Reduced Gamma Oscillations in a Mouse Model of Intellectual Disability: A Role for Impaired Repetitive Neurotransmission?

Intellectual disability affects 2–3% of the population; mutations of the X-chromosome are a major cause of moderate to severe cases. The link between the molecular consequences of the mutation and impaired cognitive function remains unclear. Loss of function mutations of oligophrenin-1 (OPHN1) disrupt Rho-GTPase signalling. Here we demonstrate abnormal neurotransmission at CA3 synapses in hippocampal slices from Ophn1 -/y mice, resulting from a substantial decrease in the readily releasable pool of vesicles. As a result, synaptic transmission fails at high frequencies required for oscillations associated with cognitive functions. Both spontaneous and KA-induced gamma oscillations were reduced in Ophn1 -/y hippocampal slices. Spontaneous oscillations were rapidly rescued by inhibition of the downstream signalling pathway of oligophrenin-1. These findings suggest that the intellectual disability due to mutations of oligophrenin-1 results from a synaptopathy and consequent network malfunction, providing a plausible mechanism for the learning disabilities. Furthermore, they raise the prospect of drug treatments for affected individuals.


Introduction
The X-linked mental retardation (XLMR) disorders are a group of single gene mutations which lead to moderate to severe, nonspecific, intellectual disability [1]. An ongoing major challenge is to understand the cellular mechanisms responsible for the cognitive impairment. Recent work has challenged the dogma that dendritic spine abnormalities are responsible for the underlying pathophysiology. It has led to the hypothesis that intellectual disabilities may be described as synaptopathies because several have been associated with alterations in synaptic function [2][3][4]. How these alterations in synaptic function affect brain function in ways that explain cognitive impairment remains unknown.
Synchronisation of neuronal activity in the gamma  Hz) frequency band is thought to underlie the encoding and retrieval of episodic memory, attention and the formation of neuronal assemblies that facilitate associative learning [5,6]. Gamma oscillations have also been proposed to provide the spike timing that facilitates the synaptic summation necessary for long-term potentiation [7,8]. Gamma oscillations are frequently studied in the hippocampus, where they have been suggested to function in exploratory behaviour and navigation [9][10][11] and gamma oscillation strength is correlated with behavioural performance [12]. Altered neuronal synchrony in the gamma band has been reported for a variety of neuropsychiatric disorders [5] and neurodevelopmental disorders, including autism, Williams syndrome [13] and the Ts65Dn Down syndrome model of intellectual disability [14].
Gamma oscillations can be generated in vitro in the CA3 region of the hippocampus by tonic activation of kainate (KA) receptors [15] or can occur spontaneously [16,17]. Gamma and other cortical oscillations associated with cognition depend on the repetitive activity of inhibitory neurons and intermittent activity in excitatory neurons, and so the ability of inhibitory synapses to operate at these frequencies is critical [6,18].
Loss of function mutations in the OPHN1 gene (Ophn1 in mice) result in a moderate to severe learning disability in humans [19] and learning impairments in mice [20]. OPHN1 encodes oligophrenin-1, a protein with a Rho-GAP domain which negatively regulates RhoA, Rac1 and Cdc42 [19,21]. In addition to the regulation of Rho-family GTPases, oligophrenin-1 regulates the size of the readily releasable pool (RRP) of vesicles in inhibitory synapses [4], possibly through regulation of synaptic vesicle endocytosis [22,23]. The altered vesicle dynamics prevents synapses from functioning at frequencies within the gamma range [4].
In the present study, we investigate the role of oligophrenin-1 in neuronal network activity, particularly spontaneous and KAinduced gamma oscillations using the Ophn1 mouse model of intellectual disability. We demonstrate that hippocampal synapses are unable to function at frequencies necessary for higher cognitive function, due to a substantial decrease in the RRP of synaptic vesicles. We propose that these synaptic changes underlie the deficits in gamma oscillations reported here. The alterations in spontaneous gamma oscillations were abolished by inhibition of the RhoA signalling pathway. The rescue of emergent neuronal network activity by small molecule pharmacological inhibition of the downstream signalling pathway of oligophrenin-1 raises the possibility of a pharmacotherapy to treat affected individuals.

Ethics statement
The Ophn1 colony was generated using a C57-BL6 background as described previously [20]. Ophn1 -/y mice and wild type (Ophn1 +/y ) littermates were generated by breeding heterozygote females (Ophn1 +/2 ) with Ophn1 +/y males; resulting in ,50:50 Ophn1 +/y : Ophn1 2/y . Breeding and experiments were performed under regulation by the Animals (Scientific Procedures) Act (1986) of the UK, and approved by the Biomedical Ethics Review sub-committee (BERSC) at the University of Birmingham. All experiments and analyses were performed blind to genotype.

Extracellular recordings
After sectioning, slices were placed in a Haas-type interface recording chamber and allowed to equilibrate for an hour at the interface between aCSF and moist 95%O 2 -5%CO 2 (300 cm 3 / min). Slices were constantly perfused with aCSF at a flow rate of ,2 ml/min; the temperature was maintained at 32uC. Slices were visualised with a stereo-microscope (Leica MZ8, Micro Instruments, Long Hanborough, Oxon, UK) mounted above the interface chamber. Extracellular microelectrodes were pulled from thick-walled borosilicate glass capillaries (1.2 mm O.D.60.69 mm I.D.; Harvard apparatus, Edenbridge, Kent, UK) using a P-97 puller (Sutter Instrument Co, Novato, CA). Electrodes were filled with aCSF and had a typical resistance of 2-4 MV. Extracellular potentials were recorded using an Axoclamp 2B amplifier (Molecular Devices, Sunnyvale, CA), low pass Bessel filtered at 1 kHz (NL-125, Digitimer Ltd, Welwyn Garden City, UK) and digitized at 10 kHz by a Power 1401 (CED Ltd, Cambridge, UK). Additionally, a Humbug 50/60 Hz (Digitimer) was used to remove noise locked to the electrical mains supply. Stimulation and data acquisition were controlled using Spike 2 software (v6.12; CED). Data were stored for subsequent off-line analysis using Spike 2.

Gamma oscillations
After slices were placed in the interface recording chamber, an extracellular recording electrode was placed in the pyramidal cell layer of CA3c. Spontaneous activity was recorded for 5 minutes before the addition of 50 nM KA. The strength of the gamma oscillation in the frequency range 20-80 Hz was measured by fast Fourier transforms of 60 s epochs of data (Hanning window, FFT size 4096) and quantified as summated power. The dominant frequency in this range was used to quantify the peak frequency. Spontaneous gamma oscillations were observed in a subset of slices (Ophn1 +/y 8 of 21 slices [,38%]; Ophn1 2/y 10 of 30 slices [,33%]), characterised by a peak power in the gamma range (average power in 25-35 Hz is greater than average power in 10-20 Hz) [16].
Waveform averages were calculated from 30-50 gamma cycles, phase zeroed at the peak of the gamma cycle recorded in s. pyramidale. To avoid potential bias between genotypes, the following method was used to select the gamma cycles for averaging. A 2 s epoch of data was band pass filtered (20-80 Hz, 2 nd order Butterworth digital filter) and events between 30 and 70% of the maximum filtered signal were selected for averaging.
The effect of Rho kinase (ROCK) inhibitors on gamma oscillations were evaluated by addition of Y-27632 (trans-4-[(1R)-1-Aminoethyl]-N-4-pyridinylcyclohexanecarboxamide dihydrochloride; 10 mM) to the superfusate for 20 minutes before the addition of 50 nM KA. The effect of Y-27632 on spontaneous gamma oscillations was tested immediately before addition of KA and on KA-induced gamma oscillations after 1 hour of KA.

Evoked field potentials
The mossy fibre tract was stimulated using a nichrome concentric stimulating electrode (80% nickel/20% chromium, tip diameter 50 mm; Advent research materials; Oxford, UK) placed in the hilus. Postsynaptic potentials (PSPs) were recorded from s. radiatum in CA3c. Stimulus-response curves were generated to determine the half maximal stimulus intensity values (20.860.6 V, range 15-27 V, n = 31). This intensity was used for subsequent high frequency stimulation recordings. High frequency stimulation was evoked by a train of 10 stimuli at 33 Hz and PSP slopes were measured and normalised to the first response. KOH (osmolarity ,285 mOsm). EPSCs were recorded at 2 75 mV to avoid contamination by IPSCs (reversal potential of IPSCs) and were confirmed as EPSCs by their abolition by 20 mM NBQX and 25 mM D-APV (not shown). Patch electrodes typically had resistances of 4-7 MV. Membrane potentials and currents were recorded using an NPI SEC-10L amplifier (Scientifica, Harpenden, UK), low pass Bessel filtered at 1 kHz (NL-125, Digitimer Ltd, Welwyn Garden City, UK) and digitized at 10 kHz by a Power 1401 (CED Ltd, Cambridge, UK). Stimulation and data acquisition were controlled using Signal software (v3.10; CED). Evoked currents were elicited by a concentric stimulating electrode placed in the hilar region.

Whole-cell patch-clamp recordings
The readily releasable pool (mature and fusion-competent) was quantified by cumulative evoked IPSC amplitude analysis during repetitive stimulation [24]. The intracellular solution was supplemented with 100 mM Alexa 488 to enable visualisation of dendrites. A stimulating electrode (2 nd patch pipette filled with aCSF) was placed near the dendritic tree ,50 mm from the soma. Stimulus intensity was varied to achieve minimal stimulation (activation of a single axon/synapse) as described in Powell et al (2012). The RRP size and probability of vesicle release from the RRP were calculated by applying repetitive stimuli (40 pulses at 20 Hz) and calculated as outlined in the main text.

Statistical analysis
Values are expressed as mean6S.E.M. Curve fitting and data analysis were performed in Origin 8 (Silverdale Scientific, Stoke Mandeville, UK). Unpaired t-tests (with Bonferonni correction for multiple comparisons) were used to evaluate differences between genotypes; a Shapiro-Wilk test was used to examine if the data followed a normal distribution. Repetitive stimulation experiments were analysed using ANOVA. Significance criterion was p,0.05.

Drugs
All chemicals were purchased from Abcam Biochemicals (Cambridge, UK).

Results
We have previously demonstrated that repetitive inhibitory neurotransmission onto dentate gyrus granule neurons is reduced due to a smaller RRP [4]. We first tested whether the central conclusions from dentate granule cells generalised to CA3 pyramidal neurons. IPSCs evoked by stimulating the mossy fibre inputs were significantly smaller in Ophn1 2/y neurons than Ophn1 +/y neurons (15736173 pA, n = 15 and 23176279 pA, n = 13, respectively; p = 0.02 ; Fig 1a,  Ophn1 null neurons display altered responses to repetitive stimuli We examined the ability of synapses to follow high frequency stimulation [25] at frequencies associated with cognition ($33 Hz) [6]. Using a submaximal stimulus (evoked ,30% of maximum evoked IPSC), IPSCs built up with successive stimuli at 33 Hz, reaching a steady level within 10 stimuli in Ophn1 +/y neurons (144614%, n = 14; Fig 2a, b). IPSC facilitation did not occur in Ophn1 2/y neurons (82612%, n = 20; p = 0.003; Fig 2c). Facilitation was also weakened in Ophn1 2/y neurons for 50 and 100 Hz stimulation (Fig 2c). As oligophrenin-1 controls the size of the RRP [4], we reasoned that this may explain the altered response to repetitive stimuli in Ophn1 2/y CA3 neurons. Using high frequency minimal stimulation (40 pulses, 20 Hz) [4,24], we examined the functional state of the RRP in CA3 neurons. The RRP was significantly smaller in Ophn1 2/y neurons than in Ophn1 +/y neurons (513.86100.9 pA, and 1466.26307.7, respectively, n = 7; p = 0.005 ; Fig 3b, c), corresponding to a smaller number of vesicles in the RRP (1262 and 3467, respectively; p = 0.005; Fig 3e). The reduction in RRP did not alter the mean amplitude of the first IPSC (Ophn1 2/y 97.8619.9 pA and Ophn1 +/y 75.4616.3 pA, p = 0.84 ; Fig 3c), which means that the probability of releasing single vesicles from the RRP (P ves ) was significantly greater in Ophn1 2/y neurons (Ophn1 2/y 0.2660.06 and Ophn1 +/y 0.0860.01, p,0.01; Fig 3d). Having demonstrated that inhibitory neurotransmission in the CA3 region is affected in a manner similar to that in the dentate gyrus [4], we examined whether these observations translated to excitatory transmission. Extracellular field potentials were recorded from CA3 s. radiatum and PSPs were evoked by activation of the mossy fibre pathway. PSPs were reduced in Ophn1 2/y slices (Fig 4a, B; p,0.01) and the frequency of spontaneous excitatory postsynaptic currents recorded from CA3 pyramidal neurons was lower in Ophn1 2/y neurons than Ophn1 +/y neurons (1.760.4 Hz, n = 9 and 6.961.5 Hz, n = 8, respectively; p = 0.003; Fig4 c-f).
Altered gamma oscillations in Ophn1 2/y slices Synchronisation of neuronal activity in the gamma frequency range has been associated with cognitive function [6] and can be recorded in vitro either spontaneously [16] or generated by superfusion of KA [6].
Hippocampal slices were placed in the interface recording chamber and an extracellular recording electrode was placed in the pyramidal cell layer of CA3c. Spontaneous activity was recorded for 5 minutes before addition of KA (Fig 6a, 7a). The summated power of spontaneous gamma oscillations was smaller in Ophn1 2/y slices than in Ophn1 +/y slices (27.567.6 mV 2 , n = 8 and 66.0613.7 mV 2 , n = 10, respectively, p = 0.03; Fig 6b). Calculation of the waveform average for spontaneous gamma oscillations in Ophn1 +/y and Ophn1 2/y slices revealed that, despite the reduction in overall gamma power, the shape of the gamma cycle was unaltered (Fig 6c, d).
Superfusion of KA (50 nM) increased CA3 network activity throughout its application in both Ophn1 +/y and Ophn1 2/y slices (Fig 7a, b). The frequency of oscillations increased in Ophn1 2/y slices from 26.260.9 Hz for baseline activity to 34.361.1 Hz after 1 hr (n = 18; Fig 7d). A similar increase in frequency was observed   Fig 7c). Analysis of the representative spectrograms for the development of KA-induced oscillations did not reveal major differences between Ophn1 +/y and Ophn1 2/y slices (Fig 7b), indeed it appeared to be a simple loss of power because the average gamma waveform was unaltered (Fig 8b).
Given that Ophn1 2/y mice show alterations in dendritic morphology [4,26], we reasoned that associated changes in connectivity may affect the spatial distribution of KA-induced gamma oscillations. To compare the spatial distribution and coherence of KA-induced gamma oscillations, field recordings were made along the pyramidal layer of the CA regions. A stationary electrode was placed in CA3c and a roving electrode was placed in CA3b, CA3a and CA1 in turn. The maximum  cross-correlation and phase difference were determined from cross-correlograms using CA3c as the reference channel. As shown previously [16], KA-induced gamma oscillations were coherent throughout the CA regions, but the maximum cross-correlation fell with increasing electrode separation. No significant differences between Ophn1 +/y and Ophn1 2/y slices were observed for spatial profile or coherence (Fig 8c, d).

Sensitivity of reduced gamma oscillations to ROCK inhibitors
As the synaptic deficits in Ophn1 2/y dentate gyrus granule cells are reversed by acute inhibition of ROCK activity [4], we examined whether the deficit in network oscillations could also be reversed. Application of Y-27632 (10 mM) for 20 minutes prior to superfusion of KA did not reverse the impairment in Ophn1 2/y slices, but rather depressed KA-induced gamma oscillations in both Ophn1 +/y and Ophn1 2/y slices. After 60 minutes of KA, Y-27632 had reduced gamma oscillations in Ophn1 +/y slices to ,50% of those in vehicle controls (2315.361056.6 mV 2 , n = 7 and 6146.861317.9 mV 2 , n = 10, respectively, p = 0.04; Fig 9a). Application of Y-27632 to Ophn1 2/y slices caused a similar reduction in KA-induced gamma oscillations from vehicle controls (1272.86197.3 mV 2 , n = 5 and 3638.06823.2 mV 2 , n = 11, respectively, p = 0.02; Fig 9a).
Inhibition of ROCK by Y-27632 in Ophn1 2/y slices significantly increased spontaneous oscillations, in contrast to the reduction of KA-induced gamma oscillations (vehicle 71.5619.2 mV 2 , Y-27632 215.8651.8 mV 2 , p = 0.005; Fig 9b). In Ophn1 +/y slices, Y-27632 did not affect spontaneous oscillations (vehicle 137.5620.9 mV 2 , Y-27632 180.8640.9 mV 2 , p = 0.32; Fig 9b). The different effect of Y-27632 on spontaneous and KA-induced gamma oscillations was intriguing, but it may suggest that distinct network mechanisms generate spontaneous and KA-induced gamma oscillations [16].

Discussion
The pathophysiology of intellectual disability is poorly understood, but mutations in single genes that result in cognitive deficits provide unique opportunities to discover the underlying mechanisms. Increasing evidence links synaptic malfunction to cognitive deficits [2,27]. Mutation of the OPHN1 gene in humans produces a non-specific X-linked mental retardation [19]. Reduced expression of oligophrenin-1, the protein encoded by Ophn1, results in altered dendritic spine morphology [20,26] and learning impairments in mice [20]. Oligophrenin-1 has been previously shown to modulate endocytosis [22,28], which alters responses to repetitive stimuli in excitatory synapses [23]. In contrast, its role in inhibitory terminals is less clear. In dentate gyrus granule neurons, it functioned as a regulator of synaptic vesicle availability [4], whereas in cultured CA1 neurons, oligophrenin-1 expression did not affect inhibitory transmission [28]. To further understand the role of oligophrenin-1 in cognitive deficits, we examined synchronous neuronal activity in the gamma frequency band. Reduced neuronal synchrony in the gamma range has been reported for various pathological conditions including Alzheimer's disease [29], ageing [30], schizophrenia [31] and autism [13]. Abnormal EEG recordings have long been associated with intellectual disabilities in humans [32], although until recently the link between mouse models and human disease was lacking [14]. Here we show that gamma oscillations are reduced in Ophn1 2/y mice and propose that this is a result of the reduction in synaptic strength which arises from deficits in vesicle recycling.
The observation that both KA-induced and spontaneous gamma oscillations are reduced in Ophn1 2/y slices is not surprising given the dependence of both on inhibitory neurotransmission [16] and that gamma oscillations place a high demand on inhibitory transmission [33,34]. Kainate, acting via presynaptic receptors [35], has been demonstrated to facilitate neurotransmitter release via an alteration of the RRP [36]. Therefore the smaller RRP reported in this study could indeed explain the reduced gamma oscillation. Previous studies have demonstrated that the strength of the inhibitory input to pyramidal neurons modulates the power of gamma oscillations without changing the dominant frequency [15,25]. Therefore, the alteration in vesicle availability appears the most likely candidate to explain the decreased power. Other potential explanations include alterations in the intrinsic excitability, although this seems unlikely as the intrinsic properties and firing frequency of interneurons and pyramidal neurons were not affected by the loss of expression of oligophrenin-1 (data not shown). An alternative explanation for the reduced gamma oscillations is that it results from fewer synapses on Ophn1 2/y neurons. The rapid reversal of all the synaptic deficits [4] and spontaneous gamma oscillations (this study) argues against changes in synaptic numbers because it is unlikely that the doubling in numbers of synapses required to explain the results can occur in the rapid timescale shown here. Furthermore, the difference in RRP measured at single synapses in both the CA3 (this study) and dentate gyrus [4] using minimal stimulation demonstrates a direct functional impairment of synaptic mechanisms in Ophn1 2/y mice. Pharmacological reversal of gamma oscillations in vitro Despite learning disabilities being common disorders (affecting 2-3% of the population), the most effective treatment options remain special-needs education and similar approaches. Recent research has suggested that intellectual disabilities may be treatable with pharmacotherapy [37]. The cognitive deficits reported for Ophn1 2/y mice have been proposed to result from over-activation of the RhoA signalling pathways [20]. Inhibition of the major downstream effector ROCK abolishes presynaptic changes in vesicle availability [4], postsynaptic alterations in synaptic plasticity [22], alterations in dendritic spine morphology [26] and more recently behavioural deficits [38]. In this study we have examined whether inhibition of ROCK activity could restore the altered gamma oscillations in Ophn1 2/y mice. Pre-treatment with Y-27632 restored spontaneous gamma oscillations to wild type levels in Ophn1-/y slices, but was without effect on these oscillations in wild type slices. The reversal of spontaneous gamma presumably reflects the restoration of synaptic transmission towards normal levels. In contrast, Y-27632 reduced KA-induced gamma oscillations in both Ophn1-/y and Ophn1+/y slices, which was an unexpected observation given that it reverses the synaptic changes [4] and dendritic spine abnormalities [26] associated with reduced oligophrenin-1 expression. As the requirement for fast AMPA receptor-mediated synaptic transmission differs between KA-induced and spontaneous gamma oscillations [16,39] the reversal of spontaneous, but not KA-induced oscillations may reflect selective reversal of excitatory transmission (Fig 4) with GABA-ergic transmission unaffected. The synaptic phenotype of impaired vesicle dynamics for both excitatory and inhibitory neurotransmission in CA3 synapses is analogous to that described in the dentate gyrus synapses where full reversal of inhibitory transmission was achieved [4]. So while not directly tested here, it is unlikely that the differential effects of Y-27632 are due to selective reversal of excitatory neurotransmission.
The neuronal networks that underlie the generation of spontaneous and KA-induced oscillations have been demonstrated to be dependent on different subpopulations of interneurons [16,34,[40][41][42]. In KA-induced gamma oscillations, electrotonically-coupled parvalbumin-expressing basket cells are responsible for the phasic inhibition that synchronises pyramidal cell firing [11,43], although multiple interneuron subtypes are active during gamma oscillations [34]. Spontaneous oscillations are also dependent on phasic inhibition, however current source density analysis suggests that s. lucidum interneurons are responsible, with 'mossy fibre associated' interneurons being the most likely candidate [16]. It is therefore possible that the differential effect of Y-27632 on KA-evoked and spontaneous gamma results from ROCK exerting different control of synaptic vesicle dynamics in the different interneuron populations, although this remains to be established.
Given that ROCK inhibition affected KA-induced oscillations to a similar extent in both wild-type and knockout slices, it is possible that the methodology used to generate gamma oscillations determines the effect of Y-27632. This has been demonstrated for the NR2B-selective NMDA receptor antagonist Ro25-6981 which potentiates KA-induced [14], but reduces tetanus-evoked gamma oscillations [44]. It would be interesting to examine the effects of inhibition of ROCK on gamma oscillations generated by activation of muscarinic receptors, as these are thought to be most similar to spontaneous oscillations [16].
Inhibition of ROCK may have non-specific effects on the network responsible for generation of KA-induced oscillations. For example, KA-induced oscillations are dependent on electrotonic coupling [45,46] which may be altered by inhibition of ROCK [47]. Additionally, Y-27632 has been shown to modulate various ionic conductances such as Nav1.5 [48] and alter K + -induced contraction of smooth muscle [49], suggesting off-target effects which may specifically alter KA-induced oscillations. Therefore further studies are required to establish the underlying cause of the divergent effects of ROCK inhibition.  In this study we have examined whether inhibition of ROCK activity could restore the altered gamma oscillations in Ophn1 2/y mice. A recent report demonstrated that peripheral application of ROCK inhibitors can improve spatial and working memory in rodents [50], raising the possibility that the use of ROCK inhibitors may provide a suitable pharmacotherapy for the reversal of cognitive deficits in this intellectual disability.