Syndapin I Loss-of-Function in Mice Leads to Schizophrenia-Like Symptoms

Abstract

Schizophrenia is associated with cognitive and behavioral dysfunctions thought to reflect imbalances in neurotransmission systems. Recent screenings suggested that lack of (functional) syndapin I (PACSIN1) may be linked to schizophrenia. We therefore studied syndapin I KO mice to address the suggested causal relationship to schizophrenia and to analyze associated molecular, cellular, and neurophysiological defects. Syndapin I knockout (KO) mice developed schizophrenia-related behaviors, such as hyperactivity, reduced anxiety, reduced response to social novelty, and an exaggerated novel object response and exhibited defects in dendritic arborization in the cortex. Neuromorphogenic deficits were also observed for a schizophrenia-associated syndapin I mutant in cultured neurons and coincided with a lack of syndapin I–mediated membrane recruitment of cytoskeletal effectors. Syndapin I KO furthermore caused glutamatergic hypofunctions. Syndapin I regulated both AMPAR and NMDAR availabilities at synapses during basal synaptic activity and during synaptic plasticity—particularly striking were a complete lack of long-term potentiation and defects in long-term depression in syndapin I KO mice. These synaptic plasticity defects coincided with alterations of postsynaptic actin dynamics, synaptic GluA1 clustering, and GluA1 mobility. Both GluA1 and GluA2 were not appropriately internalized. Summarized, syndapin I KO led to schizophrenia-like behavior, and our analyses uncovered associated molecular and cellular mechanisms.

defective synaptic GluA1 and GluA2 clustering, exaggerated novel object response, glutamatergic hypofunction, long-term potentiation (LTP), long-term depression (LTD), reduced anxiety-like behavior and social interactions, schizophrenia, syndapin I (PACSIN1) knockout mice

Introduction

Schizophrenia is marked by pronounced cognitive dysfunctions, which seem to be associated with imbalances in neurotransmitting and neuromodulatory systems. Although not fully understood, it is thought that these imbalances account not only for the “positive” (e.g., intrusions, excitement) and “negative” symptoms (e.g., social withdrawal, blunted affect) but also for the cognitive “disorganization” (deficient attention, impaired executive functions) observed in the disease (Bowie and Harvey 2006; Frankle et al. 2003; Koch 2007).

Syndapins (synaptic dynamin-associated proteins; also called PACSINs) belong to the BAR domain superfamily of membrane-shaping proteins. Their molecular functions include membrane binding, curvature induction and/or sensing via their F-BAR domain, as well as protein–protein interactions via NPF motifs and an SH3 domain (Kessels and Qualmann 2004; Qualmann et al. 2011). F-BAR and SH3 domain functions hereby seem to positively influence each other. At rest, syndapin dimers are thought to be in a closed, autoinhibited conformation, as supported by crystal structures of syndapin dimers showing the SH3 domains positioned onto the back of the dimeric F-BAR module (Rao et al. 2010; Wang et al. 2009). Additionally, self-association via the F-BAR domain can interconnect SH3 domain interaction partners (Kessels and Qualmann 2006; Schwintzer et al. 2011) and may be important for the formation of syndapin I–containing nanodomains at the plasma membrane (Schneider et al. 2014).

The brain-enriched syndapin I isoform has so far been implicated in neuromorphogenesis (Dharmalingam et al. 2009), in presynaptic, activity-dependent membrane trafficking and shaping of synaptic vesicles (Anggono et al. 2006; Koch et al. 2011), and in dendritic spine formation and organization (Schneider et al. 2014). In line with functions in different endocytic processes (Koch et al. 2011; Qualmann and Kelly 2000), syndapin I knockdown in cultured, dissociated neurons led to a reduced intracellular accumulation of overexpressed GluA2 when endocytosis was stimulated (Anggono et al. 2013).

Screening and correlation efforts furthermore suggested that there may be some link of syndapin I functions and schizophrenia. Schizophrenia patients displayed reduced syndapin I protein levels in the dorsolateral prefrontal cortex (Pennington et al. 2008), and a gene locus including the human SYNDAPIN I (PACSIN1) gene (6p21.3-6p22) was schizophrenia-associated (International Schizophrenia Consortium et al. 2009; Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium 2011; Stefansson et al. 2009). As such observations lack any confirmation of causality, we studied syndapin I KO mice (Koch et al. 2011) to experimentally address whether a lack of syndapin I may indeed cause schizophrenia-like phenotypes.

Furthermore, recent exome sequencing efforts of schizophrenic patients led to the identification of a rare mutation leading to an R241Q exchange in syndapin I (PACSIN1) (Genovese et al. 2016). Schizophrenia shows a complex etiology involving multiple and heterogeneous genetic factors with many individual mutations found merely in the sub-permille range (Fromer et al. 2014; Singh et al. 2016). As furthermore obvious loss-of-function mutants in the human PACSIN1 gene are almost absent from the ExAC database (gene, ENSG00000124507) and the probability of a loss-of-function intolerance in humans thus is high (pLi 0.99), the identification of the R241Q mutation may be of relevance, too. However, it would have to either represent a clear loss-of-function mutation despite the only single exchange or even cause dominant-negative effects. One may envision the latter by a combination of intact dimer formation with impairment of some other crucial molecular aspect(s) of syndapin function. As the identified schizophrenia-associated mutation (SzM) may thus help to unveil and dissect such molecular mechanisms, we thus here additionally characterized the syndapin I R241Q mutation. Interestingly, our structural, biochemical, and cell biological analyses show that syndapin I SzM is a loss-of-function mutation. Together with the further hints toward a role of syndapin I in schizophrenia outlined above, this made it even more urgent to test syndapin I KO mice for a putative involvement of syndapin I in schizophrenia.

Strikingly, syndapin I KO indeed resulted in characteristic schizophrenia-like behavior, such as less anxiety, reduced response to social novelty, and an exaggerated response to a novel object. With (i) defects in dendritic arborization in the cortex, (ii) glutamatergic hypofunction by negatively affecting the synaptic availability of both NMDA- and AMPA-type glutamate receptor subunits, (iii) impaired synaptic plasticity by disrupting both long-term potentiation (LTP) and long-term depression (LTD), and (iv) imbalances between NMDA and AMPA currents, our analyses furthermore identified molecular and cellular defects caused by syndapin I KO that thereby also provided insights into the pathophysiology associated with schizophrenia-like behavior in mice.

Material and Methods

Animals

All animal procedures were in strict compliance with the EU directives 86/609/EWG and 2007/526/EG guidelines for animal experiments and were approved by the local government (Thüringer Landesamt, Bad Langensalza, Germany, and Landesverwaltungsamt, Land Sachsen-Anhalt, Halle [Saale], Germany).

The generation of syndapin I KO mice has been described previously (Koch et al. 2011).

All experiments have been performed with syndapin I WT (+/+) and KO (−/−) mice derived from heterozygous breedings on a C57Bl/6J;129/SV (88/12) background except for behavioral analyses (n > 20 backcrossings on C57Bl/6J). For some experiments, we also analyzed heterozygous syndapin I mice.

Animals were housed under temperature- and humidity-controlled 14 h light/10 h dark conditions. Testing was conducted during the light phase, whereby the mice had at least 30 min of acclimation to the test room.

Behavioral Analyses

During the social interaction test in the home cage, a single mouse was first placed into a type II home cage and allowed to explore for 10 min. Afterward, a stranger mouse was added into the cage separated by a transparent wall, and social interactions were analyzed for an additional 10 min. Social interactions were defined as time in proximity of the stranger.

During the hanging wire test, the mouse was attached to a hanging wire with its forelimbs and/or hind limbs, and the latency to fall was measured.

In the marble burying test, the open-field (OF) box was filled with bedding and 25 marbles were placed on top. The test was performed in the dark. The mouse was allowed to bury the marbles for a defined time period of 30 min. Afterwards, the number of marbles that were at least buried by two-thirds was counted.

The following tests were performed essentially as described: open field (Bhattacharya et al. 2014), plus maze (Graeff et al. 1998), nest building (Won et al. 2012), three-chamber test (O'Tuathaigh et al. 2007), rotarod (Bhattacharya et al. 2014), and beam walking (Irintchev et al. 2005). The pup retrieval test was performed essentially as described (Liu et al. 2013) except that three pups and virgin females were used. The response to novel object test (Wiedholz et al. 2008) has been performed with slight modifications. Mice were video-tracked for 30 min each phase in the open-field box.

Morris water maze (males; 32 weeks old) was performed as described (Morris et al. 1986). The water maze apparatus consisted of an open water maze tank (diameter 100 cm) filled with opaque water. In order to escape from the water, the mice had to find a fixed location, a hidden acrylic glass platform submerged approximately 1 cm below the water surface. In order to define the platform location and start positions, the water maze was divided into four quadrants (N, E, S, W). Mice were habituated for 2 days and then trained for 5 consecutive days to find the platform. Each mouse was tested four times a day with randomized starting positions. Mice were placed in the water facing the wall and had 60 s to find the platform. A probe trial was conducted 24 h after the last training trail to determine the extent to which the mice had learned about the spatial location of the platform. Here, the platform was removed and mice were allowed to swim once for 60 s in the tank. Cohorts were tested by the same experimenter for the duration of the testing procedure.

A video monitoring system (VideoMot 2 plus fg3xcap software from TSE) was used for documentation and quantification.

Plasmids

Syndapin I RNAi and scrambled RNAi constructs were expressed in vectors coexpressing PM-mCherry as reporter, as described (Schneider et al. 2014). FLAG-syndapin I F-BAR (aa1-382) also was described before (Kessels and Qualmann 2006). Superecliptic pHlourin-tagged GluA1 (SEP-GluA1) and GFP–actin were gifts from J.M. Henley (University of Bristol, Bristol, United Kingdom) and S. Kindler (University Medical Center Hamburg-Eppendorf, Germany), respectively. PM-mCherry expression vector (a derivative of the pEGFP-F vector; Takara Bio Inc.) was a gift of M. Korte (Technische Universität Braunschweig, Braunschweig, Germany).

Rat syndapin I-GFP was cloned into pEGFP-N (Takara Bio Inc.) via EcoRI/SalI after PCR using the primers BQ2223, 5′GAGGAATTCACGCCACCATGTCTGGCCCCTAC3′, and BQ2235, 5′TTCGTCGACTATAGCCTCAACGTA3′ (restriction sites underlined).

Rat syndapin I SzM (R238Q) was generated by mutagenesis PCR using the primers BQ2623, 5′CAGTTTGAGGAGAAGCAGCTGGTCTTCCTGAAG3′, and BQ2624, 5′CTTCAGGAAGACCAGCTGCTTCTCCTCAAACTG3′ and inserted into pEGFP-N.

Both syndapin I and syndapin I SzM were subcloned into pGEX-6P (GE Healthcare) and modified pPAL7 (Bio-Rad). pPAL7 was modified within the MCS (bold nucleotides within the original pPAL7 sequence 5′AAGCTTTGACTAGTACCATGGCGGGATCC3′ were removed).

For rescue experiments, an RNAi-insensitive version of the syndapin I R238Q SzM mutant was constructed by fusing R238Q-mutated and RNAi-insensitive parts of syndapin I in pEGFP-N using the Eco91I and PauI restriction sites in syndapin I. RNAi-insensitive syndapin I was described previously (Dharmalingam et al. 2009).

GST-Dynamin I PRD (aa746-851 DynI ab) (Qualmann et al. 1999), GST-Dynamin III PRD (aa756-858 DynIII baa) (Koch et al. 2011), and GST-N-WASP PRD (aa265-391) (Kessels and Qualmann 2002; Kessels and Qualmann 2006) have been published previously. GST-Cobl PRD (aa54-390) was cloned by PCR using the primers BQ1463, 5′CCGGAATTCATGAAGGAGGCACTGC3′, and BQ1465, 5′CCGCTCGAGCACTGTGCTCTTCC3′ and inserted into pGEX-5X (GE Healthcare) via EcoR1/Xho1. All constructs generated via PCR were sequenced before use.

Fusion Proteins

GST-fusion proteins were purified as described (Qualmann et al. 1999). Tag-free versions were generated according to the manufacturer (Profinity eXact Protein Purification System, Bio-Rad).

Antibodies

For primary antibodies, please see Table 1.

Secondary antibodies used included conjugates with Alexa Fluor488, Alexa Fluor568, Alexa Fluor647, and Alexa Fluor680 (Molecular Probes), and antibodies coupled to IRDye680 (LI-COR Biosciences) and IRDye800 (LI-COR Biosciences), respectively.

Table 1

Primary antibodies used

Protein recognized	Species	Clone/name/ordering #	Dilutions used		Company, citation
			WB	IF
Bassoon	rb	141 002		1:300	Synaptic Systems
ERC1b/2	rb	143 003	1:1000		Synaptic Systems
FLAG	rb	F7425	1:500		Sigma
GFP	m	Clone JL-8	1:8000		Clontech
GFP	m	Clone B34		1:1000	BioLegend
GluA1	m	Clone RH95	1:1000	1:100	Millipore
GluA1	rb	Clone C3T		1:500	Millipore
GluA2	m	Clone 6C4		1:100	Millipore
GluA2	m	Clone L21/32	1:500	1:500	NeuroMab
GluA3	m	Clone 3B3	1:1000		Millipore
GluN2A	m	clone N327/95	1:300		NeuroMab
GluN2B	m	clone N59/36	1:500		NeuroMab
GST	rb	-	1:1000		(Qualmann et al. 1999)
Homer1	gp	160 004	1:1000	1:300	Synaptic Systems
MAP2	gp	188 004		1:500	Synaptic Systems
MAP2	rb	ab24640		1:500	Abcam
N-WASP	gp	P337	1:1000		(Kessels and Qualmann 2002)
PICK1	m	clone L20/8	1:1000		NeuroMab
ProSAP1/Shank2	rb	162 202	1:1000		Synaptic Systems
ProSAP2/Shank3	rb	162 002	1:1000		Synaptic Systems
Shank1	m	Clone N22/21	1:1000		NeuroMab
Syndapin I	gp	--	1:2500		(Braun et al. 2005)

Liposome Copelleting

Liposomes made from Folch fraction I (Sigma) were generated as described (Koch et al. 2011). For liposome copelleting with purified protein, tag-free proteins were pre-spun at 200 000 g for 7 min at 28 °C before liposome binding. Pre-cleared, tag-free proteins (0.1 mg/ml) were then incubated with liposomes (1 mg/ml) in H-buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 2.5 mM DTT) for 15 min at RT and centrifuged at 200 000 g for 20 min at 28 °C. Supernatant (unbound) and pellet (bound) were boiled in SDS sample buffer and subjected to SDS-PAGE and Coomassie staining. Visualization was done using the LI-COR Odyssey system.

Liposome copelleting assays with membrane-depleted brain lysates were performed as described previously (Koch et al. 2011).

Membrane Fractionation of HEK293 Cell Lysates

Twenty-four hours after transfection, HEK293 cells (grown in six-well plates) were harvested by scraping into ice-cold PBS and centrifuged at 100 g for 5 min at 4 °C. The cell pellets were flash frozen in liquid nitrogen, resuspended in 220 μl fractionation buffer (5 mM HEPES pH 7.4, 0.25 M sucrose, Complete EDTA-free [Roche], 1 mM EDTA), and incubated for 15 min at 4 °C. The cell suspension was then drawn into a syringe (37 Gx1 1/2) and ejected. This step was performed 10 times. An aliquot was kept for analysis (homogenate, H); the remainder was centrifuged (1000 g, 10 min, 4 °C) to yield supernatant S1 and pellet P1. S1 was again centrifuged (11 700x g, 20 min, 4 °C) to yield supernatant S2 and the crude membrane fraction pellet P2. P1 and P2 were dissolved in 150 μl and 100 μl fractionation buffers, respectively. Aliquots H, S1, P1, S2, and P2 were boiled in SDS sample buffer and subjected to Western blot analysis using the LI-COR Odyssey system.

Turbidity Assay

Membrane sculpting dynamics were measured as described (McLean et al. 1991; Wolf et al. 2019). Equal amounts of GST-fusion proteins (0.18 mg/ml) or GST alone (0.09 mg/ml) were incubated with liposomes (0.4 mg/ml) made from Folch fraction I in H-buffer. Directly after mixing of proteins with liposomes, absorbance at 400 nm was measured at 30 °C every minute in a plate reader (SpectraMax M2, Molecular Devices). The absorbance A₄₀₀^t = n was derived by subtraction of “sample alone” from “sample plus liposomes” at t = n. The change in absorbance expressed as A₄₀₀ at the time of analysis in relation to A₄₀₀ at start time set to zero ([A₄₀₀^t = n/A₄₀₀^t = 0]—1) (in percent) was plotted against time, and tau values were calculated after curve fitting (one phase decay) with GraphPad Prism 5.03.

Coprecipitations

Pulldown experiments were performed as described (Schneider et al. 2014). GST-fusion proteins (45 μg of GST-DynI PRD, GST-DynIII PRD, or GST-N-WASP PRD; 50 ng of GST-Cobl PRD; 30 μg of GST [control for GST-DynI PRD, GST-DynIII PRD, GST-N-WASP PRD]; or 40 ng of GST [control for GST-Cobl PRD]) were coupled to GSH matrix and incubated with extracts from HEK293 cells overexpressing equal amounts of GFP, SdpI-GFP, and SdpI SzM-GFP, respectively. Cell extracts were prepared with 10 mM HEPES pH 7.4, 1 mM EGTA, 0.1 mM MgCl₂, 1% Triton X-100, 150 mM NaCl, and Complete EDTA-free.

Cross-Linking Experiments

For cross-link studies, extracts from HEK293 cells transfected with constructs encoding for syndapin I-GFP and syndapin I R238Q-GFP, respectively, and FLAG-tagged syndapin I F-BAR were incubated with increasing amounts of the cross-linker 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC; Sigma) (20 min at room temperature). The cross-linking reaction was stopped by the addition of 4xSDS-PAGE sample buffer and boiling. The lysates were then analyzed by SDS-PAGE and immunoblotting using 5–8% polyacrylamide gels.

Dendrite Analysis of Rat Hippocampal Neurons

Primary rat hippocampal neuronal cultures were prepared, maintained, and transfected as described previously (Schwintzer et al. 2011) at DIV 4.

For overexpression studies with syndapin-GFP and the SzM mutant thereof, neurons were fixed in 4% (w/v) paraformaldehyde (PFA) at RT for 4 min at DIV 5 and immunostained for MAP2.

For RNAi studies of primary neurons and corresponding rescue experiments, neurons were transfected at DIV 4 and fixed 48 h later at DIV 6. Transfected cells were recognized by mCherry and GFP reporter expression and recorded in systematic sweeps across coverslips.

Morphometric measurements were based on anti-MAP2 immunolabeling using Imaris 8.4 software. For each neuron, filaments to reconstruct the dendritic tree were manually drawn by Imaris filament mode. The origin was placed in the soma-center. Minimal length for primary dendrites was set to 10 μm. For Sholl analyses (Sholl 1953), the number of dendritic branching points, the number of terminal points, and the total dendritic tree length (filament length sum) were software-determined. For Sholl analyses, the cell perimeter was defined as a ring with a radius of 5 μm from the central origin. The effects of syndapin I knockdown by RNAi as well as rescue experiments using RNAi-insensitive syndapin I (SdpI^*-GFP) or SdpI^* SzM-GFP, respectively, were analyzed in a fully blinded manner using the open-source software Ant Renamer for coding of the images.

Dendrite Analysis of Golgi-Stained Cortical Pyramidal Cells

Sagittal brain slices were Golgi-stained as described (Schneider et al. 2014). Z-stacks of cortical pyramidal cells (layer II/III and layer V) were recorded with a 20× objective and 3D reconstructed using Imaris 8.0 software. Images were baseline and background subtracted. The filament tracer mode with the shortest distance algorithm (diameter: 2–16 μm) with gaps of maximal 1 μm was used to reconstruct the dendritic tree. For quantification, Sholl analysis (Sholl 1953), number of terminal points, number of branch points, and filament length sum were used.

Immunocytochemistry and Antibody-Feeding Assays Using Mouse Primary Hippocampal Neurons

Dissociated mouse hippocampal neurons were prepared essentially as previously described (Schneider et al. 2014) except that postnatal day 0–1 hippocampi were used.

For surface-labeling hippocampal neurons (DIV) 16–20 were incubated with anti-GluA1 (Millipore, Clone RH95) or anti-GluA2 (Millipore, Clone 6C4) antibodies (10 μg/ml each) for 20 min at 37 °C in conditioned medium. Neurons were fixed for 5 min with 4% PFA and stained with secondary antibodies in block solution (5% [w/v] BSA in PBS pH 7.4 supplemented with 10% [v/v] horse serum) to label the surface pool of receptors. After 15 min fixation in PFA, cells were permeabilized with 0.2% (v/v) Triton X-100 in block solution and stained with further primary and secondary antibodies to visualize the total receptor pool.

For antibody-feeding assays, living hippocampal neurons were preincubated with 2 μM TTX (final concentration) in conditioned medium for 1 h at 37 °C, and surface-localized receptors were labeled with anti-GluA1 (mouse RH95, Millipore). After washing off excessive antibodies, neurons were treated with different inducers of membrane trafficking at 37 °C (50 μM NMDA [final] for 3 min or 100 μM AMPA [final] for 3 min) followed by chase (times as indicated) in conditioned medium containing 2 μM TTX. Neurons were fixed for 5 min with 4% PFA and stained with anti-mouse Alexa Fluor488 to label the surface pool of receptors. After 15 min fixation in PFA, cells were permeabilized with methanol and stained with anti-mouse Alexa Fluor568 to label those receptors that had been internalized. Sum intensities were quantified by using ImageJ (NIH).

AMPAR (AMPA Receptor) Cluster Analyses

Images were recorded as Z-series with 0.25–0.3 μm-intervals using a Zeiss AxioObserver equipped with an ApoTome, processed equally, and quantitatively analyzed by Imaris 7.6 software (Bitplane) using the “colocalization,” “spot” (generation of spheres), “surface” (generation of clusters), and “filament” tool functions.

Analyses of cultured neurons were performed with ≥2 independent neuronal preparations. GluA1 cluster densities were determined in colabelings with MAP2 and Bassoon and Homer1, respectively. Two to three dendritic regions within one image (100–250 μm dendritic length in total) were selected and analyzed with equal settings. Sum intensity and size of each GluA1 cluster were calculated from costainings with Homer1 and Bassoon. For determining the membrane-localized diffuse pool of GluA1, sum intensities of GluA1 cluster were subtracted from overall surface GluA1 sum intensities. Similar analyses were performed with anti-GluA2.

FRAP Analysis of SEP-GluA1 and GFP–Actin

Dissociated rat hippocampal neurons were prepared as previously described (Schneider et al. 2014) and transfected using Lipofectamine 2000 (Life Technologies) at DIV 12–14. FRAP was performed 3–4 days later using procedures previously described (Makino and Malinow 2009; Rocca et al. 2013). Briefly, FRAP experiments were performed in osmo-corrected medium (20 mM HEPES pH 7.4, 0.8 mM MgCl₂, 1.8 mM CaCl₂, 5 mM KCl, 140 mM NaCl, 5 mM glucose) at 37 °C using a Zeiss AxioObserver Spinning Disc microscope equipped with a scanning FRAP unit (Rapp OptoElectronic). To bleach locally, we used a 473 nm diode laser with 50% intensity for 5–8 fast pulses (≤ 2 s, in total). Recovery curves were corrected against background noise, and signal intensities were normalized to neighboring, unbleached spine regions (Makino and Malinow 2009).

Transferrin (TF) Receptor Trafficking

For transferrin (TF) uptake, hippocampal neurons were starved for 5 min in serum-free Neurobasal medium and incubated with TF-Alexa Fluor647 (Life Technologies) (50 μg/ml in serum-free Neurobasal medium) for 5 min and 15 min, respectively, at 37 °C. After washing with Neurobasal medium, neurons were incubated with holo-TF (500 μg/ml) in conditioned medium for 30 min at 4 °C to remove surface-bound TF-Alexa Fluor647 followed by 5 min 4% PFA fixation and imaging.

For TF recycling, neurons were incubated with TF-Alexa Fluor647 (50 μg/ml in serum-free Neurobasal medium) for 1 h at 37 °C and then washed with Neurobasal medium. For controls (labeled as “0 min”), neurons were then incubated with holo-TF for 30 min at 4 °C to remove surface-bound TF-Alexa Fluor647 and then fixed with 4% PFA. To test for TF recycling (labeled as “25 min”), neurons with internalized TF-Alexa Fluor647 were incubated with holo-TF (500 μg/ml) for 25 min at RT to allow for TF-Alexa Fluor647 recycling and then fixed with 4% PFA. Remaining intracellular TF-Alexa Fluor647 was quantified in syndapin I KO and compared with syndapin I WT.

For quantification, sum intensities of TF-Alexa Fluor647 were measured in Alexa Fluor488–conjugated phalloidin (Molecular Probes)-stained neurons using ImageJ.

Electrophysiology

Transverse hippocampal slices (400 μm thick; from adult males) were prepared using a vibratome (Microm HM 650 V, Thermo Scientific) and thereafter placed for about 90 min in an incubation chamber containing artificial cerebrospinal fluid (aCSF) (124 mM NaCl, 4.9 mM KCl, 1.2 mM NaH₂PO₄, 25.6 mM NaHCO₃, 2 mM CaCl₂, 2 mM MgSO₄, 10 mM glucose, saturated with 95% O₂ and 5% CO₂, pH 7.3–7.4), continuously perfused with 95% O₂/5% CO₂ at RT (22–24 °C).

Patch-clamp whole-cell measurements of miniature excitatory postsynaptic currents (mEPSCs) from CA1 pyramidal neurons were performed by using a MultiClamp 700B patch-clamp amplifier and pCLAMP10 software (Axon Instruments). mEPSCs were recorded with glass microelectrodes filled with a solution containing 135 mM K-gluconate, 5 mM MgCl₂, 10 mM K-HEPES, and 20 mM glucose (pH 7.25, 286 mOsm, pipette resistance 3–6 MΩ) at a holding potential V_h = −60 mV. The bath medium (aCSF) was supplemented with 100 μM picrotoxin and 1 μM TTX.

To determine the NMDA–AMPA ratio, evoked excitatory postsynaptic currents (eEPSCs) were activated with 0.1 ms biphasic square electric pulses delivered to Schaffer collaterals through the concentric Pt–Ir electrode. Patch microelectrodes were filled with a solution containing 135 mM CsMeSO₄, 4 mM NaCl, 4 mM Mg-ATP, 0.3 mM Na-GTP, 0.5 mM EGTA, and 10 mM K-HEPES (pH 7.24; 281 mOsm, pipette resistance 3–5 MΩ). AMPA-mediated EPSCs were measured at V_h = −70 mV in the presence of 100 μM picrotoxin, NMDA-mediated EPSCs at V_h = +40 mV in the presence of 20 μM CNQX, and 100 μM picrotoxin.

All data were low-pass filtered at 2 kHz and acquired at 10 kHz with a Digidata 1440A controlled by the pCLAMP10 software. Off-line analysis of mEPSCs was performed using MiniAnalysis software (v.6.0.7, Synaptosoft).

Extracellular long-term recordings in the CA1 region were performed as reported previously (Denayer et al. 2008). After 90 min incubation, one slice was arbitrarily selected, and a tungsten electrode was placed in the stratum radiatum of area CA1. For recording of field excitatory postsynaptic potentials (fEPSPs), a glass electrode (filled with aCSF, 3–7 MΩ) was lowered into the stratum radiatum about 200 μm from the stimulation electrode. The time course of the field EPSP was measured as the descending slope function for all sets of experiments. After input–output curves had been established, the stimulation strength was adjusted to elicit a fEPSP slope of 35% of maximum and was then kept constant throughout the experiment.

Conditioning stimuli for LTP and LTD were as follows. LTP was induced by a single-theta burst stimulation (TBS) consisting of 10 pulses of 4 stimuli at 100 Hz, separated by 200 ms, 0.2 ms pulse width. NMDAR-dependent LTD was evoked by bath application of 30 μM NMDA dissolved in aCSF for 4 min (Martin et al. 2014; Pasciuto et al. 2015).

Immediately after induction of LTP and LTD, respectively, evoked responses were monitored at 1, 4, 7, and 10 min and then subsequently recorded every 5 min up to 3 h. The recording of slices from mutant mice was interleaved by experiments with wild-type controls.

Data are presented as mean ± SEM unless otherwise indicated. Statistical significance was determined using IBM SPSS 19 (Armonk, NY, United States) and GraphPad Prism 5.01.

Biochemical Analyses of Total Cortex Homogenates, Synaptic Membranes, or PSDs (Postsynaptic Densities)

Mouse postsynaptic densities were prepared as described with slight modifications (Wyneken et al. 2001). All fractionation steps were carried out on ice or at 4 °C unless otherwise stated. Briefly, per preparation either cortices of three adult mice (for PSD preparation) of the same genotype were pooled, or the cortex of one adult mouse (for preparation of total homogenate and synaptic membranes) was used. The tissue was immediately homogenized in 10 ml/g tissue weight ice-cold buffer A (320 mM sucrose, 1 mM EDTA, 5 mM HEPES, pH 7.4, containing Complete EDTA-free) using a Potter S homogenizer (Sartorius). An aliquot of the total homogenate was kept for Western blot analysis. Cell debris and nuclei were removed by centrifugation at 1000 g. The supernatant S1 was centrifuged at 12 000 g for 20 min to yield supernatant S2 and pellet P2 (crude membrane fraction). P2 was further fractionated by centrifugation via a sucrose step gradient, essentially as described (Wyneken et al. 2001) except that 1 mM EDTA was added to the sucrose solutions. The synaptosomal fraction of the first gradient was lysed for 30 min in five volumes of a hypo-osmotic buffer B (1 mM Tris-HCl, pH 8.1; 1 mM EDTA) and centrifuged at 33 000 g for 30 min. The resulting pellet was resuspended and fractionated by a second centrifugation in a sucrose step gradient containing 1 mM EDTA to yield the interphase consisting of synaptic membranes.

To prepare PSD fractions, the interphase consisting of synaptic membranes was diluted in buffer C (320 mM sucrose, 1 mM EDTA, 5 mM Tris-HCl, pH 8.1) and was then delipidated by an equal volume of buffer D (320 mM sucrose, 1 mM EDTA, 1% [v/v] Triton X-100, 12 mM Tris-HCl, pH 8.1). The suspension was kept on ice for 15 min and centrifuged for 30 min at 33000 g. The pellet was resuspended in equal volumes of buffer C and D (60 ml of each buffer per 10 g tissue weight), kept on ice for 15 min and finally centrifuged for 30 min at 33 000 g. The PSD pellet was resuspended in buffer B to a final protein concentration of 1 mg/ml. Equal protein amounts (10 μg) were analyzed by semiquantitative immunofluorescent Western blotting using a LI-COR Odyssey system. All samples of one particular fraction were transferred onto the same blot membrane and normalized to a loading control (ERC1b/2 for total cortex homogenate and synaptic membranes, Homer1 for PSD). Mean protein level in KO was expressed as percent of mean WT level.

Quantification and Statistical Analysis

No statistical methods were used to predetermine sample size. All quantitative data shown represent the mean and SEM.

Statistical analyses were done using GraphPad Prism software (GraphPad Software, Inc.; versions 5.01, 5.03, and 6, respectively) using the tests specified in the figure legends. Comparisons of two conditions were tested by either D’Agostino normality test/Mann–Whitney U test or by unpaired t-test (normal data distribution). Multiple comparisons were tested by a two-way ANOVA with Bonferroni’s post-test (Sholl analyses) or two-way repeated measures (RM) ANOVA with Sidak’s post-test or were tested by Kruskal–Wallis with Dunn’s post-test and one-way ANOVA with Tukey’s post-test (normal data distribution), respectively, as reported in the figure legends.

Statistical significances were marked by ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001 throughout.

Results

Syndapin I KO Shows Schizophrenia-Like Behavior

Reduced syndapin I (PACSIN1) protein levels were one out of only five protein reductions in proteomic studies of the dorsolateral prefrontal cortex of schizophrenia patients (Pennington et al. 2008), and a gene locus including the human PACSIN1 gene was found to be associated with schizophrenia (International Schizophrenia Consortium et al. 2009; Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium 2011; Stefansson et al. 2009). We therefore analyzed syndapin I KO mice to address whether syndapin I dysfunction may indeed be involved in schizophrenia. Schizophrenia is marked by hyperactivity, stereotypic movements, and reduced anxiety-like behavior (Bowie and Harvey 2006; Frankle et al. 2003; Koch 2007). Syndapin I KO mice clearly were hyperactive, as their locomotor activity in the open-field (OF) test was increased, as both the distance traveled and the speed of movements were increased (Fig. 1A,B).

Figure 1

Syndapin I KO mice show schizophrenia-like phenotypes. (A,B) Syndapin I (SdpI) KO mice showed increased locomotor activity in the open field, as determined by distances traveled (A) and speed (B). (C) Reduced anxiety-like behavior in an elevated plus maze, as indicated by increased time on open arm. (D) Decreased nest building activity of syndapin I KO mice. (E) Increased latency of syndapin I KO mice to retrieve pups (time of observation, 600 s). (F) Social interactions in the home cage (unaffected). (G,H) Three-chamber test. Syndapin I KO mice behaved as WT in phase 2 (G) but lacked the increased exploration time toward stranger number two in phase 3 (H) suggesting a decrease in social memory and defective social novelty responses. (I–L) Novel object response test. (I,J) Tracks of WT and syndapin I KO mice during an extended open-field test (phase 1) (I) and after presentation of the novel object in the center (area marked by 1) (phase 2) (J). (K,L) Quantifications of time in proximity to the novel object (area marked by 2) (time-resolved, K; averaged time in proximity during phase 2, L). Syndapin I KO mice show hyperactivity, reduced anxiety-like behavior, and an exaggerated response to novelty. (A–L) n = 10/genotype, except for E, n = 8 (KO). Data, mean + SEM or ± SEM. unpaired t-test (A,B,C,F,L) and two-way repeated measures (RM) ANOVA + Sidak’s (D,E,G,H), respectively. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. (A–H,L) Statistically significant differences are reported between KO and WT tested under similar conditions. (G,H) Statistically significant significances are also reported between the two conditions for WT and KO, respectively. See also Fig. S1.

This hyperactivity may also underlie the superior performance of syndapin I KO mice observable in the rotarod tests (Fig. S1A). Although basic motor functions of syndapin I KO mice, such as motor coordination and balance as well as muscle strength, were normal (Fig. S1B,C), their latency to fall off in constant-speed rotarod tests was strongly increased (Fig. S1A).

Stereotypic behavior of syndapin I mice was also increased, as demonstrated by the marble burying test. Syndapin I KO mice buried about 75% of all marbles and thus highly significantly more than WT mice (about 45%) (Fig. S1D).

Reduced anxiety-like behavior can be tested by the elevated plus maze. Syndapin I KO mice spent significantly more time on the open arms, which WT mice normally carefully avoid. With more than 100% increase, the effect was drastic and argued for a strongly reduced anxiety of syndapin I KO mice (Fig. 1C).

Reduced social interactions—even with family members—also are typical of schizophrenia (Bowie and Harvey 2006; Frankle et al. 2003; Koch 2007). Syndapin I KO mice showed obvious deficits in nest building (Fig. 1D). Syndapin I KO female mice also were generally more distant and less responsive during mother–infant play, similar to humans suffering from schizophrenia (Davidsen et al. 2015). Quantitatively, this emotional distance and lack of infant response are clearly demonstrated by pup retrieval tests. Although left-alone pups represent a very strong social stimulus, virgin syndapin I KO females did not retrieve pups at all. When latencies to retrieve pups were measured, syndapin I KO mice usually exceeded the complete time of observation (600 s) (Fig. 1E).

To extend our analysis to social behavior toward interactions between adult mice, we subjected mice to home cage social interaction tasks and three-chamber tests. No differences between genotypes were found in their preference for a stranger mouse: neither in the home cage social interaction task (Fig. 1F) nor in phase 2 of the three-chamber test (Fig. 1G). Importantly, however, in phase 3 of the three-chamber test, during which a new stranger mouse was added, syndapin I KO mice showed no preference for the new stranger. Instead, syndapin I KO mice explored the new stranger with the same intensity as the already more familiar one. Syndapin I KO mice therefore obviously do not respond properly to social novelty (Fig. 1H).

Novelty responses can also be assessed by confrontation with novel objects instead of conspecifics. Such tests also are very informative in the contexts of schizophrenia assessments, as increased locomotor activity in response to novel objects was demonstrated to be typical for schizophrenic mouse models (Wiedholz et al. 2008). In addition to again showing general hyperactivity (Fig. 1I; also see Figs 1A and S1E for quantitation of distance traveled), syndapin I KO mice indeed showed an excessive response to the novel object placed in the center. Whereas WT mice briefly explored the new object, syndapin I KO mice obsessively visited and explored the novel object (Fig. 1J,K).

In order to take into account the hyperactivity of syndapin I KO mice, which may increase the number of visits and the distances traveled in object proximity by about 20% (compare Fig. 1A), we determined the relative time that syndapin I KO mice spend in proximity of the object in order to quantitatively assess this severe behavioral phenotype in more detail. Whereas WT mice only devoted less than 5% of their time for novel object interactions and the individual interactions usually were short and of low intensity, syndapin I KO mice in total spent almost 4-fold as much time in close proximity to the novel object, and the mice even jumped or climbed onto the object (Fig. 1I–L).

Considering the hyperactivity, the reduced anxiety-like behavior, the altered social interactions, and the exaggerated responses to novel objects, we concluded that syndapin I KO mice displayed schizophrenia-like behaviors.

Syndapin I SzM R241Q Is a Loss-of-Function Mutation

In addition to the reduced syndapin I (PACSIN1) protein levels found in schizophrenia patients (Pennington et al. 2008) and the association of a human PACSIN1-containing gene locus with schizophrenia (Stefansson et al. 2009), recent exome sequencing efforts of schizophrenic patients identified a R241Q exchange mutation in human syndapin I as disease-linked (Genovese et al. 2016). Yet, it remained to be addressed whether such a mutant would be dysfunctional and which molecular and cellular functions of syndapin I may be affected by the identified disease-linked mutation. We therefore in addition to syndapin I KO mice also analyzed the syndapin I SzM at the structural, biochemical, and cellular level.

Arginine 241 shows high-sequence conservation. It is located in the periphery of the central six-helix bundle of the F-BAR module (Wang et al. 2009) and is conserved in all species we analyzed (Fig. S2A). The R241Q exchange would lead to a loss of intra- and intermolecular H-bonds (Fig. 2A). This could potentially indeed impair syndapin I functionality.

Figure 2

Syndapin I schizophrenic mutant results in syndapin I loss-of-function. (A) Crystal structure of human syndapin I (pdb code, 3HAH) and in silico mutated syndapin I R241Q (SzM) using Pymol 1.8.4.0 (Schrodinger LLC). Black arrows point out indicated H-bonds in WT syndapin I, and red arrow highlights the lack of H-bondsin syndapin I R241Q (SzM). (B) Electrophoretic mobilities of syndapin I and syndapin I SzM in Coomassie-stained native gels. (C) Syndapin I and syndapin I SzM in liposome copelleting assays analyzed by Coomassie-stained SDS-PAGE gels. (D) Membrane sculpting dynamics of GST-SdpI and GST-SdpI SzM analyzed in turbidity assays (A_400nm) using liposomes made of Folch fraction I (FF). (E) Increased tau values indicating an impaired membrane sculpting activity of GST-SdpI SzM compared with GST-SdpI. n = 6. (F) Quantification of SdpI-GFP and SdpI SzM-GFP in immunoblottings of the crude membrane fraction (P2) of HEK293 cells cotransfected with mCherry-F and GFP, SdpI-GFP, and SdpI SzM-GFP, respectively, showing a reduction of SdpI SzM-GFP in P2. n = 9 each. (G–J) Comparative coprecipitation analyses of syndapin I and syndapin I SzM with the proline-rich domains (PRD) of the syndapin interaction partners dynamin I (DynI, G), dynamin III (DynIII, H), N-WASP (I), and Cobl (J) fused to GST (for full gels, see Fig. S2G–J). GST-fusion proteins were incubated with extracts of HEK293 cells overexpressing GFP, SdpI-GFP, and SdpI SzM-GFP, respectively. Binding of SdpI SzM-GFP to all PRDs was dramatically decreased compared with SdpI-GFP. n = 6–7. (K–N) Primary hippocampal neurons (DIV 4 + 1) overexpressing GFP, SdpI-GFP, and SdpI SzM-GFP, respectively, were stained for MAP2, imaged, and subjected to dendrite analysis using Imaris software (K). Analysis of the dendritic tree revealed increased Sholl intersections (L), summarized dendritic tree length (M), and number of dendritic branch points (N) for SdpI-GFP–expressing neurons, but not for SdpI SzM-GFP. n = 22–24. (O–R) Morphometric analyses of developing rat hippocampal neurons transfected at DIV 4 with syndapin I RNAi and scrambled RNAi coexpressing GFP, RNAi-resistant (^*) syndapin I-GFP, and syndapin I SzM-GFP, respectively, by Sholl analyses (O) and determinations of total length of the dendritic trees (P), of dendritic branch points (Q), and of the dendritic terminal points (R). n = 88–89. (S,T) Anti-FLAG (red) and anti-GFP (green) immunoblotting analyses of lysates of HEK293 cells coexpressing either FLAG-tagged syndapin I F-BAR domain and syndapin I-GFP (S) or FLAG-tagged syndapin I F-BAR domain and syndapin I SzM-GFP (T) incubated with the zero-length cross-linker EDC. Arrows mark bands immunopositive for both FLAG-tag and GFP. Data, mean + SEM or ± SEM. Mann–Whitney (E–J), one-way ANOVA + Tukey’s (M,N), two-way-ANOVA + Bonferroni post-tests (L,O) (note that the comparison between scrambled RNAi vs. SdpI RNAi + SdpI^*-GFP (n.s. for all Sholl intersections) has been omitted from the statistical significance table in O), Kruskal–Wallis test (P–R); ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. See also Fig. S2.

Interestingly, although the mutation resides in the sequence region that upon dimerization forms the F-BAR module, the overall integrity of the F-BAR module seemed undisturbed, as judged from syndapin I SzM’s unchanged electrophoretic mobility in native gels when compared with the WT protein (Fig. 2B). Liposome binding also seemed unaltered in vitro, as WT and SzM both were precipitated with liposomes effectively (Fig. 2C).

Interestingly, however, syndapin I SzM showed a reduced membrane sculpting activity, as revealed by increased tau values in liposome turbidity assays (Fig. 2D,E; for quantitative method validation, see Fig. S2B–D). Furthermore, subcellular fractionation of syndapin I WT and SzM expressed in HEK293 cells showed less syndapin I SzM in the crude membrane fraction P2 compared with WT (Figs 2F and S2E,F). This lack of syndapin I SzM in P2 may reflect an impaired formation of syndapin I–enriched nanodomains (Schneider et al. 2014).

Since F-BAR interactions may not only be crucial for effective membrane sculpting but were also suggested to play a role in an autoinhibition of the syndapin I SH3 domain (Goh et al. 2012; Quan et al. 2012; Rao et al. 2010; Wang et al. 2009), we tested next whether the schizophrenia-associated mutation would additionally result in altered SH3 domain interactions. Indeed, interactions with the endocytic components dynamin I (Qualmann et al. 1999) and dynamin III (Koch et al. 2011) as well as with the actin cytoskeletal effectors Cobl (Schwintzer et al. 2011) and N-WASP (neural Wiskott–Aldrich syndrome protein) (Qualmann et al. 1999) were all strongly impaired in syndapin I SzM (Figs 2G–J and S2G–J). These findings suggested that the release of the autoinhibition between the membrane-binding F-BAR domain and the SH3 domain is disrupted by the schizophrenia-associated mutation and that cell biological defects caused by syndapin I SzM may include dynamin-mediated endocytosis and F-actin–related neuromorphogenesis processes mediated by the actin nucleator Cobl (Schwintzer et al. 2011) and/or by the Arp2/3 complex and its activator N-WASP (Dharmalingam et al. 2009).

Consistent with putative defects in cytoskeletal pathways, schizophrenia-like symptoms were reported to be accompanied by reductions in dendritic tree complexity (Gu 2002; Jarskog et al. 2007; Penzes et al. 2011). Syndapin I shows strong neuromorphogenic potential when overexpressed in neurons. This requires both intact F-BAR and SH3 domain functions (Dharmalingam et al. 2009; Schwintzer et al. 2011). Whereas syndapin I significantly increased the global arborization of the dendritic tree, that is, the number of Sholl intersections, the summarized length of the dendritic arbor, and the number of branch points (Figs 2K–N and S2K), the syndapin I SzM, in contrast, showed no such neuromorphogenic potential, but the morphology of the cells remained similar to control (Figs 2K–N and S2K).

In order to more vigorously address the detrimental effects of syndapin I SzM in loss-of-function experiments, we conducted RNAi rescue experiments with developing rat hippocampal neurons. Sholl analyses of the dendritic tree showed that syndapin I loss-of-function led to moderate but statistically significant defects in dendritic tree development (Fig. 2O). More detailed analyses of specific parameters showed that the total length of the dendritic tree and both branch and terminal points were reduced (Fig. 2P–R). The syndapin I deficiency phenotypes were successfully rescued by RNAi-resistant wild-type syndapin I but not by the syndapin I SzM mutant (Fig. 2O–R).

Taken together, syndapin I SzM thus indeed clearly represents a loss-of-function mutation. Syndapin I SzM showed impaired SH3 domain interactions, defects in membrane sculpting, and an increased membrane extractability during fractionations (Fig. 2D–J), which may relate to its functional inactivity in neurons (Fig. 2K–R). Syndapins function as F-BAR–mediated dimers and higher-molecular-weight assemblies (Kessels and Qualmann 2006). We therefore next addressed whether the SzM mutation in the syndapin I F-BAR domain (i) still allows for self-assembly and (ii) can still integrate itself into WT syndapin I assemblies. Wild-type syndapin I F-BAR domain (FLAG-F-BAR), wild-type syndapin I-GFP, and also syndapin I SzM-GFP all gave rise to homomeric cross-link products of different sizes (Fig. 2S,T; green and red bands). Thus, also syndapin I SzM-GFP effectively self-assembles.

Additionally, also heteromeric assemblies containing both WT FLAG-tagged F-BAR as well as syndapin I-GFP and syndapin I SzM-GFP, respectively, were detected (Fig. 2S,T; yellow bands). The pattern of bands in the all–wild-type experiments (Figs 2S and S2L) hereby was identical to that of the experiment including syndapin I SzM (Fig. 2T and S2M).

These experiments show that SzM-mutated syndapin I is effectively integrated into the oligomeric syndapin complexes that are of critical importance for the in vivo functions of syndapins suggesting that even SzM mutation of only one allele may already cause defects that go beyond just lowering the gene dose of functional, wild-type syndapin I to half.

Syndapin I KO Results in Defects in Neuromorphogenesis and a Disrupted Membrane Targeting of the Syndapin I Interaction Partner N-WASP

In line with observations of impaired dendritic arborization in schizophrenic patients (Gu 2002; Jarskog et al. 2007; Penzes et al. 2011), Sholl analyses of Golgi-stained pyramidal cells in sections of WT and syndapin I KO brains showed a pronounced reduction in dendritic complexity of cortical pyramidal cells from syndapin I KO. These defects were consistent with a strong reduction of branch points, terminal points, and total dendritic tree length (Fig. 3A–H). Both pyramidal neurons in layer II/III and in layer V showed severe and highly statistically significant impairments (Fig. 3A–H). Interestingly, the impairment in all dendritic parameters analyzed (Fig. 3C–H) was already observed to full extent in heterozygous mice suggesting that in the brain, the full gene dose of syndapin I is required (Fig. S3A–F).

Figure 3

Syndapin I KO mice show reduced dendritic arborization of cortical pyramidal neurons in both layer II/III and layer V and a lack of plasma membrane recruitment of N-WASP. (A–H) Less expanded and less complex dendritic tree of cortical pyramidal cells in layer II/III and layer V in syndapin I KO mice, as revealed by Sholl analysis (A,B) and determinations of total dendritic tree length (C,F), dendritic branch points (D,G), and terminal points (E,H) of Golgi-stained neurons in brain slices of syndapin I KO and WT mice; layer II/III, n = 147 (WT) and 144 (KO); layer V, n = 112 (WT) and 134 (KO) cells derived from six mice/genotype. (I) Quantitative anti–N-WASP immunoblotting (please also see Fig. S3) of lipid-bound protein material from membrane-depleted lysates of WT, syndapin I KO, and heterozygous brains as well as from KO lysates with added recombinant syndapin I (KO + SdpI; rescue). n = 9 (WT, KO, KO + SdpI) and 5 (Het.), respectively. Data, mean + SEM or ± SEM. Two-way repeated measures (RM) ANOVA + Bonferroni’s (A,B), Mann–Whitney test (C–H), and one-way ANOVA + Tukey’s (I), respectively. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. See also Fig. S3.

Studies in cultured neurons had revealed that the neuromorphogenic potential of syndapin I requires membrane localization of the actin cytoskeleton components Cobl and N-WASP by syndapin I (Dharmalingam et al. 2009; Schwintzer et al. 2011). We thus asked whether the cytoskeletal mechanisms identified in cultured neurons were also valid at the animal level. Membrane association studies using liposomes and membrane-depleted brain cytosol from syndapin I WT and KO mice, respectively, demonstrated that this was the case. Whereas endogenous Cobl did not withstand the elaborate procedure of the assay (data not shown), membrane association of N-WASP was indeed strongly reduced upon syndapin I KO. Importantly, this defect was completely rescued by syndapin I addition to the syndapin I KO brain lysates (Figs 3I and S3G). Strikingly, quantitative experiments with brain lysates from heterozygous mice show impairments similar to complete KO of syndapin I suggesting that wild-type concentrations of syndapin I are required for efficient syndapin I–mediated recruitment of N-WASP to membranes (Fig. 3I).

Thus, syndapin I is not only critical for membrane targeting of the endocytic component dynamin (Koch et al. 2011) and for presynaptic functions (Koch et al. 2011), but syndapin I is also essential for membrane association of the Arp2/3 complex activator N-WASP controlling the formation of actin filaments.

Defective AMPAR Clustering and Spatial Organization in Syndapin I KO Neurons

Schizophrenia-related behavioral abnormalities were also reported in some mice with disrupted signaling components of the glutamatergic system, including the NMDAR (NMDA receptor) SU (subunit) GluN1 and the AMPAR SU GluA1, which is part of GluA1/GluA2 receptors and important for synaptic plasticity (Ballard et al. 2002; Mohn et al. 1999; Wiedholz et al. 2008). Syndapin I has been identified to form complexes with PICK1- and GluA2-containing AMPARs (Anggono et al. 2013). This would imply that syndapin I is important for GluA1/GluA2 but also for GluA2/GluA3 AMPA receptors.

Since the effects of syndapin I KO on postsynaptic neurotransmitter receptors had not been analyzed, yet, we addressed whether syndapin I KO may influence the synaptic availability of postsynaptic receptors and thereby disturb neuronal excitability. We investigated the subcellular localizations of endogenous neurotransmitter receptors in cultured hippocampal neurons (Fig. 4). This cell system is relatively homogenous compared with, for example, cortical cultures (Kaech and Banker 2006), and allows for a reliable assignment of synaptic puncta to individual dendritic segments (Haeckel et al. 2008; Schneider et al. 2014).

Figure 4

Reduced synaptic availability of GluA1 as well as reduced electrophysiological AMPAR activity is accompanied by increased AMPAR mobility upon syndapin I deficiency. (A–G) Primary hippocampal neurons from WT and syndapin I KO mice (DIV 16–20) stained for surface-localized GluA1 together with Homer1, Bassoon (A,D–G), or MAP2 (C) that were subjected to surface reconstructions using Imaris software. (A) Exemplary coimmunolabeling of surface GluA1, Homer1, Bassoon. Bars, 10 μm. (B) Illustration of analyzed clusters and spheres. Synaptic clusters, GluA1 clusters overlapping with Homer1–Bassoon-positive spheres. Non-synaptic clusters, Homer1–Bassoon-negative GluA1 clusters. Spheres were derived from Homer1–Bassoon colocalization channels with a diameter of 0.5 μm resembling PSD-like areas. (C) Quantitation of densities of GluA1 surface–reconstructed clusters in MAP2-stained dendrites. (D–F) GluA1 clusters grouped into synaptic and non-synaptic clusters.Note that there are fewer synaptic and more non-synaptic GluA1 clusters upon syndapin I KO (D). Synaptic and non-synaptic cluster sizes (E) as well as summarized intensities (F) were decreased in syndapin I KO. (G) Decreased surface-localized GluA1 signal intensities in Homer1–Bassoon-colocalizing spheres resembling PSD-like areas upon syndapin I KO. (C–G) n = 30–35 pictures/genotype. (H,I) AMPAR-mediated eEPSCs from CA1 pyramidal neurons evoked at V_h = −70 mV in the presence of 100 μM picrotoxin were reduced in syndapin I KO brain slices. Quantitative analyses (I), n = 26/27 cells (WT/KO) of 10 mice/genotype. (J,K) Slowed decay kinetics (J), as quantitatively described by increased decay time constant tau (K), of mEPSCs from syndapin I KO CA1 pyramidal neurons (patch-clamp whole-cell measurements in the presence of 100 μM picrotoxin and 1 μM TTX at V_h = −60 mV). n = 21/28 cells (WT/KO) of 9/10 (WT/KO) mice. (L) Immunofluorescence images of cultured neurons of WT and syndapin I KO mice stained for endogenous GluA1. To visualize the diffuse pool of surface-localized GluA1, GluA1 signal intensities were grouped into clustered (GluA1 cluster) and non-clustered (GluA1 diffuse) signals. Bars, 10 μm. (M) Signal intensities of the diffuse pool of surface GluA1 were increased upon syndapin I KO. (N–P) Spine regions of primary hippocampal rat neurons (DIV 12–14) cotransfected with SEP-GluA1 and scrambled RNAi or syndapin I RNAi for 3–4 days subjected to FRAP (N) and analyzed in comparison to unbleached spine regions for plateau fluorescence recovery (O) and half time of recovery (P). (N) Smoothed curves of mean GluA1 recovery upon scrambled versus syndapin I RNAi. Whereas recovery plateaus (O) were unchanged, GluA1 recovery was faster in syndapin I–deficient neurons (P). n = 16–18. (Q–S) Similar FRAP analyses of spine regions in mature neurons cotransfected with GFP–actin and scrambled RNAi versus syndapin I RNAi. (Q) Mean GFP–actin recovery (smoothed curves) upon scrambled versus syndapin I RNAi. (R) Recovery plateaus were unchanged. (S) Quantification of half times revealed faster GFP–actin recovery in syndapin I–deficient neurons. n = 10–26. Data, mean + SEM. Unpaired t-test (C,G,I,K,M,O,P,R,S) or one-way ANOVA + Tukey’s (D–F). Statistically significant differences in D–F are only reported between KO and WT conditions. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. See also Fig. S4.

We first studied the subcellular localization of GluA1, as it contributes to both basal synaptic transmission and synaptic plasticity (Zamanillo et al. 1999). We quantified GluA1-containing assemblies derived from surface-reconstructed plasma membrane–localized GluA1 immunofluorescence signals, subsequently referred to as “clusters” (achieved by same settings between WT and KO; Fig. 4A,B). While the overall GluA1 cluster density was unaffected in syndapin I KO neurons (Fig. 4C), the percentage of synaptic GluA1 clusters was significantly reduced in syndapin I KO (Fig. 4D).

Moreover, the size and the summarized GluA1 intensity in both synaptic and non-synaptic GluA1 clusters were significantly reduced (Fig. 4E,F). Intensity measurements within the reconstruction of an immunocytochemical correlate of the PSD (defined by Homer1–Bassoon colocalization; 0.5 μm diameter; Fig. 4B) revealed a significantly decreased GluA1 intensity within these spheres upon syndapin I KO (Fig. 4G).

Comparable defects were observed for clustering GluA2 into postsynaptic assemblies in syndapin I KO neurons (Fig. S4A–F). Similar to GluA1, also GluA2 cluster sizes and GluA2 intensities within clusters were reduced (Fig. S4D,E). Solely the synaptic/extrasynaptic cluster distributions and the GluA2 intensity in Homer1–Bassoon spheres were so similar in syndapin I KO when compared with WT mice that no statistical significances were obtained (Fig. S4B,C,F).

Thus, syndapin I KO mice show a substantially impaired organization of both GluA1 and GluA2, but defects in synaptic/extrasynaptic cluster distributions and the lack of receptors right at the contact site with the presynapse were only obvious for GluA1. Unfortunately, specifically GluA2/GluA3 receptors could not be studied due to a lack of suitable anti–GluA3-specific antibodies.

These results indicate a critical role of syndapin I in GluA1/GluA2 receptor distribution and organization.

AMPAR Hypofunction at Basal Synaptic Activity in Syndapin I KO Mice

In line with defective synaptic localization of AMPAR, whole-cell patch-clamp measurements of CA1 pyramidal neurons of syndapin I KO mice revealed reduced evoked AMPAR-mediated eEPSCs (Fig. 4H,I). Additionally, mEPSCs showed significantly slower decay kinetics, as evidenced by increased tau values (Fig. 4J,K). These findings reflect alterations in the elementary functioning of individual synapses.

GluA1 clustering requires a reservoir of diffusely localized GluA1-containing AMPARs. Less well-clustered GluA1 accompanying an increased pool of diffuse GluA1 would give rise to more broadly distributed GluA1 within the postsynaptic plasma membrane. This may negatively affect the open/close synchronization of individual receptors resulting in slowed open/close kinetics, that is, increased tau values, as observed in our examinations (Fig. 4K). We thus next addressed whether the diffuse GluA1 pool would be increased. Distinguishing between clustered and diffuse GluA1 immunosignals in imaging studies, we indeed found defects in GluA1 clustering, as the intensity of the surface-localized diffuse pool of GluA1 was significantly higher in syndapin I KO than in WT neurons (Fig. 4L,M). A similar impairment was detected for GluA2 (Fig. S4G).

The above scenario should relate to an increased mobility of surface-localized GluA1. Fluorescence recovery after photobleaching (FRAP) of surface-localized AMPARs in dendritic spines using superecliptic pHluorin-tagged GluA1 (SEP-GluA1) (Ashby et al. 2004) in neurons cotransfected with either syndapin I RNAi or a scrambled RNAi indeed showed such an increased mobility (Figs 4N and S4H). While the plateau levels were indistinguishable, the SEP-GluA1 recovery half time was significantly decreased upon syndapin I deficiency (Fig. 4O,P). GluA1-containing AMPARs thus show an increased rate of mobility or lateral diffusion upon syndapin I deficiency.

AMPAR mobility depends on forces and/or viscosity brought about by the underlying cytoskeleton (Choquet and Triller 2013). To investigate actin dynamics in dendritic spines, we analyzed FRAP in spines expressing GFP–actin (Figs 4Q–S and S4I). Upon syndapin I deficiency, faster recovery rates were observed (Fig. 4S). Thus, the mobility of GluA1-containing AMPARs is likely increased due to a less dense, more unstable cortical F-actin meshwork. Since the membrane recruitment of the Arp2/3 complex activator N-WASP was disrupted in syndapin I KO mice (Figs 3I and S3G), and syndapin I is an N-WASP activator (Dharmalingam et al. 2009), it seems that such a less stable actin network in syndapin I–deficient spines reflects a loss of N-WASP functionality.

Taken together it can be concluded that the impaired clustering and the disrupted spatial organization of AMPARs upon syndapin I deficiency affect synchronization of AMPAR gating and proper function of synaptic GluA1-containing AMPARs. Hence, syndapin I KO mice suffer from AMPAR hypofunction (Figs 4 and S4).

AMPAR Hypofunction at the Level of Synaptic Plasticity in Syndapin I KO Mice

The pathophysiology in schizophrenia is also associated, at least partially, with alterations in synaptic plasticity (Crabtree and Gogos 2014; Harrison and Weinberger 2005). Prominent forms of synaptic plasticity are long-term potentiation (LTP) and long-term depression (LTD). Both synaptic plasticity processes require proper localization and dynamics of postsynaptic scaffold proteins and especially of GluA1-containing AMPARs in time and space (Penn et al. 2017; Shepherd and Huganir 2007).

As no LTD and also no LTP data were available for syndapin I KO mice in the literature, we next addressed putative defects in both synaptic plasticity processes to further unveil the cellular and physiological defects associated with the behavioral defects observed for syndapin I KO mice. We first performed NMDA-induced cLTD in dissociated hippocampal cultures and used antibody-feeding assays (Rocca et al. 2008) to explore surface and internalized pools of endogenous AMPAR SUs. Interestingly, NMDA treatment almost completely failed to induce an internalization of endogenous GluA1 in syndapin I KO neurons (Fig. 5A,B). Instead the surface levels of GluA1 remained unchanged over the full period of time marked by endocytic uptake and reinsertion of receptors into the plasma membrane in WT neurons (Fig. 5A,B).

Figure 5

Reduced synaptic plasticity in syndapin I KO mice. (A,B) Internalized and surface-localized GluA1 in response to NMDA-induced cLTD analyzed in the presence of 2 μM TTX in WT and syndapin I KO hippocampal neurons (DIV 16–20) by anti-GluA1 antibody feeding. Cells were fixed without NMDA treatment (0 min) or 5, 10, 15, and 60 min after 3 min NMDA treatment and subsequently stained for surface and internalized GluA1. (A) Representative images. (B) Quantitative analysis of surface/internalized ratios revealing an impaired GluA1 internalization after NMDA-induced cLTD in syndapin I KO neurons. n = 38–52 cells/genotype and condition. (C) Significantly impaired NMDA-induced cLTD in syndapin I KO hippocampal CA1 regions. Horizontal bar, bath application of NMDA. n = 7/6 (WT/KO). (D) Theta burst-induced LTP (arrow) in the hippocampal CA1 region is significantly impaired in syndapin I KO compared with WT. n = 7/8 (WT/KO). Bars, 20 μm. Data, mean ± SEM. Two-way ANOVA + Bonferroni’s (B); two-way RM ANOVA + Fisher’s LSD (C,D) (statistically significant differences are only reported between KO and WT conditions). ^*P < 0.05; ^***P < 0.001. See also Fig. S5.

A similar block of internalization was observed for GluA2 (Fig. S5A). These findings for endogenous GluA2 in our syndapin I KO studies are in line with an earlier report describing a reduced internalization of overexpressed myc-tagged GluA2 in cultured dissociated neurons subjected to syndapin I RNAi and prolonged incubation with NMDA (Anggono et al. 2013). While these observations with overexpressed myc-GluA2 were attributed to modulation of GluA2/3 heterodimers via PICK interactions (Anggono et al. 2013), our analyses of both endogenous GluA1 and GluA2 in contrast clearly support an effect on GluA1/2 heteromers—as it would be crucial for hippocampal LTD.

Importantly, also AMPA-induced GluA1 internalization was impaired in syndapin I KO neurons (Fig. S5B), suggesting that the GluA1 internalization defects observed in syndapin I KO mice are directly linked to the observed defects in AMPAR organization, distribution, and/or dynamics (Fig. 4A–G,N; Fig. S4).

Since we did not observe defects in basal endocytosis or recycling reflected by normal trafficking of endogenous transferrin receptors in syndapin I KO neurons (Fig. S5C–E), syndapin I seemed specifically required for activity-dependent endocytosis of postsynaptic AMPARs.

These defects in activity-dependent endocytosis of postsynaptic AMPARs became electrophysiologically overt when NMDAR-induced chemical LTD was examined. fEPSP slope determinations clearly demonstrated that NMDA-induced LTD was significantly impaired in syndapin I KO (Fig. 5C). The observed critical role of syndapin I in hippocampal LTD is fully consistent with the impairments in GluA1 and GluA2 internalization upon syndapin I KO (Figs 5A,B and S5A). Our result for hippocampal LTD in syndapin I KO mice is furthermore somewhat reminiscent of a related requirement of the isoform syndapin II in cerebellar granule cells (Anggono et al. 2013).

Disturbed activity-dependent endocytosis of GluA1 and GluA2 (Figs 5A,B and S5A) would result in altered surface levels of GluA1 and GluA2. Indeed, we detected altered surface levels of both GluA1 and GluA2 in syndapin I KO neurons (Fig. S5F–H).

The amount of surface-localized AMPARs and the dynamics of AMPARs are also critical for another prominent form of synaptic plasticity, LTP. As actin cytoskeletal contributions also are critical for LTP and syndapins interface with both components of endocytic vesicle formation and cytoskeletal components, we evoked LTP at hippocampal SC–CA1 synapses by a single-theta burst stimulation. Whereas LTP was robustly induced in WT, it was completely abolished upon syndapin I KO (Fig. 5D).

Prompted by the marked deficits of syndapin I KO in both hippocampal LTP and LTD, we additionally examined hippocampus-dependent spatial learning in the Morris water maze. Conflicting data exist whether LTP is required for and/or correlates with spatial learning in the Morris water maze or whether this is not the case (Lynch 2004; Morris et al. 1986; Zamanillo et al. 1999). The fact that hippocampal LTD induction has been suggested to enhance both acquisition and retrieval of spatial reversal memory (Dong et al. 2013) may complicate correlations between synaptic plasticity on one side and learning and memory on the other side even further. We did not find any discernible Morris water maze performance deficit of syndapin I KO mice lacking both proper LTP and LTD (Fig. S5I,J).

Overall, we concluded that, in syndapin I KO mice, AMPAR hypofunction is observable at both the level of basal synaptic transmission and of synaptic plasticity. Strikingly, the synaptic plasticity defects include both impaired LTD and impaired LTP. Consistent with our findings, synaptic plasticity defects were typically found in schizophrenia (Crabtree and Gogos 2014; Harrison and Weinberger 2005).

AMPAR and NMDAR Hypofunction in Syndapin I KO Mice

To substantiate our results of AMPAR hypofunction in syndapin I KO mice with independent means, we quantitatively determined the protein composition of PSD fractions from WT and syndapin I KO cortices (Wyneken et al. 2001) by comparative Western blot analyses. Besides trends toward reduced protein levels of the syndapin I interaction partners PICK1 (Anggono et al. 2013) and ProSAP1/Shank2 (Schneider et al. 2014)—interestingly, genetic variants of both genes have been linked to schizophrenia (Chen et al. 2017; Peykov et al. 2015) and the levels of the syndapin binding partner ProSAP2/Shank3 were reduced in the anterior cingulate cortex of schizophrenia patients (Föcking et al. 2015)—syndapin I KO showed clearly decreased GluA1 levels in isolated, Triton X-100–extracted PSD fractions (Fig. 6A).

Figure 6

AMPAR and NMDAR hypofunction in syndapin I KO mice. (A) Quantitative Western blot analysis of PSD fractions (10 μg protein each) from WT and syndapin I KO cortices. Signals were normalized to anti-Homer1 immunoblotting signals on the same membrane. n = 4–6 detections/protein. (B) NMDAR-mediated eEPSCs evoked at V_h = +40 mV in the presence of 20 μM CNQX and 100 μM picrotoxin (same neurons as presented in Fig. 4H). (C) Quantitative analyses showing significantly reduced NMDAR eEPSCs upon syndapin I KO. n = 26/27 cells (WT/KO) from 10 mice/genotype. (D) Decreased ratio of NMDA/AMPA eEPSCs in syndapin I KO mice. Data, mean + SEM. Mann–Whitney U (A) or unpaired t-test (C,D) for WT versus KO; ^*P < 0.05; ^**P < 0.01; ^***P < 0.001. See also Fig. S6.

In contrast, total GluA1 levels in homogenates and GluA1 levels in general preparations of wider areas of synaptic membranes were not affected (Fig. S6). This demonstrated that specifically the levels of GluA1 well-embedded into the synaptic scaffold were affected (Fig. 6A). In contrast, GluA2 levels in PSD preparations were not affected in a significant manner (Fig. 6A).

These findings are fully consistent with our quantitative imaging data, which unveiled reduced GluA1 but not GluA2 signal intensities in Homer1–Bassoon spheres, that is, in synaptic contact areas, of immunolabeled syndapin I KO neurons (Figs 4G and S4F).

Interestingly, glutamatergic hypofunction was not restricted to AMPARs, since we also observed a reduction of the NMDAR SU GluN2A in PSD fractions of syndapin I KO. GluN2B, in contrast, was not affected (Figs 6A).

This prompted us to address potential defects in NMDAR-mediated synaptic functions in syndapin I KO mice by electrophysiological measurements. Indeed, NMDAR-mediated eEPSCs were significantly reduced (Fig. 6B,C).

Comparison of the reductions of GluN2A and GluA1 protein levels in PSD fractions (Fig. 6A) suggested an even stronger decline of NMDAR levels in syndapin I KO. Accordingly, the ratio of NMDAR- and AMPAR-mediated eEPSCs was significantly reduced in syndapin I KO mice (Fig. 6D; for AMPA eEPSC, see Fig. 4H,I). The behavioral defects observed and the AMPAR hypofunction found both at the level of basal synaptic transmission and at the level of both synaptic plasticity processes, LTD and LTP, in syndapin I KO mice are therefore additionally accompanied by an NMDAR hypofunction in these mice.

Discussion

Mouse models have been proven valuable in understanding malfunctions of molecular pathways leading to schizophrenia-like symptoms (Ballard et al. 2002; Mohn et al. 1999; Wiedholz et al. 2008). Syndapin I (PACSIN1) KO mice showed key symptoms of these published mouse models for schizophrenia, that is, increased locomotor activity, decreased social novelty response, impaired novel object response, and decreased anxiety-like behavior. Our experimental analyses of syndapin I loss-of-function phenotypes hereby are in line with the observation that a gene locus including the human PACSIN1 gene was found to be associated with schizophrenia (International Schizophrenia Consortium et al. 2009; Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium 2011; Stefansson et al. 2009) and with the finding that one of only five severely reduced proteins in proteomic studies of the dorsolateral prefrontal cortex of schizophrenia patients was syndapin I/PACSIN1 (Pennington et al. 2008).

Also the identification of the schizophrenia-associated mutation R241Q (Genovese et al. 2016) seems to argue in this direction, as our biochemical and functional analyses demonstrated that syndapin I SzM is (i) a loss-of-function mutation and can (ii) assemble with itself and with wild-type syndapins, that is, even mutation of only one allele will corrupt both the dimeric F-BAR domain of syndapins and macromolecular syndapin lattices instead of just lowering the gene dose of functional syndapin I. Arginine 241 is highly conserved and located in the periphery of the central six-helix bundle, which represents the core of the F-BAR domain (Wang et al. 2009). In line, our biochemical analyses clearly showed that syndapin I SzM lacked membrane sculpting activity and showed a reduced membrane association in cells. Both are functions of the syndapin I F-BAR domain (Dharmalingam et al. 2009).

F-BAR domain functions are essential for the neuromorphogenic potential of syndapin I (Dharmalingam et al. 2009; Schneider et al. 2014; Schwintzer et al. 2011). Consistently, syndapin I SzM completely lacked the ability to promote dendritic arborization, and syndapin I SzM furthermore also completely failed to rescue the phenotypes of syndapin I deficiency in dendritic arborization.

In addition to the schizophrenia-like behavioral phenotypes observed upon syndapin I KO, syndapin I KO mice also develop epileptic seizures (Koch et al. 2011). It may be noteworthy that the risk of schizophrenia in patients with epilepsy is substantially increased (Qin et al. 2005) and that correlative studies suggested that a gene locus including the human PACSIN1 gene 6p21.3-6p22 is associated with both schizophrenia (International Schizophrenia Consortium et al. 2009; Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium 2011; Stefansson et al. 2009) and idiopathic generalized epilepsy (Turnbull et al. 2005).

At the cellular level, syndapin I loss-of-function phenotypes also showed a consistence with symptoms found in schizophrenia patients. We observed a reduced dendritic arborization of both layer II/III and layer V neurons in slices from the cortex of syndapin I KO mice. Importantly, also schizophrenic patients showed impaired arborization of dendritic trees (Gu 2002; Jarskog et al. 2007; Krystal et al. 2003; Penzes et al. 2011). The finding that also heterozygous mice showed dendritic defects as syndapin I KO mice is in line with the finding that reductions of syndapin protein levels were found in proteomic analyses of schizophrenia patients (Pennington et al. 2008) and with our observation that efficient N-WASP recruitment to membranes critically depended on wild-type concentrations of syndapin I whereas brain extracts from heterozygous syndapin I KO mice were as insufficient as lysates from syndapin I KO mice. At the molecular/mechanistic level, our findings thereby suggest that cell biological defects associated with the behavioral and cellular schizophrenia-like phenotypes of syndapin I KO mice include neuromorphogenesis defects mediated by its cytoskeletal binding partner N-WASP. Consistently, syndapin I SzM showed strongly reduced associations with two cytoskeletal components critical for neuromorphogenesis, Cobl and N-WASP. These mechanistic insights also shed new light on the findings of reduced Arp2/3 complex and N-WASP mRNA levels in schizophrenia patients (Datta et al. 2017; Pennington et al. 2008).

Our studies furthermore unveiled severe glutamate receptor dysfunctions in syndapin I KO mice. The observed defects are not based on a reduced expression but instead represent a hypofunction of glutamate receptors. GluA1 and GluA2 clustering was impaired. The postsynaptic availability of both GluA1 and GluA2 hereby was negatively affected by both, reductions of GluA1 and GluA2 cluster sizes and an additional reduction of the numbers of GluA1 clusters in synaptic areas.

Receptor cluster formation requires associations with PSD proteins and the underlying actin cytoskeleton (Choquet and Triller 2013). The observed changes in actin cytoskeletal dynamics revealed by FRAP analyses of dendritic spines, the fact that the PSD components PICK1 (Anggono et al. 2006) and ProSAP1/Shank2 also associate with syndapin I (Schneider et al. 2014), and the finding that the spatial distribution of the receptor scaffold protein ProSAP1/Shank2 in postsynapses of hippocampal neurons was disrupted when syndapin I was lacking (Schneider et al. 2014) would explain the observed increase in GluA1 mobility, the reduced receptor cluster sizes, and the reduced synaptic availability of GluA1 in syndapin I KO mice.

Several genes associated with glutamate receptor function were linked to schizophrenia susceptibility by gene association studies, and this led to the idea that defects in the glutamatergic system can also underlie the cognitive disorganization in schizophrenia (Bowie and Harvey 2006; Cherlyn et al. 2010; Frankle et al. 2003; Koch 2007). In line with this view, postmortem studies of brains from schizophrenia patients showed that NMDAR and/or AMPAR expression was frequently altered in the cortex and the hippocampus (Rubio et al. 2012). Furthermore, besides glutamate receptor SU (Ballard et al. 2002; Mohn et al. 1999; Wiedholz et al. 2008), also variants of the syndapin I interaction partners PICK1 and ProSAP1/Shank2 genes have recently been linked to schizophrenia (Chen et al. 2017; Peykov et al. 2015).

Our data demonstrate that syndapin I is required for proper AMPAR function during basal synaptic transmission and during synaptic plasticity processes. Theta burst-induced LTP was impaired, and NMDA-induced LTD also was disrupted. Syndapin I plays an important role in synapse maintenance and also in synapse remodeling in response to activity, as highlighted by the impairments of synaptic GluA1 anchoring and GluA1 trafficking. The LTP defects we observed in syndapin I KO mice may result from a combination of changes in receptor anchoring in the PSD, alterations in the postsynaptic scaffold, and changes in actin cytoskeletal organization and dynamics in spines, as syndapin I interfaces with cytoskeletal components and syndapin I deficiency led to less stable F-actin in dendritic spines.

Our findings of a disrupted glutamate receptor endocytosis in syndapin I KO mice within the context of the F-actin–rich postsynaptic compartment can be explained by the findings that syndapin I associates with endocytic machinery components of the dynamin family (Qualmann and Kelly 2000; Qualmann et al. 1999), that syndapin I in addition links the dynamin-based machinery of endocytic vesicle formation to actin filament formation via syndapin I dimer formation (Kessels and Qualmann 2006), that syndapin I is crucial for membrane recruitment of dynamins in synaptic compartments (Koch et al. 2011), and that inhibition of the endocytic machinery component dynamin blocks glutamate receptor endocytosis. A failure to internalize AMPARs prevents the implementation of LTD (Wang and Linden 2000).

Apart from the mechanistic considerations, it is also noteworthy that in line with alterations of the other syndapin I binding partners, also alterations of dynamin protein levels were linked to schizophrenia (Pennington et al. 2008). The fact that we, in addition to GluA1, additionally observed a failure to internalize endogenous GluA2 in syndapin I KO neurons incubated with NMDA hereby is in line with an earlier report describing a reduced internalization of overexpressed myc-tagged GluA2 in dissociated hippocampal neurons subjected to syndapin I RNAi and prolonged incubation with NMDA (Anggono et al. 2013). Whereas these observations with overexpressed myc-GluA2 were attributed to modulation of GluA2/3 heterodimers via PICK interactions (Anggono et al. 2013), our analyses of both endogenous GluA1 and GluA2 in contrast clearly support an effect on GluA1/2 heteromers.

In addition to the defects in GluA1/2 heterodimer organization, trafficking, and function, our biochemical and electrophysiological analyses also unveiled strong defects in NMDAR function in syndapin I KO mice. This argues for hypofunction of both AMPARs and NMDARs in syndapin I KO mice, which, for example, also manifest in the imbalances of NMDA and AMPA eEPSCs. Hypofunction of both AMPARs and NMDARs may also explain why syndapin I KO mice show symptoms similar to both AMPAR and NMDAR SU loss-of-function models (Ballard et al. 2002; Mohn et al. 1999; Wiedholz et al. 2008).

Since syndapin I KO furthermore causes hypofunctions of both AMPARs and NMDARs without affecting total AMPAR and NMDAR SU protein levels, syndapin I KO mice may represent a valuable model system that, in contrast to NMDAR of AMPAR KO, is not limited by having the receptors themselves manipulated by genetic means. This provided new opportunities to uncover molecular and cellular mechanisms underlying schizophrenic dysfunctions and highlighted that the schizophrenia-like behavioral phenotypes observed upon syndapin I KO are accompanied by a lack of dendritic tree complexity, altered cytoskeletal dynamics, AMPAR and NMDAR hypofunctions, and impairments in both LTD and LTP.

Funding

Fonds Wetenschappelijk Onderzoek (FWO) Vlaanderen (G0D76.14 to D.B.); Katholieke Universiteit Leuven (KU) (GOA 12/08, TRPLe to D.B.); Deutsche Forschungsgemeinschaft (DFG) (Ke685/3-1 and 3-2 to M.M.K. as well as Qu 116/5-1, 5-2 and 6-2; TRR166 project B05 to B.Q.).

Notes

We thank S. Berr, M. Öhler, M. Roeder, C. Scharf, B. Schade, A. Kreusch, K. Sowa, and A. Reichel for technical help, S.A. Hofbrucker-MacKenzie for help with freeze fracture EM, C.A. Hübner for generously providing access to equipment, and M. Korte, J.M. Henley, and S. Kindler for plasmids. Conflict of Interest: None declared.

Author notes

Nicole Koch, Dennis Koch, Sarah Krueger and Jessica Tröger have contributed equally to this work.