPSC and sEPSC activity before and after application of DHPG (10 μM) from WT and Fmr1 KO mice, respectively. (E-F) Representative voltage clamp traces for sIPSC frequency before and after application of carbachol (10 μM) from WT and Fmr1 KO mice. (G) Bar population plot of sIPSC frequency from A and B. (H) Bar population plot of sEPSC frequency from C and D (I) Bar population plot of sIPSC frequency from E and F. (J) Single (open) and average (filled) baseline sIPSC frequency from WT and Fmr1 KO mice. (K-L) Logarithmic population plot of sIPSC frequency ratio and sIPSC amplitude ratio, respectively, before and after application of different concentrations of DHPG from WT (black circles) and Fmr1 KO (blue circles). All recordings are from L2/3 pyramidal neurons.

When responses were recorded as a function of DHPG concentration for frequency and amplitude as shown in the normalized DHPG to basal sIPSC activity dose-response plots (Fig. 2K and L), differences were found at 10 μM and 100 μM of DHPG between Fmr1 KO and WT sIPSC frequency, but differences in amplitude were only found at 100 μM. Indeed, when the 10 μM DHPG response was normalized to the baseline frequency activity, the increment of sIPSC frequency was significantly lower in Fmr1 KO than WT counterparts (sIPSC frequency: 1.25 ± 0.19 Hz, n = 11 vs 2.17 ± 0.27 Hz, n = 12; unpaired t-test, p = .039;. Fig. 2K). Moreover, a one-way anova test indicated that the frequency response to DHPG at 10 μM concentration was different between genotypes (F = 6,9: p = .0007).

When recording sEPSCs of L2/3 pyramidal cells from WT and Fmr1 KO mice we found a significant difference in basal sEPSC frequency (WT: 9.84 ± 1.94 Hz, n = 5 and Fmr1 KO:15.19 ± 1.48 Hz, n = 5, t-test p = .035, Fig. 2H) indicating elevated basal excitability. We then tested if DHPG affected sEPSCs and found that with 10 μM of DHPG, WT sEPSC frequency significantly increased (9.84 ± 1.94 Hz to 17.50 ± 1.80 Hz, n = 5, p = .012). However, neither sEPSC amplitude or frequency were significantly changed in the Fmr1 KO mice (10.45 ± 0.48pA to 10.25 ± 0.54pA, n = 5, p = .80; and from 15.19 ± 1.48 Hz to 15.18 ± 2.01 Hz, n = 5, p = .99, respectively; Fig. 2C, D and H), suggesting that mGluR responsiveness is overall affected in the network. In order to understand whether this response is specific for mGluRs or if other neuromodulators are also failing to modulate the inhibitory activity, we bath applied carbachol (10 μM), an agonist of cholinergic receptors. We found that carbachol induced increases of sIPSC frequency (WT: from 4.53 ± 0.54 Hz to 7.73 ± 0.18 Hz, n = 7, p = .0004; KO: from 4.18 ± 0.53 Hz to 7.66 ± 0.23 Hz, n = 10, p = .0001) and amplitude (WT: from 16.67 ± 0.79pA to 52.45 ± 10.28pA, n = 7, p = .014; KO: from 16.89 ± 0.92pA to 59.60 ± 10.34pA, n = 10, p = .002) in both WT and Fmr1 KO pyramidal cell recordings (Fig. 2E, F and I). These results suggest that cholinergic neuromodulation on inhibitory activity of the somatosensory cortex is functional and that mGluR-mediated responses are not the result of the altered inhibitory synaptic properties of the network. Altogether, these data indicate there is a basal increase in sIPSCs frequencies and that there is a loss of inhibitory drive mediated by mGluR activation in FXS.

Heterosynaptic I-LTD is impaired in Fmr1 KO mice

Several studies have reported the critical roles of mGluR-mediated plasticity in FXS (Nosyreva and Huber, 2006; Bianchi et al., 2009; Zhang et al., 2009; Auerbach and Bear, 2010; Connor et al., 2011; Chevere-Torres et al., 2012; Yau et al., 2016), however there is no information on how plasticity at inhibitory synapses is affected in FXS. A well-known and studied form of heterosynaptic I-LTD is an eCB-mediated I-LTD that is triggered by postsynaptic activation of group I metabotropic glutamate receptors (mGluR-I) in the hippocampus (Chevaleyre and Castillo, 2003). In this case, the mGluR activation leads to the production of diacylglyercol (DAG) by phospholipase C (PLC). In turn, diacylglycerol lipase (DGL) converts DAG to the major eCB, 2-AG, which is released from the postsynaptic cell and travels back across the synapse to activate type 1 cannabinoid receptors (CB1Rs) on the GABAergic terminal. CB1R activation subsequently reduces protein kinase A activity, which combined with an enhancement of Calcineurin (CaN) activity may lower the phosphorylation status of an unidentified substrate in the release machinery to persistently depress GABA release (Castillo et al., 2011).

Considering from our previous data that the inhibitory-mediated mGluR responsiveness was decreased, we wanted to understand whether this heterosynaptic I-LTD may be altered in FXS. To do this, we first induced I-LTD through a protocol of high frequency electrical stimulation (HFS) in layer 4 (L4) of S1 and recorded evoked IPSCs (eIPSCs) from pyramidal cells in L2/3. I-LTD was measured by analyzing the percentage of change of the integral under the curve of the eIPSC before and after the stimulation protocol. We found that L2/3 pyramidal cells responded with a long-lasting depression (60.14 ± 10.42% of the baseline, n = 8, p = .032; Fig. 3A) indicating our electrical induction of I-LTD worked. We then tested whether the activation of CB1R was necessary for I-LTD to be maintained, like previously reported. To do this, we bath applied AM251 (4 μM), an inverse agonist of cannabinoid receptor type 1 (CB1R), for 10 min before electrically inducing I-LTD. We found that AM251 blocked the electrically mediated I-LTD (96.08 ± 21.01% of the baseline, n = 4, p = .92; Fig. 3B) suggesting that the activation of CB1R is necessary to mediate the depression of the inhibitory drive into pyramidal cells.

Fig. 3.

Heterosynaptic I-LTD is disrupted in L2/3 of Fmr1 KO mice.

(A) Percentage of eIPSC activity change over time from L2/3 pyramidal cells of WT mice before and after an electrical HFS stimulation and (B) in the presence of AM251 (an eCB receptor inverse agonist). (C) Percentage of eIPSC activity change over time from L2/3 pyramidal cells of Fmr1 KO mice before and after an electrical HFS stimulation. Inset shows representative eIPSC waveform before (1) and after (2) HFS protocol. (D) Percentage of eIPSC activity change over time from L2/3 pyramidal cells of WT mice before and after an electrical HFS stimulation (black filled circles), chemically-induced protocol (grey open circles), in the presence of a mGluR antagonist cocktail of MPEP/LY367385 (blue open circles) or in the presence of H89 (a PKA inhibitor) (green open circles). (E) Percentage of eIPSC activity change over time from L2/3 pyramidal cells of WT (black circles) and Fmr1 KO (blue circles) mice before and after 10 min application of muscarine (10 μM).

Next, we wondered if we could recapitulate the I-LTD seen in WT in Fmr1 KOs. To do this, we electrically induced I-LTD in Fmr1 KO mice and found that eIPSCs recorded from pyramidal cells of Fmr1 KO mice did not change in response to HFS stimulation (88.60 ± 8.04% of the baseline, n = 7, p = .29; Fig. 3C) indicating that I-LTD in Fmr1 KO mice is faulty. To confirm whether the observed WT I-LTD was a heterosynaptic I-LTD, we tested whether mGluRs were critical to this response using a combination of MPEP (4 μM) and LY367385 (100 μM) to block all group I mGluRs. When the MPEP/LY367385 cocktail was applied, eIPSCs in pyramidal cells of WT mice respond differently than the control group to electrical stimulation (one-way anova test, F = 21.75: p < .0001) and did not exhibit I-LTD (83.56 ± 5.90% of the baseline, n = 5, p = .13; Fig. 3D). These data suggest that group I mGluRs participate in the long-term depression synaptic plasticity process. Moreover, since AM251 (4 μM) blocked electrically mediated I-LTD (Fig. 3B), we postulate that mGluR activation may be coupled to the activation of CB1R to mediate the depression of the inhibitory drive into pyramidal cells. Because the heterosynaptic I-LTD also depends on presynaptic PKA activity, we applied H89 (10 μM), an inhibitor of Protein Kinase A (PKA) and found that I-LTD was blocked (97.05 ± 15.06% of the baseline, n = 3, p = .83; Fig. 3D). This data suggests that the observed I-LTD is heterosynaptic and is dependent on mGluR, CB1R and PKA activation in an input specific manner.

To determine whether heterosynaptic I-LTD could also be triggered by other metabotropic receptors, we tested if the mAchR agonist, muscarine, could elicit I-LTD, since it shares common intracellular signaling pathways with mGluRs (Younts et al., 2013; Younts and Castillo, 2014). Bath application of muscarine (10 μM) induced a long-term depression of inhibitory activity in L2/3 pyramidal cells (60.79 ± 6.34% of the baseline, n = 5, p = .0008; Fig. 3E), thereby supporting the fact that mAchR activation can elicit I-LTD in L2/3 of the somatosensory cortex. However, muscarine application onto L2/3 pyramidal cells of Fmr1KO mice failed to induce I-LTD (113.73 ± 13.98% of the baseline, n = 6, p = .51; Fig. 3E) suggesting that I-LTD failure is not only dependent on mGluR activation, but also on the intracellular pathways that are shared with mAchRs. Therefore, these results suggest that group I mGluRs participate in the long-term depression synaptic plasticity process and that faulty I-LTD in FXS is not apparently due to defects in the molecular machinery involved in this phenomena and could be result of a diminished mGluR responsiveness associated to the hyperactive network.

To further evaluate this hypothesis, we tested whether chemical activation of mGluRs could elicit I-LTD. To do this, we applied DHPG (10 μM) for 10 min while recording eIPSCs in WT and Fmr1 KO L2/3 pyramidal cells at 0.25 Hz for a total of 40 min. We found that mGluR activation evoked a chemically-induced I-LTD in L2/3 pyramidal cells from WT mice (62.94 ± 4.71% of the baseline, n = 11, p = .0005; Fig. 4A [upper panel] and C) and that AM251 blocked the chemically-induced I-LTD (99.96 ± 8.60% of the baseline, n = 3, p = .61; Fig. 4A [lower panel] and C), confirming the participation of mGluR in the phenomena. On the other hand, we found that although direct mGluR activation via DHPG in Fmr1 KO L2/3 pyramidal cells induced significant short-term depression (41.81 ± 4.42% of the baseline, n = 7, p = .001; Fig. 4B and D), it failed to evoke long-term responses (93.60 ± 7.82% of the baseline, n = 7, p = .49; Fig. 4B and D). This suggests that the triggering mechanism for heterosynaptic I-LTD might be intact but cannot be sustained long-term in Fmr1 KOs when chemically induced as opposed to electrical induction. Taken together, these data suggest three important phenomena: 1) that inhibitory drive onto L2/3 pyramidal cells of the somatosensory cortex undergo a mGluR-mediated heterosynaptic I-LTD in WT mice that is similar to previous reports (Chevaleyre and Castillo, 2003), 2) that I-LTD plasticity is abnormal in the L2/3 of the somatosensory cortex of Fmr1 KOs and 3) that faulty I-LTD responses in Fmr1 KO mice can either be due to a hyperactivated inhibitory network that cannot further be stimulated or a diminished sensitivity of mGluRs.

Fig. 4.

Chemically induced mGluR activation fails to replicate the heterosynaptic I-LTD in Fmr1 KO mice.

(A) Representative traces of DHPG-induced I-LTD in pyramidal cells of WT L2/3 in the absence (upper trace) and presence (lower trace) of AM251 (4 μM). (B) Representative traces of DHPG-induced I-LTD in L2/3 pyramidal cells of Fmr1 KO in the absence (upper trace) and presence of AM251 (lower trace). (C) Percentage of change in eIPSC activity over time from L2/3 pyramidal cell population of WT mice in the absence (open black circles) and presence of AM251 (filled black circles) before and after 10 min application of DHPG (10 μM). (D) Percentage of change in eIPSC activity over time from L2/3 pyramidal cell population of Fmr1 KO mice in the absence (open blue circles) and presence (filled blue circles) of AM251 before and after 10 min application of DHPG.

The eCB neurotransmission system is not responsible for altered I-LTD in Fmr1 KO mice

Since blocking CB1R activation with AM251 revealed that electrical and chemically induced I-LTD is dependent on intracellular endocannabinoid (eCB) mobilization (Figs. 3 and 4), we next wanted to determine whether eCB mobilization mechanisms are responsible for the faulty I-LTD seen in Fmr1 KOs (Fig. 5). To do this, we examined depolarization suppression of inhibition (DSI). DSI is a known form of short-term plasticity dependent on trans-synaptic eCB mobilization that is triggered by the depolarization of the postsynaptic neuron resulting in decreased GABA release from the presynaptic inhibitory interneuron (Varma et al., 2001). This mechanism resembles the first steps of the I-LTD phenomena and allows us to test whether eCB mobilization mechanisms are functional. In this experiment, because interneuronal populations that are sensitive to mAchR modulation overlap with CB1R-expressing interneurons, we bath applied carbachol (10 μM) to increase the amplitude of the inhibitory drive that is mediated by eCB-sensitive interneurons in L2/3. This strategy has been used before by other researchers to better study the DSI phenomena (Nagode et al., 2014; Vargish et al., 2017). As a result, when L2/3 pyramidal cells were depolarized from −60 to 0 mV for 1 s, sIPSC amplitudes were depressed from the original amplitude in WT (39.24 ± 7.66%, n = 5) and in Fmr1 KOs (51.87 ± 9.38%, n = 9), indicating no significant differences in DSI between Fmr1 KO versus WT mice (p = .32; Fig. 5A, D, and E). These responses also have similar latencies and time courses for WT and Fmr1 KO L2/3 pyramidal cells (Fig. 4D). These results indicate that eCB storage and mobilization are functional in the Fmr1 KO mouse model suggesting that the abnormalities seen in I-LTD are not due to deficiencies in these eCB system properties. Finally, to determine if disrupted I-LTD may be due to deficits in CB1R function, we tested the effect of Win55 (5 μM), a CB1R agonist, in eIPSCs recorded from L2/3 pyramidal cells of WT and Fmr1 KO mice. We found that an equal subset of pyramidal cells were depressed in response to Win55 in WT (7/11) and Fmr1 KO (3/5) with no statistical differences in the magnitude of the CB1R response (42.47 ± 2.21%, n = 7, p = 1.52e-6, and 54.83 ± 2.04%, n = 3, p = .011) (Fig. 5B,C, and F). Thus, we conclude that CB1R sensitivity and the eCB neurotransmission system are not directly responsible in determining the faulty long-term I-LTD seen in Fmr1 KOs.

Fig. 5.

eCB machinery is intact in Fmr1 KO mice.

(A) Representative carbachol induced sIPSCs before, and after a DSI protocol of 1 s of depolarization to 0 mV from WT (black traces), Fmr1 KO (blue traces) and WT in the presence of AM251 (indigo trace) L2/3 pyramidal cell recordings. (B) Representative traces of eIPSCs before and after application of Win55 (5 μM), a CB1R agonist, from WT (black traces) and Fmr1 KO (violet traces) L2/3 pyramidal cells. (C) Data population of percentage of change in eIPSC activity from WT (grey circles) and Fmr1 KO (violet circles) L2/3 pyramidal cells in the presence of Win55 (5 μM) from baseline. (D) Data population of percentage of change in eIPSC activity over time 90 s after the DSI induction protocol for WT (black circles) and Fmr1 KO (blue circles) L2/3 pyramidal cells from baseline. (E) Data population bar plots of maximum percentage of change in eIPSC activity after induction of DSI protocol. (F) Data population bar plots of maximum percentage of change in eIPSC activity after application of Win55 (5 μM).

Discussion

The hyperexcitable phenotype observed in FXS is largely attributed to excessive excitatory drive, yet the role of inhibitory contributions is less understood. Moreover, plasticity at inhibitory synapses is under-studied in light of several studies that have documented faulty synaptic plasticity in Fmr1 KOs. Here, we demonstrate abnormalities in the inhibitory network activity and in mechanisms that dictate inhibitory plasticity and function in L2/3 pyramidal cells in somatosensory cortex of Fmr1 KO mice. Overall, we observe basal hyperexcitability and an increase in neuronal firing rate suppression during whisker activation in vivo and increased basal inhibitory spontaneous currents, a faulty mGluR-mediated activation of inhibitory events, and impaired depressive plasticity at inhibitory synapses in vitro. These phenomena result in an overall enhanced cortical inhibitory synaptic function in L2/3 of the primary somatosensory cortex of Fmr1 KO mice that could potentially mediate a homeostatic counterbalance to the hyperexcitability of FXS cortical circuits found in vivo.

Fmr1 KO S1 L2/3 neurons show hyperexcitability and augmented sensory response

Previous behavioral experiments have shown that Fmr1 KO mice exhibit an exaggerated response to whisker stimulation, a failure to adapt to repetitive stimuli and an impaired learning in a whisker-dependent behavioral task (Arnett et al., 2014; He et al., 2017). At the neuronal level, in vivo experiments have shown that neurons of the somatosensory cortex are hyperexcited (Goncalves et al., 2013; Zhang et al., 2014). In agreement with this, we found that neurons in Fmr1 KOs fired at a higher rate in basal conditions throughout different brain states (anesthetized and awake). Interestingly, a recent study presented results suggesting that spontaneous basal cortical activity is similar in both Fmr1 KO and WT animals (Antoine et al., 2019). Thus presenting a situation which deviates from our findings and others (Goncalves et al., 2013; Zhang et al., 2014). We believe these differences in anesthetized and awake recording results may rely on experimental design. Specifically, we used different anesthesia drugs (isoflurane vs urethane/chlorprothixene) and awake behavioral paradigms (freely moving vs head fixed) in our experiments, both possibly having an impact on network excitability and the outcome of these experiments. Moreover, they show that neurons in the L2/3 of the Fmr1 KO exhibit spatially broader cortical activation to single-whisker stimulation and a decreased whisker-evoked firing rates in response to multiple deflection velocities (Antoine et al., 2019). In agreement with these results, we found that neurons responded differently to stimulus, since more cortical neurons in the Fmr1 KO mice decreased their FR in response to a single air puff when compared to WT mice. In this case, whisker stimulation is probably robust enough in both scenarios to engage the network overpowering the basal level activity differences. We also found that DHPG infusion on average increased the FR in both WT and Fmr1 KO L2/3 neurons in vivo. However, there was a trend showing a diminished FR increments (Fig. 1F) in response to mGluR activation in Fmr1 KO mice; although statistical differences were not detected between genotypes. This could suggest either diminished mGluR sensitivity or an activity ceiling effect that overall restrains the network to reach increased firing rate activity. Understanding that the complexity of the data collected from the somatosensory cortical network could hide relevant information at the cellular level in this context, therefore we performed the in vitro experiments.

Taken together, our data suggests that interneuron activity might attempt to counteract alterations in somatosensory coding. Additional in vivo electrophysiological experiments could elucidate how different types of somatosensory inputs affect neuronal and network activity in L2/3 and how population coding within all cortical layers leads to hypersensitive driven behaviors seen in FXS. Moreover, selective identification and neuronal activity regulation of different neuronal populations, with techniques such as optogenetic, could shed light on the role excitatory and inhibitory neurons play in vivo.

Enhanced inhibitory cortical activity impacts mGluR-mediated modulation

Similar to our results found in vivo, we found increased basal neuronal activity in Fmr1 KO mice. In particular, inhibitory spontaneous activity was higher than in WTs. Although others reported that inhibition is decreased in the somatosensory cortex of Fmr1 KO, our results do not contradict this finding (Antoine et al., 2019). While their finding is in the context of feedforward inhibition from L4 to L2/3, which is driven by a specific cortical circuit, we find that basal sIPSCs are elevated in Fmr1 KO mice, which is a read out of the overall inhibitory activity in the network within L2/3. We postulate that the observed elevated basal sIPSC activity is most likely caused by an increased excitatory neurotransmission that drives more inhibition into L2/3 pyramidal cells and/or hyperactive interneuron subtypes that are attempting to homeostatically counterbalance cortical hyperexcitability. Moreover, we found a decreased mGluR-driven spontaneous inhibitory postsynaptic input to pyramidal cells in Fmr1 KO mice. This loss of mGluR responsiveness could be due to a ceiling effect caused by the hyperactive network, although we cannot discard mGluR changes in expression, decreased membrane localization or structural abnormalities.

Inhibitory plasticity is altered in Fmr1 KO mice

At this time, we are the first to demonstrate that, in contrast to excitatory synapses, LTD at inhibitory synapses (I-LTD) is abolished in the Fmr1 KO mice. Although these two LTD processes (excitatory and inhibitory LTDs) follow different mechanistic and intracellular pathways of activation, they have at least one critical point in common - mGluR activation. To this point, our results reveal a faulty mGluR-mediated I-LTD in L2/3 of the somatosensory cortex of Fmr1 KOs. This I-LTD is also dependent on eCB release, CB1R activation, and PKA function as is suggested by the lack of the I-LTD in WT mice when mGluRs, CB1Rs and PKA activity were blocked. These are all molecular components of a heterosynaptic I-LTD, suggesting that this phenomenon is similar to the one observed in the hippocampus (Chevaleyre and Castillo, 2003; Chevaleyre et al., 2007).

Furthermore, previous reports suggest the possibility of direct participation of a faulty enzymatic complex to synthetize eCB in FXS and show that dysfunctional eCB molecular machinery is altered in excitatory synapses without affecting inhibitory connectivity (Maccarrone et al., 2010; Jung et al., 2012). However, our results indicate that eCB activity is unaltered in terms of release and storage availability of eCBs, and that CB1R receptor sensitivity is maintained in Fmr1 KO mice. Therefore, faulty eCB synthesis is not likely to be the main cause of this faulty heterosynaptic I-LTD in FXS (Fig. 5). Although this finding does not discard a role for the synthesis of eCBs, it may further support a secondary role for eCB receptor activation in the faulty heterosynaptic I-LTD of the somatosensory cortex.

Taken together, it seems that mGluR-mediated I-LTD is lacking in cortical pyramidal cells from Fmr1 KO because of diminished mGluR responsiveness due to a ceiling effect associated with a hyperactive neuronal network rather than molecular components malfunction. This type of association has been shown before where network activity levels modify the threshold for synaptic transmission and plasticity (Felix-Oliveira et al., 2014).

Is faulty inhibitory plasticity in Fmr1 KOs cell-type specific?

Based on the mechanisms of inhibitory plasticity, our results suggest different interneuron subtypes are contributing to the abnormal activity of the somatosensory cortex of Fmr1 KO mice. However, previous reports indicate that Sst-LTS interneurons are especially sensitive to activation via mGluRs (Fanselow et al., 2008; Paluszkiewicz et al., 2011). Based on those findings, we speculate that the diminished inhibitory response to DHPG recorded from L2/3 pyramidal cells in Fmr1 KO S1 could be due to the loss of mGluR activation responsiveness in the Sst-LTS cells. On the other hand, because the mechanism of heterosynaptic I-LTD requires the eCB molecular machinery to be activated, as documented previously (Chevaleyre et al., 2007), it is highly likely that interneurons participating in this mechanism of plasticity express CB1Rs. The main candidate for this type of response is CB1R-expressing perisomatic-targeting basket cells (BCs) that belong to the 5HT-3R expressing group of interneurons in the cortex (Lee et al., 2010), which are also characterized by the expression of cholecystokinin (CCK), as is documented in hippocampus and cortex (Foldy et al., 2007; De-May and Ali, 2013). The abnormal function of these interneurons, resulting from a faulty mGluR modulation, would have different implications in the activity of the cortical network. Although these two interneuron cell types (Sst-LTS and CCK-BCs) modulate the network in similar fashion by having a role in the fine tuning of the information being processed in the cortex (Lee et al., 2010; Rudy et al., 2011), they have different postsynaptic targets (perisomatic vs dendritic) and timing of the response associated to mGluR activation (short vs long-term) (Chevaleyre and Castillo, 2003; Marinelli et al., 2008). Thus, we speculate that the faulty modulation on the dendritic targeting Sst-LTS cells would lead to an immediate but transient truncated inhibitory control over the excitatory activity of the somatosensory cortex, and the unregulated I-LTD mediated by CCK-BCs would lead to a loss of the long-term control of the inhibitory drive in the Fmr1 KO mice. Interestingly, PV-positive interneurons have been shown to decrease their output activity in the Fmr1 KO cortex (Gibson et al., 2008; Goel et al., 2018; Antoine et al., 2019) suggesting a diminished inhibitory drive in the network opposite to what is presented here; however, our data does not contradict those findings, considering that interneurons have different properties in the network and their contribution to the phenotype is cell-type specific in the somatosensory cortex.

In a like manner, the mAchR response reported here could also be determined by interneuron classes. Interestingly, we found that mAChR-mediated I-LTD is disrupted in Fmr1 KO mice and that carbachol-enhanced inhibitory drive in our DSI experiment is not. We think that although these are opposites effects on the inhibitory drive in our I-LTD and DSI experiments, this is expected considering the differences in experimental designs. I-LTD is done by electrical stimulation of L4 of somatosensory cortex, leading to eIPSCs mediated by different inhibitory cells where only a fraction can be eCB downregulated, and DSI is measured by sIPSCs where mAchR activation selectively enhances the inhibitory responses of those cells that can be modulated by eCB release, giving a more specific downregulation of the inhibitory activity. Thus, mAchR activation during I-LTD and DSI responses incorporate different cellular components. Lastly, despite the fact that M4 mAchR translation is upregulated in Fmr1 KO mice and its activation can restore both protein synthesis levels and excitatory LTD in the hippocampus (Thomson et al., 2017), the mAchR activation in Fmr1 KO in this study did not restore the I-LTD phenomena seen in WT mice, possibly as a result of our nonspecific treatment.

Conclusion

The findings of this manuscript suggest that cortical network inhibition is attempting to counterbalance the hyperexcitability observed in the Fmr1 KO mouse model of FXS. At baseline levels, inhibitory synaptic activity (in the form of sIPSCs) is higher in the somatosensory cortex of Fmr1 KOs suggesting there is an attempt to balance the heightened levels of excitatory activity observed in this disorder (Wang et al., 2017). Additionally, in the long-term, inhibition is enhanced in the form of a faulty I-LTD, which allows a persistent inhibitory drive into postsynaptic cells in periods of long-lasting high cortical network activity. Indeed, the loss of mGluR activation responsiveness in the Fmr1 KOs could be explained by the increased basal inhibitory drive in the disorder, seemingly at ceiling levels. This would limit the role of mGluRs on activating the network at both baseline and long-term levels. Additionally, this heightened activity likely comes at the cost of a loss of inhibitory synaptic plasticity and signal to noise cortical detection. The influence of this hyperactive cortical network was recently demonstrated by Domanski and colleagues where they showed that during recurrent activity at behaviorally-relevant frequencies, there is decreased and disorganized spiking of L4 neurons (the major excitatory input to L2/3) (Domanski et al., 2019). This would also suggest that sensory information being relayed to L2/3 is most likely dramatically altered (Domanski et al., 2019). Altogether, these data suggest that heightened inhibition is attempting to maintain E/I balance. However, enhanced excitatory and inhibitory activity in the network could lead to faulty sensory signal detection, codification and interpretation resulting in altered somatosensory perception as seen in FXS.