Defective synapse maturation and enhanced synaptic plasticity in Shank2-/- mice

Autism spectrum disorders (ASDs) are neurodevelopmental disorders with a strong genetic aetiology. Since mutations in human SHANK genes have been found in patients with autism, genetic mouse models are employed for a mechanistic understanding of ASDs and the development of therapeutic strategies. In sharp contrast to all studies so far on the function of SHANK proteins, we observe enhanced synaptic plasticity in Shank2-/- mice, under various conditions in vitro and in vivo. Reproducing and extending previous results, we here present a plausible mechanistic explanation for the mutants' increased capacity for long-term potentiation (LTP) by describing a synaptic maturation deficit in Shank2-/- mice.


Introduction
The activity-dependent formation and remodelling of synaptic connections is pivotal to adaptive neural circuit function. Dysregulation of these processes is considered a prime cause of neurodevelopmental diseases such as autism spectrum disorders (ASDs) (Ebert and Greenberg, 2013). The list of mutations associated with ASDs and the number of mouse models is growing rapidly, yet understanding the connection between genetic mutation, synaptic defects, and disease phenotypes remains a challenge. Mutations in SHANK genes occur in autistic patients (Durand et al., 2007;Berkel et al., 2010;Sato et al., 2012) and Shank mutant mice reproduce several autism-related phenotypes (Bozdagi et al., 2010;Peça et al., 2011;Wang et al., 2011;Schmeisser et al., 2012;Won et al., 2012;Lee et al., 2015;Peter et al., 2016). SHANKs (short for SH3 and multiple ankyrin repeat domains protein, also referred to as ProSAPs) are scaffold proteins in the postsynaptic density of mammalian excitatory synapses that interact with numerous synaptic proteins (Monteiro and Feng, 2017). Recent reports have made a strong case for NMDA receptor hypofunction and/or defective long-term potentiation as a cause for disease-relevant behavioural phenotypes in Shank2 and Shank3 mutant mice (Won et al., 2012;Lee et al., 2015;Peter et al., 2016) and Shank3 (Bozdagi et al., 2010;Peça et al., 2011;Kouser et al., 2013) and it has even been suggested that pharmacological upregulation of LTP can be used to reverse the phenotype (Won et al., 2012;Bozdagi, Tavassoli and Buxbaum, 2013). In sharp contrast to all other studies on the function of SHANK proteins we observe an enhancement of synaptic plasticity in Shank2 -/mice (Schmeisser et al., 2012). Here, we report robustly increased LTP in vitro and also following in vivo induction of LTP in Shank2 -/mice. Moreover, we find that Shank2 -/mice suffer from deficient synaptic maturation, suggesting a mechanistic explanation for their increased LTP capacity. This calls for further investigations into how autism-associated phenotypes reflect alterations in synaptic maturation and stability before we should attempt to pharmacologically increase LTP in autism therapies.

Results and Discussion
Motivated by methodological differences between our previous findings (Schmeisser et al., 2012) and a parallel study (Won et al., 2012), that reported increased (Schmeisser et al., 2012) versus decreased LTP (Won et al., 2012), respectively, in genetically similar Shank2 -/mice, we first investigated whether differences in slice storage, recording temperature, and animal age might account for the discrepancy in the expression of LTP. To this end, we reproduced a number of conditions from Won et al. (2012), i.e. slices were stored in an ACSF/oxygenated air (Haas-type) interface chamber instead of submerged in ACSF, animal age was 8-9 instead of 3-4 weeks, and recordings were performed at elevated instead of room temperature.
Hippocampal slices from Shank2 -/mice were then subjected to the common LTP induction protocol (see Methods for details). In our hands, also in these conditions, Shank2 -/mice showed markedly higher LTP than wild-type controls ( Figure 1A). Slice storage seemed to affect different electrophysiological phenotypes to various degrees, however. While storing slices submerged in ACSF versus a Haas-type interface chamber had no effect on the expression of LTP ( Figure 1B) or basal synaptic transmission ( Figure 1C), it did affect the observed AMPA/NMDA receptor ratios ( Figure 1D).
In light of these results and reports on substantial changes in synaptic spine morphology following brain slice preparation (Kirov, Sorra and Harris, 1999), we reasoned that an in vitro examination of Shank2 -/phenotypes might be problematic, given that SHANK2 has a well-described role in synaptogenesis and the regulation of structural dynamics in dendritic spines (MacGillavry et al., 2016). In particular we wondered whether the decreased basal synaptic transmission and increased LTP we consistently observed in Shank2 -/mice might be secondary to slicing-induced synaptic remodelling. This motivated us to examine synaptic transmission and LTP in vivo. Most rodent in vivo LTP studies have been performed in rats rather than mice, however, and not all protocols that generate LTP in brain slices consistently generate lasting synaptic plasticity in vivo, possibly due to differences in neural dynamics and neuromodulatory tone in the intact brain.
Using recently established experimental procedures (Kemp and Manahan-Vaughan, 2008) (see Methods for details), we compared Shank2 -/and wild-type mice with regard to their capacity to express electrically induced LTP in vivo. We tested different induction protocols, eliciting both short-and long-lasting forms of LTP (Kemp and Manahan-Vaughan, 2008). Both forms of LTP could be elicited in Shank2 -/mice and wildtype controls (Figure 2A, B), validating that synapses without SHANK2 can express LTP, as our data from acute slices suggest. For long-lasting LTP, the potentiation in Shank2 -/mice significantly exceeded that of wild-type controls ( Figure 2A). Further corroborating our in vitro results, basal synaptic transmission was significantly decreased in Shank2 -/mice versus wildtype controls when assessed in awake, behaving animals ( Figure 2C). In summary, we observe enhanced LTP and decreased basal synaptic transmission in Shank2 -/mice under a range of different conditions both in vitro and in vivo.
How can we consolidate the discrepancy between reduced synaptic transmission and increased LTP? Is there a mechanistic explanation linking these two findings? In Shank3 -/mice, reduced LTP has been associated with NMDA as well as AMPA receptor hypofunction Kouser et al., 2013). Both mechanisms seem plausible, since hippocampal CA1 LTP is dependent on both NMDA and AMPA receptors in its induction and expression, respectively (Nicoll, 2017). Shank2 -/mice show reduced AMPA receptor dependent basal synaptic transmission ( Figure 1C), and, in certain conditions, a reduction in AMPA/NMDA receptor ratios ( Figure 1D). We thus set out to test whether synapses lacking SHANK2 might possess fewer AMPA receptors or be functionally silent (i.e. lack them altogether), which would render these synapses salient substrates for the expression of LTP. Indeed, minimal stimulation revealed markedly higher failure rates in Shank2 -/mice than wild-type controls at hyperpolarized potentials (-60 mV, when only AMPA receptor containing synapses contribute to EPSCs) compared to depolarized potentials (+40 mV, when synapses lacking AMPA but possessing NMDA receptors can also pass currents) ( Figure 3A). Given that the apparent synaptic potency was not different between genotypes ( Figure 3B), this suggests that the reduced AMPA/NMDA receptor ratios ( Figure 3C) are a direct consequence of an increased fraction of silent synapses in Shank2 -/mice ( Figure 3D). At the same time, these silent synapses might provide the structural framework that facilitates the expression of LTP in Shank2 -/mice.
A similar excess of silent synapses has been described in mice lacking Sapap3 (Wan, Feng and Calakos, 2011), a GKAP family protein that directly interacts with SHANKs (Boeckers et al., 1999;Naisbitt et al., 1999;Yao et al., 1999) and whose absence in mice causes obsessive-compulsive behavioural traits (Welch et al., 2007) -which are also typical in autistic individuals. Likewise, FMR1 -/mice, which model the autism-related Fragile X syndrome, show altered plasticity and synapse maturation in the barrel cortex (Harlow et al., 2010).
In another mouse model for intellectual disability (SYNGAP1 haploinsufficiency) hippocampal synapses are unsilenced prematurely, adversely impacting learning and memory in the adult animal (Rumbaugh et al., 2006;Clement et al., 2012). These and the present study draw a picture of synapse maturation as a tightly controlled process, the dysregulation of which seems of relevance for a range of neurodevelopmental disorders; this process must therefore be investigated further in future studies, in particular with regard to how it can be influenced by therapeutic approaches for the potential treatment of autism and related disorders.
In summary, we consistently observe increased LTP in hippocampal Schaffer collateral-CA1 synapses of Shank2 -/mice in vitro, as well as in awake behaving animals. We have uncovered a developmental synapse phenotype that could link the phenomena of decreased synaptic transmission and increased LTP in Shank2 -/mice. These data add weight -and mechanistic insight -to our initial conclusion that SHANK2 loss of function results in enhanced synaptic plasticity, highlighting the necessity to caution against generalizing therapeutic approaches that rely on boosting LTP as a strategy to treat ASD phenotypes.

Shank2
-/mice were bred in a C57BL/6J background with a heterozygous breeding protocol. The study was Westfalen, respectively). Wild-type littermates were used as a control throughout and experimenters were blind to the genotype of the tested animals for data collection and analysis.
Hippocampal brain slices were prepared from animals of both sexes as described (Schmeisser* et al., 2012).
Briefly, mice were anesthetized with isoflurane and decapitated.  Immediately after preparation, slices were transferred to in an ACSF/oxygenated air interface chamber and allowed there to recover until recording, for at least one and at most five hours. Recordings were performed in a submerged recording chamber, storage and recording temperature for these experiments was 34°C.
Mouse age for in vitro experiments was 8-9 weeks for experiments shown in Figure 1A  For minimal stimulation, stimulation frequency was 0.2 Hz and the stimulation electrode was placed to produce a single-peak response. At +40 mV holding potential, stimulation intensity was reduced until transmission failures were observed (in ~ 10-40% of events), and 20-50 events were recorded. Cells were subsequently clamped to -60 mV holding potential, and 30-50 events were recorded at the same stimulation intensity. In a subset of experiments, this order was reversed, i.e. stimulation intensity was adjusted and miniature EPSCs (minEPSCs) were recorded at -60 mV first, before cells were clamped to +40 mV. We did not observe systematic differences in failure rates or amplitudes of minEPSCs between the two regimes.
Experiments with linearly increasing or decreasing failure rates and/or minEPSC amplitudes at any holding potential were excluded from the analysis. Post-hoc analysis counted a failure at depolarized potentials whenever the EPSC charge 10 to 90 ms after stimulation did not exceed a fixed threshold of 0.15 pC.
Failures at hyperpolarized potentials were defined as events with a minEPSC peak smaller than twice the signal noise of that recording (signal noise: standard deviation of the signal in a 3 ms time window averaged over all sweeps at a certain holding potential). Signal noise was not different between experimental groups, experiments with high background noise were excluded from the analysis. While the absolute failure rates depended on the criteria for failure vs successes were set, the relative difference between genotypes did not.
Apparent synaptic potency was calculated as the EPSC amplitude of all events categorized as non-failures Analyses were performed using custom written procedures in IGOR Pro and MATLAB. Data in graphs and in the text are presented as mean ± standard error unless indicated otherwise. Unpaired two-tailed Student's ttest (short: Student's t-test) and multiple way ANOVA were used to test for statistical significance. Results were considered significant at p < 0.05. Stimulus artifacts were blanked or cropped in sample traces. Sample sizes are given as number of experiments (n, applies to in vitro experiments only) and number of animals (N).

Figure 3. Minimal stimulation reveals insufficiently matured synapses in Shank2 -/mice
A. EPSCs were recorded at different holding potentials under minimal stimulation in vitro. Failure rates indicate the proportion of AMPA receptor containing synapses (-60 mV) as a fraction of total synapses (+40 mV). Failure rates at different holding potentials are plotted for control (left) and Shank2 -/mice (right).