Aberrant synaptic release underlies sleep/wake transition deficits in a mouse Vamp2 mutant

Sleep-wake transitions are modulated through extensive subcortical networks although the precise roles of their individual components remain elusive. Using forward genetics and in vivo electrophysiology, we identified a recessive mouse mutant line characterised by a reduced propensity to transition between all sleep states while a profound loss in total REM sleep time was evident. The causative mutation, an Ile102Asn substitution in VAMP2, was associated with substantial synaptic changes while in vitro electrophysiological investigations with fluorescence imaging revealed a diminished probability of vesicular release in mutants. We conclude that the synaptic efficiency of the entire subcortical brain network determines the likelihood that an animal transitions from one vigilance state to the next.

To identify mouse lines with sleep deficits, we used home-cage video-tracking to measure immobility-defined sleep (7) in cohorts from G3 pedigrees generated in a large ENU mutagenesis programme (8). By plotting percentage time spent immobile during light and dark phases (Fig. 1A,B) we identified a pedigree, called restless (rlss, MGI:5792085), where multiple individuals expressed reduced immobility. Differences were particularly evident during the light phase and were confirmed by screening a second cohort from the same pedigree. Further analysis in 1 hr time bins indicated a strong effect throughout the first eight hours of the light phase and towards the end of the dark phase (Fig. 1C). Using DNA from affected individuals, we mapped the non-recombinant mutant locus to a 35 Mb region on Chromosome 11 (Fig. 1D). Whole genome sequencing revealed a single high-confidence coding sequence mutation within the non-recombinant region. Sequence validation in multiple affected individuals consistently identified a single coding sequence variant cosegregating with the mutant phenotype, a T441A transversion in Vamp2 (Syb2) resulting in an Ile102Asn substitution in the protein's transmembrane domain (Fig. 1E,F). VAMP2 is the major neuronal vesicular component of the Soluble N-Ethylmaleimide-sensitive Factor (NSF) Attachment Protein Receptor (SNARE) complex, fundamental to neurotransmitter exocytosis. Outcrossing affected individuals to wild-type mates confirmed the recessive phenotype, with immobility-defined sleep in heterozygotes being no different to wildtypes.
Statistical analysis of immobility in 1hr bins confirmed a significant genotype effect and pairwise comparisons confirmed that sleep in homozygotes was significantly reduced in the first seven hours of the light phase and the final seven hours of the dark phase ( Fig. 1G; for this and all other statistical analysis, refer to Table S1). Further analysis of mutants also identified unusual sleep behaviours where individuals showed no preference for nesting in sheltered parts of the home-cage, frequently moved sleep position within the cage and assumed atypical sleep postures (Fig. 1H). Conventional circadian parameters using cages with wheels could not be reliably measured in homozygotes (Fig. 1J). However, Passive Infrared monitoring of circadian rhythms (9), demonstrated that homozygote circadian period was no different from controls while period amplitude and activity parameters were affected ( Fig. 1J,K).
To investigate sleep-wake architecture in Vamp2 rlss mice, we implanted cortical EEG and nuchal EMG electrodes ( Fig. 2A) and performed 24-h baseline home cage sleep recordings (Fig. 2B). EEG/EMG signatures were typical of wakefulness, NREM and REM sleep (10), although more pronounced slow-wave activity (0.5-4Hz) during NREM sleep and slower theta oscillations during REM sleep were noted in Vamp2 rlss (Fig. 2C, Fig. S1A) (Fig. 2D). Plotting the distribution of vigilance states across 24-h showed the difference between genotypes was especially apparent for REM sleep (Fig.   S1B, S2A), a conclusion supported by a decreased REM-to-total sleep ratio (Fig. 2E), while the decrease in NREM sleep during the dark period was partially compensated during the second half of the light phase (Fig. S1B, S2A). Investigations as to whether reductions in sleep are related to a reduced probability of state transitions showed that the number of wake-to-REM transitions was markedly reduced in homozygotes (Fig. 2F). Moreover, increased wakefulness and longer wake episodes in homozygotes (Fig. 2D, 2G, Fig. S2B) suggested increased wake state continuity. In considering all wake episodes other than brief awakenings (≤16 sec), their frequency was halved in Vamp2 rlss mice, suggesting that once wakefulness is initiated it is sustained for longer durations. Similar dynamics also manifested in state transitions within sleep; once a NREM sleep episode was initiated, it was less likely to transition into REM sleep (Fig. 2H, Fig S2C). Almost 90% of all NREM episodes terminated in REM sleep in WT mice, compared to less than 70% in Vamp2 rlss mice (Fig.   2J). This indicates that, along the continuum of wake -> NREM -> REM occurrences, the inertia to transition to the next state is increased in Vamp2 rlss mice (Fig. 2K). Consistently the distribution of episode durations for NREM and REM sleep showed an increased incidence of longer episodes for both sleep states in Vamp2 rlss mice (Fig. 2L,M), further suggesting that once a specific state is initiated it is less likely to terminate.
Investigations into the nature of the Vamp2 rlss mutation were driven by observations that this allele was phenotypically distinct from either heterozygous or homozygous null mutant mice (11). Western blots from whole brain, cortex or hippocampus indicated that the mutant protein was not as stable as wildtype, with levels ranging from 25-65% of controls. Notably the associated tSNARE protein syntaxin 1a (STX1A) was unaffected (Fig. 3A, Fig S3A,B).
Immunofluorescence labelling of hippocampal neuronal cultures confirmed that synaptic VAMP2 levels, when normalised to that of the synaptic active zone marker BASSOON, were about 50% that of controls ( Fig. 3 B,C, Fig. S4 C-E) and these deficits were mirrored in synaptic fractions from cortex or hippocampus ( Fig. 3 C,D). We investigated whether residual mutant VAMP2 protein was functionally anomalous. Based on evidence that mutations in the transmembrane domain of VAMP2 may affect vesicle fusion, fusion pore dynamics and exocytosis (12)(13)(14), we conducted reciprocal immunoprecipitations using antibodies to VAMP2 or STX1A. Using either native protein or tagged expression protein for immunoprecipitation, we were unable to detect a consistent difference in protein interaction ( Fig. S3E-H) although neither may reflect the true dynamism of the interaction between vSNARE and tSNARE components in the context of their physiological lipid environment.
Elsewhere, we detected what may be a compensatory homeostatic response in mutants. In cultured neurons, dendritic branching and synaptic density were higher in mutants (Fig.   S4A,B). Golgi-stained brain sections also revealed an increase in dendritic spine count (Fig.   3D,E) with greater numbers of immature spines in mutants (Fig S5). At an ultrastructural level, no gross differences in hippocampal synaptic measures were identified. However, the density of synaptic vesicles and the density of docked vesicles were ~2-fold greater in homozygotes ( Fig. 3F-H).
We tested the effect of Vamp2 rlss on synaptic vesicle release and functional vesicular pool sizes in hippocampal synapses in neuronal cultures. Cultured neurons were transduced with the fluorescence vesicular release reporter sypHy (15). Active synaptic boutons were identified using a short burst of high frequency stimulation (20 APs x 100 Hz) which triggers exocytosis of the readily releasable pool (RRP) of vesicles. Fluorescence responses to single APs were next measured in identified presynaptic boutons (Fig. 4A,B) enabling an estimate of the average release probability (pv) of individual RRP vesicles (16). Whilst functional RRP size was similar, pv in response to a single AP was profoundly decreased in Vamp2 rlss neurons (Fig. 4C). Furthermore, in agreement with electron microscopy data we observed a ~2-fold increase of the total recycling pool size (TRP) and the total number of SV in Vamp2 rlss neurons as measured by fluorescence sypHy responses to high K+ and NH4Cl containing solutions respectively (Fig. 4D). Decreases in pv are normally associated with an increase in short-term synaptic facilitation (17). Indeed, experiments in acute hippocampal slices revealed a ~2-fold increase in short term facilitation in Vamp2 rlss slices during short high frequency stimulation of the Shaffer collateral synapses (Fig. 4e,f).
Given the role of VAMP2 in fundamental neural mechanisms and the electrophysiological deficits seen in homozygotes, we expected mutants to display behavioural and sensory discrepancies in addition to changes in sleep. Surprisingly, performance in a wide test battery indicated only mild behavioural anomalies (Fig. S6) while sensory function was unchanged or even improved (Fig. S7,S8). However, homozygous behaviour in a number of tests pointed to deficits in working memory (Y-maze and Novel Object Recognition), attention (Marble burying, Novel object recognition) and social instability (Social dominance tube test) ( Fig. 4G-J). These irregularities were defined further through display of numerous stereotypical behaviours analysed using home-cage continuous video monitoring (Movie S1).
The sleep deficit identified here in Vamp2 rlss mice uncovers a previously unrecognised role for VAMP2 in sleep and may indicate a critical function for the transmembrane domain in which the mutation occurs. VAMP2 function has been profoundly well-characterised with respect to its fundamental role in vesicle fusion and exocytosis (18). Nevertheless, its functional characterisation in mammals in vivo has been hampered as the Vamp2 -/mouse mutant is perinatal lethal while Vamp2 +/show no behavioural phenotype beyond mild improvements in rotarod performance (19). The recent identification of multiple de novo variants of VAMP2 in humans has highlighted how mutations affecting the SNARE zippering mechanism can have severe developmental consequences (20). In contrast, the relatively mild consequences of the Vamp2 rlss mutation in mice has enabled us to examine features of VAMP2 function in adults. Moreover, recent publications (21,22) identified VAMP2 lead SNP associations for chronotype and sleep duration measures suggesting that subtle variations in VAMP2 function could influence sleep quality in humans.
Vamp2 rlss affords a new model to study the genetic mechanisms regulating vigilance state switching. The search for the key "nodes" in the brain which are responsible for instantaneous sleep-wake transitions have undoubtedly provided important insights into these mechanisms (23)(24)(25). However, as Vamp2 rlss mice display an increased stability of all vigilance states, it is possible that globally impaired vesicular release triggers an inherent inertia in their neuronal circuitry. Thus, it remains a possibility that sleep-wake states are Inbred strains and mutant colonies were maintained at MRC Harwell and cohorts were shipped as required. ENU mutagenesis and animal breeding regimes were performed as previously reported (8). Phenotyping was performed on mouse cohorts that were partially or completely congenic on the C57BL/6J Or C3H.Pde6b+ (26) background.

Video tracking screen for immobility-defined sleep
Video tracking was performed as described (7). Briefly, mice were singly housed and placed in light controlled chambers with CCD cameras positioned above the cages (Maplin, UK).
Monitoring during the dark was performed using infrared illumination. Analysis of videos by ANYmaze software (Stoelting) was used to track mouse mobility and immobility-defined sleep (successive periods of >40s immobility). Mice for screening were provided by a largescale ENU mutagenesis project (8).

Gene mapping and sequencing
Affected individuals from G3 pedigrees were used for mapping and mutation detection.

Mutations were mapped utilising the Illumina GoldenGate Mouse Medium Density Linkage
Panel (Gen-Probe Life Sciences Ltd, UK), providing a map position resolution of about ~20Mb. Following mapping, whole genome sequencing of the G1 founder male was carried out as described (8), variants were confirmed by Sanger sequencing.

EEG surgery, recording and analysis
Average ages and weights of mice at the time of surgery were 23.8 ± 1.2 weeks & 31.9 ± 0.8 g and 14.7 ± 1.5 weeks & 31.7 ± 1.5 g for Vamp2 rlss and WT mice, respectively. Age differences were permitted so that animals were matched by weight. Previous data indicates that this age difference does not contribute to the differences in sleep observed (27,28).
Vamp2 rlss mice were fed a high-nutrient diet in addition to regular pellets. Surgical methods, including drug administration and aseptic techniques, were as described (28). EEG screws (Fine Science Tools Inc.) were placed in the frontal (anteroposterior +2 mm, mediolateral 2 mm) and occipital (anteroposterior -3.5 to -4 mm, mediolateral 2.5 mm) cortical regions, reference and ground screws were implanted above the cerebellum and contralaterally to the occipital screw, respectively. Two stainless steel wires were inserted in the neck muscle for EMG. Mice were singly housed post-surgery.

Behavioural phenotyping.
Circadian wheel running: Circadian wheel running was performed as previously reported (29). Briefly, mice were singly housed in cages containing running wheels, placed in light controlled chambers and wheel running activity monitored via ClockLab (Actimetrics).
Animals were monitored for five days in a 12 hour light/dark cycle followed by twelve days in constant darkness.
Passive infrared analysis: Mice were additionally analysed for circadian activity using the COMPASS passive infrared system as described (9). Animals were monitored for four days in a 12 hour light/dark cycle followed by ten days in constant darkness.
Open field behaviour: Mice were placed into one corner of a walled arena (45cm X 45cm) and allowed to explore on two consecutive days for 30 minutes (30). Animal movements and position were tracked using EthoVision XT analysis software (Noldus).
Light/Dark box: Individual mice were placed into one corner of an enclosed arena separated into light and dark compartments (31). Over 20 minutes animal movements and position were monitored by EthoVision XT.
Marble burying: Briefly, a cage was prepared with approximately 5cm deep sawdust bedding (31). 9 marbles were placed on the surface of the sawdust, evenly spaced in a regular pattern. The mouse was introduced and left in the cage with the marbles for 30 minutes.
After 30 minutes the number of marbles remaining unburied, partially buried or completely buried was counted. Statistical differences were determined using the Mann-Whitney test.
Acoustic startle response (ASR) and prepulse inhibition (PPI): ASR and PPI were measured as in (32). Mice were placed in the apparatus (Med Associates, VT, USA) and responses to sound stimuli were measured via accelerometer.
Novel object recognition: A modified version of the novel object recognition task adopted for use in the home cage was used using video analysis of behaviours. On day one, animals were presented with one novel object (glass jar or lego bars) in a corner of their home-cage for 30 minutes at the beginning of the dark phase. On the following night the same object was introduced at the same position in the home-cage, again at the beginning of the dark phase for 30 min. On the third night a second object, different to the first one, was introduced at a different position in the home-cage but at the same time of night and for the same duration. Time spent inspecting objects was measured manually from video-recordings using a stop watch.
Social dominance test: Dominant and submissive behaviours were assessed for pairs of male mice using a specialised Plexiglass tube (31). Pairs of mice, one wildtype and one homozygous mutant from different group-housed cages, were placed in the tube and behaviour registered as dominant ("win") or subordinate. 5 pairs were used and total numbers of "wins" per genotype was recorded. Statistical differences were ascertained using a Chi-Squared test.

Y-maze:
A forced alternation y-maze test was used to evaluate short term working memory in mice (31). Mice were video tracked at all times using Ethovision XT, preference for the novel arm is indicated by an occupancy of greater than 33%. (34)  Heat sensitivity, Hot Plate Test: A hot plate (BioSeb, Chaville, France) set at 51 o C was used for this test (31). Animals were placed on the plate and the latency to the first paw reflex (withdrawal reflex of one paw) was measured.

Mechanical sensitivity, von Frey Test: Mechanical sensitivity was assessed as in
Optokinetic response. Visual acuity was assessed by head tracking response to a virtualreality optokinetic system (Cerebral Mechanics Inc) (33).
Auditory brainstem response (ABR). Auditory Brainstem Response tests were performed using a click stimulus in addition to frequency-specific tone-burst stimuli as described (34).
ABRs were collected, amplified and averaged using TDT System 3 (Tucker Davies Technology) driven by BioSig RZ (v5.7.1) software. All stimuli were presented free-field to the right ear of the anaesthetised mouse, starting at 70 dB SPL and decreasing in 5 dB steps. Auditory thresholds were defined as the lowest dB SPL that produced a reproducible ABR trace pattern and were determined visually.
Grip strength: Grip strength was assessed using a Grip Strength Meter (BioSeb, Chaville, France). Readings were taken from all four paws, three times per mouse as per manufacturer's instructions (30). Measures were averaged and normalised to body weight.
HCA home cage analysis: Group housed animals were monitored as described (35). Briefly, group housed mice were tagged with RFID micochips at 9 weeks of age and placed in the Home Cage Analysis system (Actual Analytics, Edinburgh) which captured mouse behaviour using both video tracking and location-tracking using RFID co-ordinates.

Golgi-Cox Staining, Spine Count Analysis
Brains dissected from 16-week old females were used for analysis. Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies Inc, USA) according to the manufacturer's instructions. 100m sections were taken using a vibratome, mounted upon charged slides, cleared in Histo-Clear (National Diagnostics, UK) and coverslipped. Neurons were viewed on an Axio-Observer Z1 (Zeiss) microscope. Z stack images were processed using extended depth of focus and Zen software (Zeiss).
Visualisation and measurements were taken using ImageJ (http://rsbweb.nih.gov/ij/). The number and type of spines on each neurite -stubby, long, mushroom and branched (36)were counted. At least 50 neurites per region per animal were analysed.

Electron Microscopy
Brains were fixed by cardiac perfusion (buffer solution: sodium cacodylate buffer pH 7. components. The tagged protein complexes were then recovered with the elution buffer supplied. Immunoprecipitation of native protein complexes from hippocampal lysates was carried out as described previously (32).

Electrophoresis and Western Blotting
Samples were prepared in NuPAGE LDS Sample Buffer (Invitrogen: NP0007) supplemented with NuPAGE Sample Reducing Agent (Invitrogen TM : NP0004) and heated at 70 ℃ for 10 mins. Equal amounts of protein were loaded and resolved on NuPAGE Bis-Tris 4-12 % gel (Invitrogen: NP032A) and then transferred to nitrocellulose membarane contained in iBlot Gel Transfer Stacks (Invitrogen: NM040319-01) using an iBlot TM Gel Horizontal Transfer Device (Invitrogen: IB21001). After blotting, membranes were incubated with primary antibodies, diluted in TBST containing 5% skimmed milk, at 4 ℃ overnight. Subsequently, membranes were washed with TBST and incubated with IR700 and IR800 secondary antibodies (LI-COR Biosciences; Lincoln, NE) for 2 hours at room temperature. After further washes, immunoreactive bands were visualized using an Odyssey Infrared Imaging System (LI-COR Biosciences; Lincoln, NE).

Immunofluorescence labelling
To evaluate VAMP2 expression and localization, primary hippocampal neurons were fixed and stained after DIV 15 to 23. Hippocampal cultures were first washed in PBS then fixed with 4% PFA for 10min and blocked in a 20% donkey serum blocking solution in TBST for 1 h at room temperature. All primary antibody incubations were conducted over 2 nights at applied in 1% blocking serum to minimize antibodies binding to the plastic-ware. Samples were washed 3x with TBST for 10 min at room temperature between steps. After antibody treatment and final washes, coverslips with cells were mounted onto glass slides with mounting medium (ProLong TM, Gold antifade reagent, Invitrogen) then dried in the dark, overnight at 4 o C before imaging.

Image and Sholl analysis
Fluorescence images were acquired at room temperature with an inverted confocal microscope (Zeiss, LMS 710) using a Plan-Apochromat 40x/1.4 (Oil, DIC, M27) objective or 20x/0.8 (M27) objective to get 4 layer z-stacks. For the synaptic intensity experiment, the same settings for laser power, PMT gain and offset and z-stack thickness were used. The pinhole size was set to 1 Airy unit for the shortest wavelength channel and the faintest image. The z-stacks acquired were compressed into single layer images by maximum projection. For intensity quantification, multi-channel fluorescence images were first converted into monochromatic and colour inverted pictures by imageJ. All fluorescence intensity analysis was conducted at the same setting.
For synaptic distribution analysis, conditions were optimized independently in order to capture data from all neurites for a single cell. PMT gain was optimized individually, using 10% area oversaturation on the shortest wavelength as a reference for all channels (n > 12 for each genotype). Then z-stacks acquired were compressed into single layer images by maximum projection. To quantify synaptic protein distribution further, we investigated VAMP2 distribution pattern by analysing VAMP2 containing synaptic projections and associated neuritic branching using a MATLAB-based SynD program (38). Multi-channel fluorescence images were acquired using ZEN software and converted into monochromatic and color inverted pictures by ImageJ for further analyses. Statistical comparisons were performed using a Mann-Whitney test with genotype and intensity as independent factors.

SypHy fluorescence imaging experiments in neuronal cultures.
Primary hippocampal neurons were maintained in modified Tyrode solution containing (mM) 125 NaCl, 2.5 KCl, 2 MgCl2, 2 CaCl2, 30 glucose and 25 HEPES (pH 7.4) supplemented with NBQX (10 µM, Ascent Scientific) and DL-AP5 (50 µM). APs were evoked by field stimulation via platinum bath electrodes separated by 1 cm (12.5-15 V, 1-ms pulses). To estimate the relative TRP size and total numbers of SVs neurons were perfused with Tyrode modified solution containing either 45 mM KCl or 50 mM NH4CL respectively. Images were acquired via 63x objective using a Prime 95B CMOS camera (Photometrics) mounted on an inverted Ziess Axiovert 200 microscope equipped with a 488nm excitation LED light source and a 510 long-pass emission filter. Exposure time was 25ms.

Image and data analysis of sypHy experiments
Images were analysed in ImageJ and MATLAB using custom-written plugins. A binary mask was placed on all varicosities that were stably in focus throughout all trials and responded to 20 AP 100 Hz burst stimulation. To estimate sypHy fluorescence changes induced by 20 AP x 100 Hz, 1AP, KCl and NH4CL the difference between the mean of 8 frames before and 8 frames after the stimulus was calculated for all synapses in the field of view (10 -100 range).
After subtracting the background, the data were normalised to the resting sypHy signal (F0).

Data analysis and statistics
Unless otherwise stated statistical differences were established using a student's t-test.
Behavioural phenotyping, synaptic spine analysis, electron microscopy, western blot quantification and immunohistochemical quantification were analysed using GraphPad Prism 7 (GraphPad Software). EEG sleep analysis was performed using SPSS (IBM Corp). SypHy fluorescence imaging experiments were analysed using GraphPad Prism 7 and SigmaPlot (Systat Software Inc). The electrophysiological recordings were analysed using LabVIEW (National Instruments), PClamp 10 (Molecular Devices) and Matlab (MathWorks) software.
Significance level for all analysis was set at P<0.05.