Behavioral Phenotyping of an Improved Mouse Model of Phelan–McDermid Syndrome with a Complete Deletion of the Shank3 Gene

Abstract Phelan–McDermid syndrome (PMS) is a rare genetic disorder in which one copy of the SHANK3 gene is missing or mutated, leading to a global developmental delay, intellectual disability (ID), and autism. Multiple intragenic promoters and alternatively spliced exons are responsible for the formation of numerous isoforms. Many genetically-modified mouse models of PMS have been generated but most disrupt only some of the isoforms. In contrast, the vast majority of known SHANK3 mutations found in patients involve deletions that disrupt all isoforms. Here, we report the production and thorough behavioral characterization of a new mouse model in which all Shank3 isoforms are disrupted. Domains and tasks examined in adults included measures of general health, neurological reflexes, motor abilities, sensory reactivity, social behavior, repetitive behaviors, cognition and behavioral inflexibility, and anxiety. Our mice are more severely affected than previously published models. While the deficits were typically more pronounced in homozygotes, an intermediate phenotype was observed for heterozygotes in many paradigms. As in other Shank3 mouse models, stereotypies, including increased grooming, were observed. Additionally, sensory alterations were detected in both neonatal and adult mice, and motor behavior was strongly altered, especially in the open field and rotarod locomotor tests. While social behaviors measured with the three-chambered social approach and male-female interaction tests were not strongly impacted, Shank3-deficient mice displayed a strong escape behavior and avoidance of inanimate objects in novel object recognition, repetitive novel object contact, marble burying, and nest building tasks, indicating increased novelty-induced anxiety. Similarly, increased freezing was observed during fear conditioning training and amygdala-dependent cued retrieval. Finally, deficits were observed in both initial training and reversal in the Barnes maze and in contextual fear testing, which are memory tasks involving hippocampal-prefrontal circuits. In contrast, working memory in the Y-maze spontaneous alternation test was not altered. This new mouse model of PMS, engineered to most closely represent human mutations, recapitulates core symptoms of PMS providing improvements for both construct and face validity, compared to previous models.


Introduction
Phelan-McDermid syndrome (PMS) is a rare and complex neurodevelopmental disorder that manifests with global developmental delay, mild dysmorphic features, motor deficits, variable degrees of intellectual disability (ID), and absent or delayed speech. Additionally, autism spectrum disorder (ASD), epilepsy, attention deficits, and recurrent medical comorbidities are common in patients with PMS (Phelan and McDermid, 2012;Betancur and Buxbaum, 2013;Soorya et al., 2013;Sarasua et al., 2014a). Recent studies show that PMS is emerging as one of the most frequent and penetrant monogenic causes of autism and ID (Sykes et al., 2009;Betancur and Buxbaum, 2013;Soorya et al., 2013;Leblond et al., 2014).
Despite overlapping etiologies between patients, there is a tremendous heterogeneity in the expression and severity of the phenotype (Cusmano-Ozog et al., 2007;Dhar et al., 2010;Phelan and Betancur, 2011;Soorya et al., 2013). This is no doubt in part due to the complex nature of in the genetic etiology of PMS (De Rubeis et al., 2018). While a large body of data indicates that haploinsufficiency of SHANK3 is the key contributor for the neurobehavioral manifestations of PMS, it can be caused by a variety of genetic rearrangements including unbalanced translocations, ring chromosome 22, terminal deletions (ranging from deletions of just SHANK3 to large deletions of up to 9 Mb), and interstitial deletions or point mutation within the SHANK3 gene (Durand et al., 2007;Moessner et al., 2007;Sykes et al., 2009;Bonaglia et al., 2011;Phelan and McDermid, 2012;Soorya et al., 2013;Leblond et al., 2014;De Rubeis et al., 2018).
Genotype-phenotype analyses have shown positive correlations between the size of the deletion and the number and/or severity of some phenotypes (Luciani et al., 2003;Dhar et al., 2010;Bonaglia et al., 2011;Soorya et al., 2013;Sarasua et al., 2014b). However, findings on specific clinical variables have not been consistent across studies. Importantly, it has become clear that indels or point mutations that impact SHANK3 alone can lead to all of the neurobehavioral phenotypes of PMS (De Rubeis et al., 2018). The SHANK3 gene has multiple promoters and is alternatively spliced and the number of Shank3 isoforms can be extensive (Maunakea et al., 2010;Benthani et al., 2015). Some de novo microdeletions or mutations of SHANK3 can therefore affect some but not other SHANK3 isoforms. The genetic heterogeneity of PMS underscores the importance of studying a wide range of mutations and deletions. SHANK3 (ProSAP2) is a major scaffolding protein that forms a key structural part of the postsynaptic density of excitatory glutamatergic synapses. SHANK3 contains multiple proteinprotein interaction domains that each mediates specific protein-protein interactions at synapses. Moreover, the expression and alternative splicing of Shank3 isoforms or even their subcellular distribution has been shown to be cell-type specific, activity dependent as well as regionally and developmentally regulated  raising the possibility that differing SHANK3 isoforms may play distinct roles in synaptic developmental and function and hence may make distinct contributions to the pathobiology of PMS.
More than a dozen isoform-specific Shank3 mouse models have been independently generated (Table 1). As expected, these models shared some similarities but also showed significant differences in molecular, synaptic, and behavioral phenotypes. Depending on the targeted exons, alterations have been reported in motor functions, social interactions, ultrasonic vocalizations, repetitive grooming, cognitive functions, and anxiety. However, very high variability has been observed regarding the presence or the intensity of such impairments across several types of Shank3-deficient models or even across different cohorts of the same model. These models are based on exonic deletions that have not been reported in human and do not reflect the vast majority of known PMS cases, which are caused by deletions affecting all SHANK3 isoforms. There was therefore an urgent need to develop an animal model with broader construct validity for PMS to fully understand the consequences of a complete deletion of SHANK3 across the range of behavioral phenotypes, which we achieved through a deletion of exons 4-22.
Interestingly, as our work was progressing, a completely independent mouse model, similarly targeting exons 4-22, was reported (Wang et al., 2016b). These mice highlight cortico-striatal circuit abnormalities and demonstrate a behavioral phenotype that resemble features of PMS. We therefore decided to conduct a comprehensive and behavioral evaluation of our mouse model evaluating many more phenotypes relevant to PMS and ASD. Criti- cohort consisted of 57 newborn mice (16 WT, 32 Het, and nine KO) from nine independent litters. Cohorts 3 (30 adult male mice, 11 WT, 10 Het, and nine KO) and 4 (27 adult male mice, 11 WT, 10 Het, and nine KO) were tested between 3 and 10 months of age according to the schedule described in Table 2. In each adult cohort, all mice were born within two weeks of each other, and generally only one triplet came from any given individual litter of PSD fractions from wild-type, heterozygous, and homozygous mice were subjected to immunoblotting with either the N367/62 anti-Shank3 antibody directed against an epitope in the SH3 domain or the H160 C-terminal antibody. Immunoblots show that all Shank3 protein bands are absent in KO brains. The migration of molecular weight markers is shown on the left (in kilodaltons) and an immunoblot for ␤III-tubulin as a loading control is shown below.
Original full scans of immunoblots are displayed in Extended Data Figure 1-1. C, RT-PCR analysis for specific Shank3 transcripts in Shank3 ⌬4-22 mice. Brain-derived mRNAs from wild-type and homozygous mice were subjected to RT-PCR targeting different isoforms. All transcripts were absent in Shank3 ⌬4-22 homozygous mice. D, Distribution of genotype. A deficit in the number of Shank3 ⌬4-22 knockout mice was observed at the time of weaning. E, Survival curve of Shank3 ⌬4-22 wild-type, heterozygous and homozygous mice between 2 and 22 months. WT, wild-type mice; Het, heterozygous mice; KO, homozygous knockout mice. ‫:ء‬ p Ͻ 0.05, ‫:ءء‬ p Ͻ 0.1. Confirmation mice. Behavioral experiments were conducted between 9 A.M. and 5 P.M. during the light phase of the 12/12 h light/dark cycle in dedicated testing sound-attenuated rooms. Mice were brought to the front room of the testing area at least half an hour before the start of experiments. All three genotypes were tested on the same day in randomized order by two investigators who were blind to the genotypes. Behavioral tests were conducted in the order and at the ages indicated in Table 2 and included developmental milestones, cage observation, neurologic and motor reflexes, open field, elevated zero-maze, Y-maze, beam walking, grip strength, gait analysis, rotarod, three-chambered social interaction task, nest building, novel object recognition, fear conditioning, pre-pulse inhibition, tail flick, olfactory habituation/dishabituation, buried food, social transmission of food preference, marble burying, four-object repetitive novel object contact task, male-female social interaction, and Barnes maze. Behavioral results are not described in the order they were tested in an effort to ease presentation and interpretation of the data.
where and when possible. At this time, pups were identified by paw tattoo using a nontoxic animal tattoo ink (Animal Identification & Marking Systems Inc) inserted subcutaneously through a 30-gauge hypodermic needle tip into the center of the paw. Individual pups were removed from the litter and placed on cotton pads in a heated cage under a heating lamp throughout the testing. Each subject was tested at approximately the same time of day. For all the timed tests, a 30-s cutoff was used and nonresponding animal received a score of 30 s. Most responses were considered positive only after they had been observed for two consecutive days. The physical development was measured by following the weight (postnatal day 1 to 21), eye opening (postnatal days 9 to 20), tooth eruption (postnatal days 7 to 18), the ear development (postnatal day 1 to 9), and the fur development (postnatal days 1 to 14) using the following scales. Eye opening, per eye: 0 ϭ eye fully closed, 1 ϭ eye partially opened, 2 ϭ eye full opened, tooth eruption, scored separately for bottom and top incisors: 0 ϭ incisors not visible, 1 ϭ incisors visible but not erupted, 2 ϭ incisors fully erupted. Ear development, per ear: 0 ϭ ear bud not detached from the pinna, 1 ϭ ear flap detached from the pinna, ear fully developed on the back of the ear). Fur development: 1 ϭ bright red, 2 ϭ nude, pink, 3 ϭ nude, gray, 4 ϭ gray, fuzzy on back and shoulder, 5 ϭ black hair on back, gray fuzzy belly, 6 ϭ body fully covered.
Sensory development was assessed using cliff aversion (postnatal days 2 to 14), auditory startle (postnatal days 6 to 18), rooting reflex (postnatal days 2 to 10), ear twitch (postnatal days 7 to 15), and forelimb grasp (postnatal days 4 to 14) using the following measures. For cliff aversion, the subject was placed on the edge of a Plexiglas platform with a 30-cm cliff with its nose and forefeet over the edge. The latency to move away from the edge was recorded. Auditory startle was measured in response to an 80-dB click 30 cm above the mouse and was considered present when the pup moved immediately after the presentation of the auditory stimulus. For the rooting reflex, the side of the pup's face were bilaterally stimulated with two cotton swabs. The reflex was considered present when the pup crawled forwards pushing the head during the stimulation. For the ear twitch, the ear of the pup was stimulated with the tip of a cotton swab that was previously pulled to form a filament. Both ears were successively stimulated and the test was considered positive when the pup turned its head or jumped in response to the stimulation. The forelimb reflex was tested by gently stimulated the front paws with the loop of a small bended metallic wire. Each front paw was scored separately as follow: 0 ϭ no response to stimulation, 1 ϭ paw folding in response to the stimulation, 2 ϭ paw grasping the wire in response to the stimulation, 3 ϭ grasp strong enough to hold for at least 1 s when the wire was lifted up.
Motor development was studied using surface righting (postnatal days 2 to 13), negative geotaxis (postnatal days 2 to 14), air righting (postnatal days 8 to 20), open field crossing (postnatal days 8 to 20), and rod suspension (P11-P20) using the following criteria. The surface righting was measured by the time for pups placed on their back to fully turn with all four paws on the ground. For negative geotaxis, pups were placed head down on a mesh covered plan that was slanted at a 45°angle, and the latency to either roll down, stay, or turn and move up the slope was recorded. For the air righting, the pup was dropped upside down at a height of 30 cm over a padded surface. Subjects received a score of 2 if they successfully righted themselves during the fall, 1 if they landed on the side and 0 if they did no turn. The open field crossing was measured by the time to exit a 13 cm in diameter circle when place on the center of the circle. For the rod suspension, the pups were gently grabbed by the trunk, brought up close to a 3-mm wooden rod 30 cm above a padded surface and released once they grabbed the rod with their front paws. The latency to stay suspended was recorded.

Physical factors, gross appearance, and spontaneous activity
Adult animals were handled daily for one week before starting behavioral testing and general health, weight (grams), length (centimeters), physical factors, gross appearance, and spontaneous activity were recording during handling using the following scales.

Motor testing Gait analysis
Motor coordination and gait patterns was observed as the subject was allowed to run the length of an elevated runway (dimensions: 152 cm long ϫ 10 cm wide) lined with white paper (Carter et al., 2001). After three training runs, the subject's paws were coated in nontoxic paint (different colors for hind and front paws) to record paw prints on two consecutive runs. The record displaying the clearest prints and most consistent gait for analysis of 50 cm was chosen to measure sway (mean distance between left and right paws), stride (mean distance between same side front and hind paws) and diagonal stance (mean distance between diagonally opposed front and hind paws).

Open field
Mice were tested in an open field (45 ϫ 45 cm) virtually divided into central and peripheral regions. Animal activity was recorded by video tracking (Noldus Ethovision). Each mouse was allowed to explore the apparatus for 60 min. The distance traveled, the number of rears and revolutions, the number of grooming bouts and cumulative grooming time, the number of head shaking or twitches, the number of entries in the center, and the time spent in the central and peripheral regions were recorded. Measures were recorded in 10-min intervals.

Rotarod
Motor coordination, endurance and learning was assessed in the Rotarod test (Omnitech Electronics Inc). Mice were placed on an elevated accelerating rod (3 cm in diameter) for three trials per day on two consecutive days. Each trial lasted for a maximum of 5 min, during which the Rotarod underwent a linear acceleration from 4 to 40 rpm. A 20-min interval was used between trials to avoid fatigue. Animals were scored for their latency to fall.

Beam walking
Subtle deficits in fine motor coordination and balance that might not be detected by other motor tests were assessed by the beam walking assay in which the mouse had to walk across an elevated horizontal wood beam (100 cm long, 1 m above bedding) to a safe dark box (Carter et al., 2001). Subjects were placed near one end in bright light, while the far end with the dark box was placed in darkness, providing motivation to cross. Performance was quantified by measuring the latency to start crossing, the time to reach the dark box or the time to fall, the total distance traveled and the number of paw slips or incomplete falls (mice able to climb back on the rod). Animals were successively trained on three different beams: 1 inch, ½ inch and ¼ inch diameter and scored on four consecutive trials per beam with 1 min of rest between trials on the same beam and 20 -30 min between each beam. Mice that did not reach the box after 2 min were gently placed inside the box and allowed to stay inside for 1 min.

Righting reflex
The subject was grasped by the nape of the neck and base of the tail, inverted so back faced down, and re-leased 30 cm above subject's home cage floor. Righting ability was scored as follow: 0 ϭ no impairment, 1 ϭ lands on side, 2 ϭ lands on back, 3 ϭ fails to right even when placed on back on the floor.

Hindlimb placing
Subject was lowered by the base of the tail until it grasped a horizontal wire grid with both forepaws. The grid was rotated to vertical and the tail was released. Mice were evaluated over three trials, 3 min apart for their latency to fall or latency to pull body on the grid and the ability to place hind paws was scored as follow: 0 ϭ grabs but falls, 1 ϭ grabs but hangs, 3 ϭ grabs and pulls body onto grid. Maximum cutoff was 60 s.

Hanging
The subject, held from the base of the tail, was allowed to grasp a wooden rod with both forepaws, rotated to horizontal and release. Test was repeated three times with a 3-min interval between trials and a 60-s maximum cutoff. Both the latency to fall and overall performance scored as follow were recorded: 0 ϭ does not grasp, 1 ϭ grasps but falls immediately, 2 ϭ grasps but then falls off, 3 ϭ grasps and stays on for 60 s.

Negative geotaxis
The subject was placed on a wire mesh grid and the grid was lift vertically, with subject facing down. Test was repeated three times with a 3-min interval between trials and a 60-s maximum cutoff. Both the latency to fall and overall performance scored as follow were recorded: 0 ϭ falls off, 1 ϭ does not move, 2 ϭ moves but does not turn, 3 ϭ turns but does not climb, 4 ϭ turns and climbs up.

Inverted screen
The subject was placed on a grid screen. The grid was waved lightly in the air, then inverted 60 cm over a cage with soft bedding material. Mice were tested only one time with a 60-s maximum cutoff, and the latency to fall was recorded.

Grip strength
Forelimb muscle strength and function was evaluated with a strength meter (Ametek). This test relies on the instinctive tendency of mice to grasp an object with their forelimbs. The animal was pulled backward gently by the tail, while grasping a pull bar connected to a tension meter and the force at the moment when the mouse lost its grip was recorded as the peak tension. Test was repeated three times with a 3-min interval between trials. Each trial consisted in five attempts in quick successions for which the best value was recorded therefore increasing the chances that the measure will accurately reflect maximum strength. The mean of three trials and the largest value from all trials were used as parameters.

Sensory reflexes
Sensory abilities were evaluated through the reflex response to several sensory modalities using the following scales. Pinna reflex in response to a gentle touch of the auditory meatus with a cotton-tipped applicator repeated three times with a 10-to 15-s interval: 0 ϭ none, 1 ϭ active retraction, moderately brisk flick, 2 ϭ hyperactive, repetitive flick. Corneal reflex in response to a gentle puff of air repeated three times with a 10-to 15-s interval: 0 ϭ no eye blink, 1 ϭ active eye blink, 2 ϭ multiple eye blink. Toe pinch normal retraction reflexes in all four limbs when lightly pinching each paw successively by applying a gentle lateral compression with fine forceps while the mouse is lifted by its tail so the hind limbs are clear of the table. Score is cumulative of four limbs: 0 ϭ no retraction, 1 ϭ active retraction, 2 ϭ repetitive retractions. Preyer reflex in response to a 90-dB click 30 cm above mouse repeated three times with a 10-to 15-s interval: 0 ϭ None, 1 ϭ Preyer reflex (head twitch), 2 ϭ jump Ͻ1 cm, 3 ϭ jump Ͼ1 cm.

Tail flick test
The automated tail flick test (Omnitech Electronics Inc) was used to assess nociceptive threshold. Awake mice were placed in a contention tube to limit movement with their tail resting on the groove of a heating panel. When the mice were calm, a narrow heat producing beam was directed at a small discrete spot ϳ15 mm from the tip of the tail. When the subject's tail was removed from the beam, an automatic timer recorded the latency. The test was repeated five times with a 3-min interval between each trial. The latency of the mice to flick their tail was recorded and the two trials with the shorter latencies were discarded since the tail is not always fully in the beam and this is often an outlier.

Acoustic startle response and pre-pulse inhibition of startle
Subjects were placed in isolation boxes outfitted with accelerometers to measure magnitude of subject movement (Med Associates). After 5 min of acclimation mice were first tested for acoustic startle response. Mice were presented with six discrete blocks of six trials over 8 min, for a total of thirty-six trials. The trials consisted in six responses to no stimulus (baseline movement), six responses to 40-ms sound bursts of 74 dB, six responses to 40-ms sound bursts of 78 dB, six responses to 82-ms sound bursts of 100 dB, five responses to 40-ms sound bursts of 86 dB, and six responses to 40-ms sound bursts of 92 dB. The six trials type were presented in pseudorandom order such that each trial type was presented once within a block of six trials. Mice were then tested for pre-pulse inhibition of startle. They were presented with seven discrete blocks of trials of six trials over 10.5 min for a total of 42 trials. The trials consisted in six response to no stimulus (baseline movement), six startle response to a 40-ms, 110-dB sound burst, six prepulse inhibition trials where the 110-dB tone was preceded by a 20-ms 74-dB tone 100 ms earlier, six prepulse inhibition trials where the 110-dB tone was preceded by a 20-ms 78-dB tone 100 ms earlier, six prepulse inhibition trials where the 110-dB tone was preceded by a 20-ms 82-dB tone 100 ms earlier, six prepulse inhibition trials where the 110-dB tone was preceded by a 20-ms 86-dB tone 100 ms earlier and six prepulse inhibition trials where the 110-dB tone was preceded by a 20-ms 92-dB tone 100 ms earlier. The seven trial types were presented in pseudorandom order such that each trial type was presented once within a block of seven trials. Startle amplitude was measured every 1 ms over a 65-ms period, beginning at the onset of the startle stimulus. The intertrial interval was 10 -20 s. The maximum startle amplitude over this sampling period was taken as the dependent variable. A background noise level of 70 dB was maintained over the duration of the test session.

Visual acuity
Visual acuity was tested using the visual placing test that takes advantage of the forepaw-reaching reflex: the mouse was held by its tail ϳ20 cm above the surface and progressively lowered. As it approaches the surface, the mouse should expand its forepaws to reach the floor. The test was repeated three times with a 30-s interval and the forepaw reaching reflex was quantified as the percentage of forepaw-reaching episodes that did not involve the vibrissae and/or nose touching the surface before the forepaws.

Buried food test
The buried food test (Yang and Crawley, 2009) measures how quickly an overnight-fasted animal can find a small piece of familiar palatable food, that is hidden underneath a layer of bedding using olfactory clues. Fruit Loops (Kellog's) were used as familiar food. For three consecutive days before the test, three to four pieces were offered to the subjects to make sure it was highly palatable for all the subjects. At 18 -24 h before the test, all chow pellets were removed from the subjects' home cages. The water bottle was not removed. On the testing day, the subject was placed in a clean cage (28 cm long ϫ 18 cm wide ϫ 12 cm high) containing 3 cm deep of clean bedding and the subject was allowed to acclimate to the cage for 10 min. While the subject was temporary placed in an empty clean cage, four to five pieces of Fruit Loops were buried ϳ1 cm beneath the surface of the bedding, in a random corner of the cage and the bedding surface was smoothed out. The subject was placed back in the testing cage and given 15 min to retrieve and eat the hidden food. Latency to find the food was recorded. If a subject did not find the food, 15 min was recorded as its latency score and the food was unburied and presented to the mouse by the experimenter to make sure that it was palatable for the mouse. At the end of testing, subjects were hold in a temporary cage until all animals from the same home cage were tested.

Olfactory habituation and dishabituation
This test consisted of sequential presentations of different nonsocial and social odors in the following order: water, lemon extract (McCormick; 1:100 dilution), banana extract (McCormick; 1:100 dilution), unfamiliar males and unfamiliar females (Yang and Crawley, 2009). Lemon and banana solutions were freshly prepared everyday using distilled water. Social odors were obtained from cages of unfamiliar C56BL/6 mice of the same and opposite sex as the subject which have not been changed for at least 3 d and were maintained outside of the experimental testing room. Social odor stimuli were prepared by wiping a cotton swab in a zigzag motion across the cage. The subject was placed in a clean bedding-covered testing cage covered with the cage grid. A clean dry applicator (10-cm cotton swab) was inserted through the cage grid water bottle hole and the animal was allowed to acclimate for 30 min to reduce novelty-induced exploratory activity during the olfaction test. Each odor (or water) was presented in three consecutive trials for a duration of 2 min. The intertrial interval was 1 min, which is about the amount of time needed to change the odor stimulus. At the end of testing, subjects were hold in a temporary cage until all animals from the same home cage were tested. The test was videotaped and subsequently scored. Sniffing and direct interaction time (touching, biting, climbing the applicator) were quantified separately.

Social tests Three-chambered social approach test
Sociability and preference for social novelty and social recognition were tested in a three-chambered apparatus (Nadler et al., 2004). The subject mouse was first placed in the central, neutral chamber and allowed to explore for 10 min with all doors closed. Next, doors were opened and the mouse was allowed to freely explore the three empty chambers for an additional 10 min. Lack of side preference was confirmed during this habituation. The subject was then temporary placed in a holding cage while two empty wire cages which allow for olfactory, visual, auditory, and tactile contacts but not for sexual contact or fighting containing either an inanimate object (black cone) or a male mouse were placed in each of the testing chambers and the subject was returned to the apparatus for a 10-min testing phase. Adult mice from the same strain that was previously habituated to the wire cup and did not exhibit aggressive behaviors but had no previous contact with the subject were used for unfamiliar mice. Unfamiliar mice were not used more than twice a day with at least 2 h before two tests. At the end of testing, subjects were hold in a temporary cage until all animals from the same home cage have been tested. The side position of the interacting animal and the object was randomly determined. All the sessions were videotracked (Noldus Ethovision) and the amount of time spent in each chamber, close to the holding cages or in direct interaction with the holding cage was automatically calculated.

Male-female social interaction
Male-female social interactions were evaluated in in a regular clean cage during a 10-min test session as previously described (Scattoni et al., 2011). Each subject male was paired with an unfamiliar estrus C57BL/6J female under low light (10 lux) conditions. A total of 20 females were used for this test allowing to avoid to reuse the same female more than twice on the same day. The sessions were videotaped and ultrasonic vocalizations were recorded using an ultrasonic microphone with a 250-kHz sampling rate (Noldus Ultravox XT) positioned 10 cm above the cage. The entire set-up was installed in a sound-attenuating room. Videos from the male subjects were subsequently manually scored to quantify (number of events and total time of male to female nose-to-nose sniffing, nose-to-anogenital sniffing, and sniffing of other body regions. Ultrasonic vocalizations were played back and spectrograms were displayed using the Ultravox XT software and ultrasonic vocalizations were manually quantified.

Social transmission of food preference
The social transmission of food preference is a test of olfaction memory that involves a social component through the use of a demonstrator mouse (Wrenn et al., 2003). The demonstrator mouse is a conspecific mouse of same sex and similar age that was labeled by bleaching before testing. To minimize neophobia during the experiments, both subjects and demonstrator mice were habituated to eat powdered rodent chow (AIN-93M, Dyets, Inc.) from 4-oz (113.40-g) glass food jar assemblies (Dyets, Inc.). This habituation was performed for 48 h in the mice home cage while the regular pellet chow was removed from the cages. After the habituation, both subject mice and demonstrator mice were food deprived for 18 -24 h before testing with free access to water. The test was divided into three phases.
Demonstrator exposition During the first phase the demonstrator was presented with a jar of powder food mixed with either 1% cinnamon or 2% cocoa. The flavor was randomly assigned to the demonstrators so half of them received the cocoa flavored food while the other half received the cinnamon flavored food. Each demonstrator was used only once a day. The demonstrators were allowed to eat the flavored food for 1 h. The jars were weighed before and after presentation to the demonstrators. The criterion for inclusion in the experiment was consumption of 0.2 g or more.
Interaction phase After eating the flavored food, a demonstrator was placed in an interaction cage with the observer subject mouse and mice were allowed to freely interact for 30 min.
Choice phase Immediately after the interaction phase, the observer mouse was placed in a clean cage and presented with one jar containing the flavor of food eaten by the demonstrator (cued) and another jar containing the other flavor and given 1 h to freely explore the jar and eat. The demonstrator flavor and the position of the jar (front or back of the cage) was randomly assigned.
All phases were videotaped and food jars were weighed before and after the sessions to determine the amount of food eaten. At the end of testing, demonstrators and observers were hold in temporary cages until all animals from the same home cage had been tested. Video recordings from the interaction phase were used to score the number and total time of sniffing bouts from the observer to the nose or head of the demonstrator. Video recordings from the choice phase were used to score the total time spent in interaction with each food jar (mouse observed in the top of the jar with nose in jar hole).

Avoidance, escape behavior, and hyper-reactivity
Object avoidance and escape behavior was observed in several tests initially designed to assess other behaviors, including the novel object recognition, the marble burying, and the nest building.

Novel object recognition
The novel object test for object recognition and memory takes place in an opacified open field arena (45 ϫ 45 cm). The test involves a set of two unique novel objects, each about the size of a mouse, constructed from two different materials and nonuniform in shape. The test consisted of one 10-min habituation session, a 5-min familiarization session and a 5-min recognition test, each videotracked (Noldus Ethovision). During the habituation, animals were allowed to freely explore an empty open field. At the end of the session, they were removed from the open field and place in a temporary clean holding cage for about 2 min. Two identical objects were placed on the median line at ϳ10 cm from each wall and the animal was returned to the open field and allowed to explore the objects for 5 min before being returned to its home cage. After 1 h, one familiar object and one novel object were placed in the open field to the location where the identical objects were placed during the familiarization session and the mouse was allowed to explore them for a 5-min recognition test. The side of the novel object position was randomly assigned so half of the animals were exposed to a novel object placed on the right of the open field and half of the animals were exposed to a novel object placed on the left of the open field.
Between each session, the open field and the objects were carefully cleaned with 70% ethanol and let dry. Familiarization and recognition sessions were scored for total time spent investigating each object, the number of object interactions and the latency o the first object interaction. Time spend in each side during habituation and familiarization and time spent sniffing two identical objects during the familiarization phase were used to examine an innate side bias. Total time spent sniffing both objects was used as a measure of general exploration.

Marble burying test
The marble-burying assay is a tool for assessing either anxiety-like and/or repetitive-like behaviors in mice (Thomas et al., 2009). Subjects were tested in a regular clean cage (28 cm long ϫ 18 cm wide ϫ 12 cm high) with 3 cm of fresh bedding. The subject was first placed in the empty cage for a 5-min habituation. It was then temporary placed in an empty clean cage while 20 dark blue glass marbles (15 mm in diameter) were positioned over the bedding equidistant in a 4 ϫ 5 arrangement to cover the whole cage surface. The subject was then returned in the test cage and allowed to explore and bury the marbles during a 15-min session that was videotaped. At the end of the session the subject was removed and the number of marbles buried (Ͼ50% marble covered by bedding material) was recorded.

Nest building
For small rodents, nests are important for heat conservation as well as for reproduction and shelter (Deacon, 2006). Mice were initially single housed in cages containing no environmental enrichment items such as bedding, cardboard houses or tunnels. To test their ability to build nests animals were temporarily single housed. One hour before the dark phase, any building material present in the home cage was removed and replaced by two cotton nestlets (Ancare, NES3600 nestlets). The test was repeated twice and scored on the next morning of the second repeat using the following multicriteria scale adapted from (Deacon, 2006; maximum score ϭ 11): nestlet shredding: 0 ϭ not at all, 1 ϭ partially, 2 ϭ fully shredded; nestlet dispersion: 0 ϭ nestlet dispersed all over the cage, 1 ϭ mostly used to build nest, 2 ϭ fully used to build a nest; nest density: 0: not dense, 1 ϭ medium density, 2 ϭ high density; nest shape: 0: no nest, 1 ϭ ball shape, 2 ϭ nest shape but no bottom, 3 ϭ full nest; presence of walls: 0 ϭ no walls, 1 ϭ partial walls, 2 ϭ nest fully surrounded by walls; maximum score ϭ 11.

Escape behavior
Escape behavior evaluated in three different tests all taking place in regular home cages (28 cm long ϫ 18 cm wide ϫ 12 cm high) by counting the number of unsuccessful (mouse climbing on cage walls) or successful (mice jumping out of the cage) attempts. The three tests, selected for their increasing anxiogenic properties, were the habituation phase of the buried food test (first test in the home cage set-up, no object at the surface of the bedding), the repetitive novel object contact task (four objects visible at the surface of the bedding) and the marble burying test (20 objects visible at the surface of the bedding). Each test was scored for 10 min.

Hyper-reactivity
Hyper-reactivity was recorded by looking at touch escape response, positional passivity, trunk curl and catalepsy during the handling of the mice using the following scales. Touch escape to cotton-tipped applicator stroke from above starting light and slowly getting firmer recorded over five trials: 0 ϭ no response, 1 ϭ mild (escape response to firm stroke), 2 ϭ moderate (rapid response to light stroke), 3 ϭ vigorous (escape response to approach). Positional passivity or struggle response to sequential handling: 0 ϭ struggles when restrained by tail, 1 ϭ struggles when restrained by neck (finger grip, not scruffed), 2 ϭ struggles when held supine (on back), 3 ϭ struggles when restrained by hind legs, 4 ϭ does not struggle. Trunk curl: 0 ϭ absent, 1 ϭ present. Catalepsy when subject front paws are positioned on a rod elevated 3 cm from floor, the amount of time the animal stayed immobile and kept its paws on rod was recorded, with a maximum cutoff of 120 s over three trials separated by 30 s. Hyper-reactivity was also observed in other tests such as the beam walking tests or the negative geotaxis test.

Repetitive novel object contact task
This novel object investigation task looks for specific unfamiliar objects preference as well as patterned sequences of sequential investigations of those items (Pearson et al., 2011;Steinbach et al., 2016). Subjects were tested in a regular clean cage (28 cm long ϫ 18 cm wide ϫ 12 cm high) with 1 cm of fresh bedding. The subject was first placed in the empty cage for a 20-min habituation. It was then temporary placed in an empty clean cage while four unfamiliar objects (a Lego piece, 3 cm in length; a jack, 4 cm in length; a dice, 1.5 cm in length; and a bowling pin, 3.5 cm in length) were place in the cage's corners at ϳ3 cm from the edges. The subject was then able to investigate the environment and objects during a 10-min session that was videotaped. The videos were manually scored for the occurrence of investigation of each of the four toys. Investigation was defined as clear facial or vibrissae contact with objects or burying of the objects. The number of contacts and the cumulative contact time was evaluated for each object. to determine if there was a genotype effect on the tendency to display preferences for particular toys, the frequencies of contact with each object were ranked in decreasing order from maximum to minimum preference for each subject and the frequencies were averaged by group and compared. To assess the pattern of object investigation, each specific toy was given an arbitrary number (1-4) and all possible three-digit and four-digit combinations without repeat numbers were identified. For both three-and fourobject sequences the total number of choice, the number of unique sequences, and the number of choices of the three most repeated sequence was calculated for each subject as described in (Steinbach et al., 2016). To take in account the overall mouse activity, the percentage of top, top two, and top three preferred choices over the total number of choices were also calculated.

Barnes maze
The Barnes maze is a test of spatial memory comparable to a dry version of the Morris water maze (Barnes, 1979). In this assay, mice use spatial memory and navigation skills to orient themselves thanks to extra-maze cues placed in the test room, with the goal of locating one of 20 identical holes evenly spaced around the edge of a brightly-lit 100 cm in diameter circular arena (Maze Engineers). While most of the holes (nontarget) have nothing beneath them and lead nowhere, the target escape hole leads to shelter in a desirably darkened and enclosed goal box below the table. Two days before the beginning of the training, habituation was performed by allowing each subject to freely explore the arena (without escape box) under modest light for 5 min. At the end of the second habituation, subjects were pre-trained to learn of the presence of the escape hole by placing them for 1 min in a clear box in the middle of the arena under bright light conditions. After 1 min, the box was lifted up and the subject was gently guided near the escape hole selected randomly on the table, allowing it to enter the hole and remain inside for 1 min. For the initial training, animals were trained for 4 d to locate the escape box (in a position different from the pre-training). All trials began with the subject in a clear box in the center of the table. The trial started when the box was lifted up. If the subject located and entered the escape box within 3 min, it was left in the box for 1 min. If the subject failed to find the escape box within 3 min, it was gently guided to near the escape hole, and allowed to stay in the box for 1 min. Animals received four trials per day with an intertrial interval of 20 min for 4 d. After each trial, the maze and the escape box were cleaned using cleaning wipes to remove odors and fresh bedding was placed in the escape box. On the fifth day, animals were tested for 3 min without the escape box for a probe test.
Time spent in the different quadrants was recorded. For the reversal training, the escape hole was moved to the opposite position on the maze and animals received four additional days of training followed by a reversal probe test on the fifth day. All trials were recorded by overhead camera (Noldus Ethovision) and scored for distance and latency to find escape box.

Y-maze test
Y-maze alternation is a test of working memory based on the natural tendency of mice to explore new territory whenever possible. Mice were placed in the center of a Y-maze (three 5-cm-wide and 50-cm-long arms, each set 130°from each other) and given 15 min to freely explore the three arms of the maze. The number of arm entries and the number of triads were recorded to calculate the percentage of alternation. An entry occurs when all four limbs are within the arm. A successful score is defined by three successive choices that includes one instance of each arm by the total number of opportunities for alternation. A type 1 error is determined by three consecutive choices where the first and third choices are identical. A type 2 error is defined by three consecutive choices where the second and third choices are identical. Perseverance is defined as three or more repetitive entries in the same arm.

Contextual and cued fear conditioning
To isolate the effects of cued and contextual fear conditioning, a 3-d assay was employed. During the training session, the mice were placed in an ethanol cleaned contextual box with a bar floor, black and white striped walls in which all movements can be recorded (Med Associate fear conditioning boxes coupled with Noldus Ethovision for control an analysis) and given 5 min to habituate. Movements were then recorded for 540 s. At 120, 260, and 400 s after the beginning of the recording, the mice were exposed to a 20-s tone (80 dB, 2 kHz) and coterminating shock (1 s, 0.7 mA). Twenty-four hours after the training phase, the animals were tested for contextual memory in the identical enclosure and movements were recorded for 240 s to assess the ability of the animal to remember the context in which the shocks had occurred the previous day. Forty-eight hours after the training phase, the animals were tested for cued memory in a different context (isopropanol cleaned, white wall insert over a mesh grid floor). They were recorded for 330 s and were presented with the identical tone from the training session at 120 s, and 260 s after the beginning of the recording session to assess the ability of each animal to remember the tone and pair it with the shock from training session. The three sessions were recorded using a camera located on the side of the boxes. Freezing, defined as lack of movement except for respiration, was scored using Noldus Ethovision software during each phase.

Elevated zero-maze
Fear and anxiety were tested in an elevated zero-maze. The apparatus consisted of a circular black Plexiglas runway, 5 cm wide, 60 cm in diameter, and raised 60 cm off the ground (Maze Engineers). The runway was divided equally into four alternating quadrants of open arcs, enclosed only by a 1 cm inch lip, and closed arcs, with 25-cm walls. All subjects received one 5-min trial on two consecutive days starting in the center of a closed arm and were recorded by video tracking (Noldus Ethovision).
Measures of cumulative open and closed arc times, latency to enter an open arc for the first time (for trials with a closed arc start), total open arm entries, latency to completely cross an open arc for the first time (for trials with a closed arc start) between two closed arcs, closed arc dipping (body in closed arc, head in open arc), open arc dipping (body in open arc, head outside of the maze) were calculated using the mean of the two trials.

Open field
The vertical activity in the open field was scored by counting the numbers of wall rears (while touching a side of the open field) and free-standing rears. The thigmotaxis was measured by quantifying the amount of time or distance traveled on the side of the open field compared to the center of the open field.

Statistical analyses
Shank3 ⌬4-22 wild-type, heterozygous, and knock-out littermates were compared for each parameter. Statistical analyses were performed with SPSS 23.0 software using different types of ANOVA with or without repeated time measures with genotype as independent variable followed by Tukey pair-wise comparisons and correction for multiple comparisons if needed or equivalent nonparametric tests when required. Newborn developmental milestones were analyzed by two-way ANCOVA using genotype and gender as between-subject factors and litter number as co-variate to take in account possible gender and litter effects. As we did not observe a gender effect, males and females were grouped together in figures and tables. to account for possible cohort effects, cohorts 3 and 4 were analyzed either together using two-way ANOVA with genotype and cohort as betweensubject factors or separately using ANOVA or Kruskal-Wallis tests. Figures represent results for both cohorts analyzed together. Each cohort data and all statistical results including cohort effects are reported in tables and corresponding extended data tables. In tests comparing activity in two or more locations (open field thigmotaxis, social preference test, social transmission of food preference, novel object recognition, zero-maze) genotype ϫ zone interactions were assessed using repeated measures. When sphericity was found violated, the Greenhouse-Geisser values were reported. The distribution of the genotypes was compared to Mendelian expectation using Pearson's 2 test, the survival curves were analyzed using survival Kaplan-Meyer 2 . The comparison to chance level was evaluated using either one-sample t test or Wilcoxon test. Normality was assessed using data visualization and Shapiro-Wilk test. All values are expressed as mean Ϯ SEM.

Results
Generation of a Shank3 ⌬4-22 mouse with a complete deletion of the Shank3 gene A mouse line with a complete disruption of the Shank3 gene was generated by retargeting ES cells previously used to disrupt exons 4 through 9 (Bozdagi et al., 2010). To do this, an additional loxP site was inserted directly after exon 22 while leaving intact the two existing loxP sites flanking exons 4 and 9 (Fig. 1A). To generate the Shank3 ⌬4-22 mouse line used in the present study, the floxed allele was then excised by breeding with a CMV-Cre transgenic line resulting in a deletion of exons 4-22 and therefore a constitutive disruption of all the Shank3 murine isoforms.
Immunoblot analyses using antibodies which crossreact either with an epitope in the SH3 domain (antibody N367/62; Fig. 1B, left panel) or with the COOH terminal (antibody H1160, Fig. 1B, right panel) showed no expression of Shank3 protein in post synaptic density fractions from Shank3 ⌬4-22 homozygous mice and reduced expression consistent with haploinsuficiency in the heterozygotes. As in humans, in mice, the Shank3 gene has 22 exons, spans ϳ58 kb of genomic DNA, and undergoes complex transcriptional regulation controlled by a combination of five intragenic promoters and extensive alternative splicing resulting in in a complex pattern of mRNA and protein isoforms Kouser et al., 2013;Waga et al., 2014;Speed et al., 2015). The loss of all known major Shank3 mRNA isoforms was confirmed by RT-PCR (Fig. 1C).
The Shank3 ⌬4-22 mouse line was maintained on a C57BL/6 background by heterozygote ϫ heterozygote mating, allowing for the production of all genotypes (wildtype, heterozygous, and homozygous) as littermates. Shank3 ⌬4-22 heterozygous and homozygous animals were viable, however abnormal Mendelian ratios were observed at the time of weaning, with a significant deficit for Shank3 ⌬4-22 knock-out mice ( Fig. 1D; Table 3). Adult survival curves between 1 and 22 months did not show a significant genotype difference with the current sample size, but there was evidence for higher numbers of deaths in Shank3 ⌬4-22 homozygous mice between 18 and 22 months ( Fig. 1E; Table 3). Although the human clinical SHANK3 mutation is hemizygous, for completeness, we have conducted our studies in Shank3-null mutant mice (homozygous knock-out, KO), along with their heterozygous (Het) and wild-type (WT) littermates. The KO mice are instrumental to understand the function of Shank3, while the Het mice have significantly greater construct validity for PMS, a haploinsufficiency syndrome. To ensure the robustness of behavioral abnormalities in the adult mice, two cohorts representing all three genotypes were compared. All the cohorts used in the present study are described in Table 2.

Developmental milestones in Shank3 ⌬4-22 neonates
Ten litters were used to study developmental milestones. The average litter size was 7.2 pups (ranging from five to nine), with 54 surviving passed postnatal day 2 (28 males and 26 females). As very limited gender effects were observed (for detailed analysis, see Table 4), males and females were analyzed together using both genotype and gender as fixed factors and the litter number as a covariate.
Developmental delays were observed in the Shank3 ⌬4-22 homozygote neonates in several of the parameters studied ( Fig Table 4). While the birth weight was not significantly different, the growth rate of Shank3 ⌬4-22 homozygote pups was slower and by P14, the weight of Shank3⌬4-22 homozygous mice was significantly lower than the weight of their wild-type littermates ( Fig. 2A). Additionally, an unusual postnatal mortality was observed when breeding heterozygous animals together, with 6.9% of the pups dying between birth and P1. Eighty-six dead pups were genotyped, showing that the percentage of Shank3 ⌬4-22 homozygote knock-out mice dying at or shortly after birth was higher than expected if the death was equally affecting all the genotypes (WT: n ϭ 20, Het ϭ 33, KO: n ϭ 33, 2 df2 ϭ 8.66, p ϭ 0.0137), this could explain, at least partially, the deficit observed at weaning. No differences were observed in any of the other physical developmental milestones, including eye opening, ear opening, tooth eruption or fur development (Extended Data Fig. 2-1A-D; Table 4).
A significant delay was observed for Shank3 ⌬4-22 homozygotes in the response to auditory startle ( Fig. 2B) and in the mid-air righting task ( Fig. 2C) although all the mice were able to properly respond at the end of the observation period. In the wire suspension ( Fig. 2D) and grasping reflex ( Fig. 2E) tasks, however, not only was the acquisition of the response delayed, but Shank3 ⌬4-22 homozygous animals remained significantly impaired until the time of weaning. In the negative geotaxis test, an initial delay was observed at P5 were most wild-type animals were able to turn while homozygous and heterozygous Shank3 ⌬4-22 animals were still falling or staying in the starting position (Fig. 2F). Moreover, after P9 when most of the animals were able to master the task, higher reactivity (characterized by a shorter latency to turn) was observed for the Shank3 ⌬4-22 homozygous mice. The acquisition of the rooting reflex was similar for the three groups; however, a premature disappearance of the reflex was observed in both the Shank3 ⌬4-22 heterozygous and homozygous pups (Extended Data Fig. 2-1E; Table 4).
Other sensory-motor and neurologic milestones such as cliff aversion, ear twitch, surface righting, negative  Table 4) were not significantly affected by the disruption of the Shank3 gene.
Ultrasonic vocalizations were recorded at postnatal day 6 on an independent cohort of mice and a genotype difference was detected in the number and quality of ultrasonic vocalizations emitted by the pups (Table 4). Shank3 ⌬4-22 heterozygous and homozygous mice emitted fewer ultrasonic vocalizations than wild-type littermates (Extended Data Fig. 2-1K; Table 4). The total calling time was also affected with Shank3 ⌬4-22 -deficient mice both spending less time calling and having shorter calls than wild-type littermates. Additionally, the peak amplitude was shorter in Shank3 ⌬4-22 -deficient mice. However, none of these parameters were significantly different, probably due to a high interindividual variability within each group with some animals emitting no vocalizations during the 3-min recording. The percentage of noncallers was higher, although not significantly, in Shank3 ⌬4-22 -deficient animals. Genotype did not affect the latency to the first call or the peak frequency of calls and no difference was observed in the time course of the emission of ultrasonic vocalizations.

Adult general health in Shank3 ⌬4-22 -deficient mice
Adult Shank3 ⌬4-22 mice were evaluated for general health at three months of age (Table 5). The three genotypes did not differ on physical measure of weight and length. Additional weight measures at the age of 15 and 20 months showed a trend in reduced weight of Shank3 ⌬4-22 homozygous mice compared to their littermates. Genotypes scored similarly and in the normal range for other physical characteristics including coat appearance (grooming, piloerection, patches of missing fur on face or body), skin pigmentation, whisker appearance, wounding, and palpebral closure. Observation in a beaker or after transfer to a housing cage revealed no abnormalities in term of spontaneous general activity, stereotypies (rears, jumps, circling, wild running), transfer arousal, gait, pelvic, and tail elevation.

Motor functions in Shank3 ⌬4-22 -deficient mice
Motor functions were examined using several different paradigms (Table 6). Footprint gait analysis showed normal stance and sway but increased stride in Shank3 ⌬4-22 homozygous mice compared to wild-type and heterozy-  Ϫ  Ϫ Nonnormal 0.38 Ϯ 0.14 0.06 Ϯ 0.04 0.     Ϫ         gous animals (Fig. 3A) and reduced spontaneous locomotion was observed during a 1-h open field session in both Shank3 ⌬4-22 heterozygous and homozygous mice (Fig.  3B). Across the 60-min session, the time course for total distance traversed by all three genotypes declined as expected, representing habituation to the open field. However, while the distance traveled during the first 10 min was similar for the three groups, the decline was faster for Shank3 ⌬4-22 homozygous mice, possibly reflecting a higher fatigability. Similarly, in the accelerating rotarod test, which assay for gait, balance, motor coordination and endurance, shorter latencies to fall where observed in Shank3 ⌬4-22 -deficient mice after the first trial, with a milder phenotype observed in the heterozygotes compared to homozygotes. When examining learning in this paradigm, characterized by an improvement of performance (latency to fall) over the trials, Shank3 ⌬4-22 heterozygous and homozygous animals failed to improve over time, in contrast to wild-type animals which showed typical learning (Fig. 3C). Impairment of motor coordination and balance was also observed in Shank3 ⌬4-22 homozygous in the beam walking test (Fig. 3D; Table 6) as well by reduced strength and endurance in both the inverted screen and hanging tests (Fig. 3E), but with no differences in forelimb grip strength (Extended Data Fig. 3-1A). There was also a trend toward an increased number of failed attempt in the hindlimb placing for Shank3 ⌬4-22 homozygous mice, compared to their littermates (Extended Data Fig. 3-1B).

Sensory abilities in Shank3 ⌬4-22 -deficient mice
For all sensory-related assays, detailed results are reported in Table 7.
No genotype differences were detected in tactile tests including the pinna reflex, the palpebral reflex, and the toe pinch retraction test. In the tail flick pain sensitivity test, a trend toward a decreased latency to flick the tail in re-sponse to a noxious thermal stimulation a was observed in Shank3 ⌬4-22 homozygous animals (Fig. 4A).
Normal Preyer reflexes were observed in all genotypes; however, Shank3 ⌬4-22 heterozygous and homozygous mice showed a reduced startle response throughout all the sound intensities (74 -92 dB, analyzed as repeated measures) indicating an impaired sound discrimination (Fig. 4B). Changes in pre-pulse inhibition of acoustic startle in Shank3 ⌬4-22 -deficient mice are consistent with abnormalities in auditory processing, rather than sensorimotor gating deficits (Extended Data Fig. 4-1A).
Normal visual placing/reaching reflexes were observed for all the mice, thus ruling out strong visual impairments (Fig. 4C).
Shank3 ⌬4-22 homozygous mice demonstrated strong deficits in the buried food test (Fig. 4D, left panel) with only seven out of 19 mice able to retrieve the food in less than 2 min and nine out of 19 mice not being able to find the food at all (Extended Data Fig. 4-1B). However, all animals showed interest for the food and ate it when it was made visible. To further investigate olfactory function, animals were subjected to the olfactory habituation/dishabituation paradigm using three nonsocial scents (water, banana, and lemon) and two social scents (unfamiliar males and unfamiliar females). Wild-type and Shank3 ⌬4-22 heterozygous animals displayed a normal response, characterized by a robust sniffing elicited by the first scent presentation of each nonsocial and social scent that declined over the second and third presentation of the same scent. In contrast, Shank3 ⌬4-22 homozygous animals had little response to any of the nonsocial scents, even on their first presentation (Fig.  4D, middle panel), thus confirming the results of the buried food test. Interestingly the lack of interest for olfactory stimuli does not appear to be the consequence of anosmia as a normal response to both social scents was observed in Shank3 ⌬4-22 homozygous mice (Fig. 4D, right panel).  5A, left panel). A significant increase in latency for the first event of anogenital sniffing was found in male Shank3 ⌬4-22 homozygous mice (Fig. 5A, right panel), and we can note that this latency may contribute to trend toward reduced anogenital sniffing time in those animals. Ultrasonic vocalizations did not show significant difference across genotypes (Extended Data Fig. 5-1A).
Similarly, In the three-chambered test for social preference, sociability, defined as spending more time interacting with the mouse than with the object, was found in all genotypes. Hence, in all groups, significantly more time was spent in the chamber containing the novel mouse than in the chamber containing the novel object, and more time was spent sniffing the novel mouse than the novel object (Fig. 5B). All genotypes showed the normal absence of innate chamber side bias during the 10-min habituation phase before the start of the sociability test.
Finally, mice were tested in the social transmission of food preference test that combines social behavior, olfactory recognition and memory skills. A modest decrease of the number of sniffing bouts initiated by the observer mouse toward the demonstrator mouse was observed during the observer-demonstrator interaction phase in Shank3 ⌬4-22 homozygous mice but not in heterozygotes (Extended Data Fig. 5-1B). All genotypes showed a strong preference for the cued food flavor that was exposed to them through the demonstrator, as compared to the noncued food flavor, as shown both by significantly more time spent interacting with the jar containing the cued food than the noncued food (Fig. 5C) or by eating significantly more cued food than noncued food during the choice phase ( Table 8). Note that two flavors were randomly used as cued and noncued food flavor and all genotypes showed an absence of flavor preference. However, the total amount of food (cued and noncued) eaten by Shank3 ⌬4-22 homozygous mice was significantly lower than the total amount of food eaten by their wild-type and homozygous littermates.

Object avoidance in Shank ⌬4-22 -deficient mice
While testing mice in different set-ups involving object interactions, a strong avoidance toward inanimate objects was observed in Shank3 ⌬4-22 homozygous mice (Table 9).
This avoidance behavior was initially observed in the novel object recognition task. This highly validated test for recognition memory is designed to evaluate differences in the exploration time of novel and familiar objects. Mice      are expected to spend more time investigating a novel object than a familiar object, and this is what was observed for wild-type and heterozygous mice (Fig. 6A, left panel). However, in homozygous mice, results were difficult to interpret due to strikingly reduced object interactions (Fig. 6A, left and middle panels). Homozygous mice spent most of both of the test sessions (the first involving familiarization with identical objects and the second involving interaction with one familiar and one novel object) away from both objects, spending excessive time in the corners of the open field as shown on heatmaps (Extended Data Fig. 6-1A) and demon-strating longer latency to explore any of the objects (Fig. 6A, right panel). Object avoidance was further confirmed in multiple independent tests, including the marble burying, a test used to assess stereotypic behavior and/or anxiety. In this paradigm, 20 marbles were spread across the cage floor in a 4 ϫ 5 pattern, leaving little space for the mice to move around the marbles. While both wild-type and Shank3 ⌬4-22 heterozygous mice quickly buried most of the marbles as is typical, Shank3 ⌬4-22 homozygous mice left the marbles almost completely undisturbed for the whole 15-min duration of the test (Fig. 6B; Extended Data Fig. 6-1B).  Consistent with these result, a significant decrease of the time spent exploring objects in the four-object exploration test was observed in the Shank3 ⌬4-22 homozygous mice as compared to their littermate (Fig. 6C).

Jar observation
During assessment of nest building, nests build by Shank3 ⌬4-22 homozygous mice were significantly less elaborate than nests built by wild-type or heterozygous mice, with some homozygotes leaving the nesting material completely untouched ( Fig. 6D; Extended Data Fig.  6-1C). Note that, in an attempt to reduce stress and improve breeding rates, dams used to produce the cohorts described here were provided with plastic huts in their home cage. Interestingly, while most of the wild-type dams (seven out of ten) chose to build their nest inside the huts, only a single Shank3 ⌬4-22 heterozygous dam out of 20 used the hut to establish their nests (wild-type vs heterozygotes: t (28) ϭ -5.085, p Ͻ 0.001). Additionally, three of the Shank3 ⌬4-22 heterozygous dams did not build a nest until after the birth.
Object avoidance might also explain the reduction of the total time of direct interactions (grabbing, touching, biting, or climbing) with the applicator used to present the different scents during the olfactory habituation and dishabituation test in Shank3 ⌬4-22 homozygous mice, compared to their wild-type and heterozygous littermates (Fig.  6E).

22-deficient mice
Unusual hyper-reactivity was observed in Shank3 ⌬4-22 homozygous mice during handling and confirmed in several behavioral tests (Table 10). This hyper-reactivity was characterized by a higher score in the touch escape test (Fig. 7A, left panel), a lower score (reflecting a higher tendency to struggle in response to sequential handling) in the positional passivity (Fig. 7A, middle panel), and a shorter latency to move from the beam during the catalepsy test (Fig. 7A, right panel). As in newborn mice, a shorter latency to turn was seen for Shank3 ⌬4-22 homozygous mice in the negative geotaxis test (Fig. 7B, left  panel). Similarly, in the beam walking test, the latency to start crossing on the smallest beam was shorter in Shank3 ⌬4-22 homozygous mice (Fig. 7B, right panel) but often led to a premature fall (Fig. 3D).
Escape attempts were observed in several tests and high-wall enclosures had to be built around testing cages to prevent successful attempts. Escape behaviors were scored in three different home cage tests. During the habituation portion of the buried food test (where no objects were visible at the surface of the cage bedding), no escape behavior nor genotype differences were observed (Fig. 7C, left panel). However, when the mice were tested in the same cages during the four-object interaction test both the number of escape attempts and the percentage of mice engaged in this behavior increased and significant genotype differences were observed (Fig.  7C, middle panel). This behavior was even more marked in the marble burying test (Fig. 7C, right panel), during which 94% of heterozygous mice and 100% of homozygous mice tried to escape. This indicated that the escape behavior is elicited by the presence of unfamiliar objects in the testing cage.

Repetitive behaviors, stereotypies, and inflexibly in Shank3 ⌬4-22 -deficient mice
Repetitive and restricted behaviors are one of the core features of ASD. Therefore, during all of the behavioral tests, mice were also carefully monitored for stereotypies, as well as perseverative and repetitive behaviors. Detailed results are reported in Table 11. While no genotype difference was observed in the number of spontaneous grooming bouts observed during the 10 first minutes of the open field test, Shank3 ⌬4-22 homozygous mice engaged in longer episodes of selfgrooming, as shown by a significant increase in the cumulative time spent grooming all body regions when compared to their wild-type and heterozygous littermates. However, skin lesions were frequently observed in older mice (over eight-month-old) of all three genotypes without obvious genotype effect. Significantly more rotations were also observed in Shank3 ⌬4-22 homozygous animals as well as a trend toward an increase of head twitching/ shaking in both Shank3 ⌬4-22 heterozygous and homozygous mice, as compared to their wild-type littermates (Fig.  8A).
Object preferences, exploration patterns and frequency of repetitive contacts with novel objects were evaluated in the repetitive novel object contact task. Although the cumulative time spend interacting with the objects was decreased in Shank3 ⌬4-22 homozygous mice (Fig. 6C), this test failed to display genotype difference in either the total number of interactions, the preference for any specific objects or the preference for any specific preferred sequence of threeobject or four-object explorations (Fig. 8B).
continued Spontaneous locomotor activity in the open field was reduced in Shank3 ⌬4-22 homozygous mice relative to other genotypes. C, Latency to fall over six trials (three trials per day for two consecutive days) in the accelerating rotarod task. Motor learning on the accelerating rotarod was deficient in Shank3 ⌬4-22 homozygous mice compared to wild-type animals as they failed to improve over time. Heterozygous mice had an intermediate phenotype. D, Percentage of falls and distance crossed during the beam walking test. While not different on the large (L, 1 inch) and medium (M, ½ inch) beams, Shank3 ⌬4-22 homozygous mice were strongly impaired in the small (S, ¼ inch) beam walking test, as shown by a significant increase of the number of falls and a decrease of the distance crossed. E, Strength and endurance measured in the inverted screen and hanging tests. Endurance strength was significantly impaired in Shank3 ⌬4-22 homozygous mice as they exhibited significantly shorter latency to fall in both the inverted screen and hanging tests. Additional results of motor tests (hindlimb placing and grip strength) are available in Extended Data Figure 3-1    No genotype difference was observed for Preyer reflex, however startle response was decreased in both heterozygous and homozygous Shank3 ⌬4-22 mice compared to their wild-type littermate with genotype differences being more marked for the higher startle intensities. Pre-pulse inhibition results are displayed in Extended Data Figure 4-1A. C, Gross visual function assessed by the visual placing test. Normal visual placing was observed for all genotypes. D, Olfactory abilities evaluated by the time to find hidden food in buried food test and the cumulative time sniffing the applicator without direct interactions during olfactory habituation and dishabituation to nonsocial and social odors. Strong impairments were observed in the buried food test for Shank3 ⌬4-22 homozygous mice as shown by a significant increase in the latency to retrieve the buried food, compared to their heterozygous and wild-type littermates. Individual performances are available in Extended Data Figure 4-1B. Similarly, a significant lack of interest for nonsocial scents (water, banana, and lemon) was observed in Shank3 ⌬4-22 homozygous mice but not in heterozygotes and wild-type during olfactory habituation/dishabituation, while they still Individuals with ASD can maintain rigid habits and frequently show strong insistence on sameness and upset by changes in routine. To examine this domain, Shank3 ⌬4-22 mice were trained for 4 d in the Barnes maze, a test of spatial learning and memory, until all the mice were able to quickly locate an escape box hidden under one of the target locations, then the location of the escape box was moved and mice were tested for reversal learning for four additional days. During the initial learning, all the genotypes were able to find the escape hatch equally well, although Shank3 ⌬4-22 homozygous mice took 1 d longer to reach criteria (Fig. 8C, left panel). All genotypes preferred the correct quadrant in the first probe test ran immediately after the initial training (Fig. 8C, middle panel). When the escape hatch was moved to the opposite side of the maze, both Shank3 ⌬4-22 wild-type and heterozygotes immediately learned the new position, while a 1-d delay was, once again, observed for the Shank3 ⌬4-22 homozygous mice. Genotypes differed markedly in the second probe test, however; while wild-type mice spent most continued displayed normal habituation/dishabituation for social scents (unfamiliar male and female bedding). The olfactory habituation and dishabituation to nonsocial and social odors was measured as cumulative time spent sniffing a sequence of identical and novel odors delivered on cotton swabs inserted into a clean cage. WT, wild-type mice; Het, heterozygous mice; KO, homozygous knock-out mice. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.1, ‫‪p‬ءءء‬ Ͻ 0.001. time in the new target quadrant, Shank3 ⌬4 -22 heterozygous mice split their time 75/25% between new and old targets, whereas Shank3 ⌬4 -22 homozygous animals spent equal time in both targets (Fig. 8C, right panel). This impaired reversal learning implies that Shank3 deficiency increases susceptibility to proactive interference where learning of a previous rule interferes with the new rule.

Learning and memory in Shank3 ⌬4-22 -deficient mice
In addition to the Barnes maze, animals were tested in two additional learning and memory tests, specifically, the Y-maze spontaneous alternation test and the fear conditioning test. Detailed results are reported in Table 12.
When looking at the spontaneous alternation behavior in the Y-maze, no differences were observed between the genotypes in any of the background strains regarding either the total number of choices, the percentage of correct choices or the percentage of errors (Fig. 9A).
Moreover, no arm preference was seen for any of the groups.
In the training session of the fear conditioning test, minimal levels of freezing behavior were seen for all the genotypes during the 5-min habituation period; however, while this percentage of spontaneous freezing decreased before the presentations of cue-shock pairings for the Shank3 ⌬4-22 wild-type and heterozygotes, it remained at significantly higher level for Shank3 ⌬4-22 homozygous mice. A significant genotype effect was then found during the training session in postshock freezing, with Shank3 ⌬4-22 homozygous mice displaying higher levels of freezing compared with wild-type and heterozygous mice (Fig. 9B, left  panel). The opposite was observed during contextual recall where even if all the mice freeze significantly more than during the habituation of the training sessions a trend toward a decrease (significant during the first minute) of freezing was observed for Shank3 ⌬4-22 homozygous mice compared to wild-type or heterozygous littermates (Fig. 9B, middle Figure 5. Social behavior of Shank3 ⌬4-22 -deficient mice. A, Male social interaction in response to the presentation of an unfamiliar conspecific female in estrus and scored by the cumulative sniffing time and latency from the male toward different body regions of the female. No genotype differences were evident in the dyadic male-female social interaction for the overall sniffing time from the male toward the female, however a trend toward a decrease in anogenital sniffing as well as a significant increase of the latency to initiate the first anogenital sniffing event was observed in Shank3 ⌬4-22 homozygous mice. B, Preference for social stimulus in the three-chambered social interaction test measured by cumulative time interacting with either a mouse or an inanimate object. All three genotypes demonstrated a significant preference for an unfamiliar mouse over a nonsocial object. C, Social transmission of food preference measured by the time spent by the test mouse sniffing the demonstrator mouse and the time spent interacting with both cued and noncued food. All genotypes had a strong preference for the food flavor presented by the demonstrator mouse. USVs and time spent sniffing the demonstrator during the demonstrator interaction phase are displayed in Extended Data Figure 5-1. WT, wild-type mice; Het, heterozygous mice; KO, homozygous knock-out mice. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.1, ‫‪p‬ءءء‬ Ͻ 0.001. panel). An increase of freezing was seen in both during and after the cue presentation (trend for the first cue, significant during and after the second cue) Shank3 ⌬4-22 homozygous mice (Fig. 9B, right panel).

Anxiety-related behaviors in Shank3 ⌬4-22 -deficient mice
Anxiety-like behaviors were monitored in the open field and in the elevated zero-maze, and detailed results are displayed in Table 13.
No significant difference between the genotypes was observed in the open field thigmotaxis level (Fig. 10A), but a decrease in the total number of times the mice reared (mainly driven by against wall rears) was observed in the Shank3 ⌬4-22 homozygous animals (Fig. 10B). No significant effects of an interaction between the time and genotype were observed for any of the parameters.
In the elevated zero-maze, all animals showed a preference for the closed arcs versus the open arcs; however, Shank3 ⌬4-22 homozygotes spent less time in the open arcs than their wild-type and heterozygous littermates. Similarly, a significant decrease of the duration of head dipping exploratory behavior in the open arcs was seen in those animals (Fig. 10B). No genotype differences were seen for other parameters.
This indicates increases in anxiety in the Shank3 ⌬4-22 homozygotes. In support of this, the long-lasting spontaneous freezing observed in Shank3 ⌬4-22 homozygous animals during the habituation and before the soundshock association in the fear conditioning training (Fig.  9B) could also be explained by a higher anxiety level those animals.

Discussion
Given the prevalence of complete SHANK3 deletions in PMS, we generated Shank3 ⌬4 -22 mice by targeting exons 4-22, thereby disrupting all isoforms and providing improved construct validity compared to previously reported models. We conducted an extensive behavioral phenotyping of neonatal (P0 -P21) and adult (three to eight months) mice to address both core symptoms and comorbidities observed in PMS. We confirmed our prediction that Shank3 ⌬4 -22 mice homozygous and in some instances heterozygous mice have a more severe phenotype than previously published models with partial deletions of Shank3 (summarized in Fig. 11). Our findings are  The test consisted of a training with two identical objects followed 1 h later by a testing session where one of the object was replaced by a novel object. During the testing session, both wild-type and Shank3 ⌬4-22 heterozygous mice had a strong preference for the novel object over the familiar object, while Shank3 ⌬4-22 homozygous mice failed to display a preference. However, this failure was due to an avoidance of both objects as shown by the strong decrease in object interaction and the increase in latency to explore any of the object for the first time in Shank3 ⌬4-22 homozygous animals, rather than to a real lack of object preference. Representative heatmaps for the three genotypes are available in Extended Data consistent with recent results from an independent model also generated by disrupting all Shank3 isoforms (Wang et al., 2016b). PMS is a neurodevelopmental disorder that manifests as early as in infancy by neonatal hypotonia and a generalized developmental delay. Previous studies have shown normal neonatal development in ⌬4-9 mice (Bozdagi et al., 2010;Wang et al., 2011;Yang et al., 2012) or only minor delays limited to ear opening and paw positioning in ⌬4-22 mice (Wang et al., 2016b). In the current study, both physical and behavioral developmental milestones were investigated. Physical delays were limited to a slower growth rate in Shank3 ⌬4 -22 -deficient animals. In addition, a non-Mendelian genotype distribution showing a deficit for Shank3 ⌬4 -22 homozygous mice was explained, at least partially, by an increased postnatal mortality observed in the Shank3 ⌬4 -22 mice homozygous animals. Similar non-Mendelian genotype distributions have been previously observed in other mouse and rat Shank3 models (Drapeau et al., 2014;Harony-Nicolas et al., 2017). As Shank3 is known to be highly expressed in placenta (Beri et al., 2007), this suggests that Shank3 deficiency could lead to placental insufficiency responsible for in utero developmental delays and increased perinatal mortality. Despite a slower growth curves during the first weeks of life, the weight of surviving homozygous animals is no longer different from their littermates when examined at three months of age, indicating a post birth correction, and survival curves between 2 and 22 months do not show any significant genotype difference.
Extensive sensory-motor deficits were observed in newborn Shank3-deficient mice. Some of them, such as the response to an auditory startle or the air righting ability, were only delayed, while others, such as performances in the wire suspension tests and the grasping reflex, were still present at the time of weaning. On homecage observation and physical examination of adult mice we did not observe severe deficits that would preclude advanced testing.
Hypotonia, motor-coordination impairments and gait abnormalities are a hallmark of PMS that persists beyond continued applicator for all nonsocial scents and for a social male scent but have interaction level similar to wild-type and heterozygous animals when presented with a female scent. WT, wild-type mice; Het, heterozygous mice; KO, homozygous knock-out mice. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.1, ‫‪p‬ءءء‬ Ͻ 0.001.    (Phelan and McDermid, 2012;Soorya et al., 2013). In previous studies, motor performances have been frequently found to be impaired in adult Shank3deficient mice (Fig. 11). Hence, decreased locomotion in the open field has been reported in many existing models including models with ⌬4-9, ⌬13-16, ⌬21 deletions, or point mutations Kouser et al., 2013;Speed et al., 2015;Bidinosti et al., 2016;Mei et al., 2016;Zhou et al., 2016;Copping et al., 2017) even if not always replicated in other models with similar or different deletions (⌬4-9, ⌬9, ⌬13, ⌬13-16, ⌬21; Peça et al., 2011;Drapeau et al., 2014;Duffney et al., 2015;Lee et al., 2015;Jaramillo et al., 2016Jaramillo et al., , 2017. Similarly, motor learning in accelerating rotarod was found to be impaired in ⌬4-9, ⌬11, ⌬13, ⌬13-16, and ⌬21 models (Bozdagi et al., 2010;Wang et al., 2011;Yang et al., 2012;Kouser et al., 2013;Zhu et al., 2014;Speed et al., 2015;Mei et al., 2016;Jaramillo et al., 2017;Vicidomini et al., 2017) although not replicated in other studies (⌬4-9, ⌬13-16, or ⌬2; Peça et al., 2011;Drapeau et al., 2014;Duffney et al., 2015;Bidinosti et al., 2016;Jaramillo et al., 2016;Li et al., 2017). In agreement with Wang et al. (2011), both spontaneous locomotion and rotarod learning were strongly impaired in our Shank3 ⌬4 -22 mouse model. Interestingly, while most models only reported deficits in homozygous animals, heterozygous mice were also affected, albeit less se-  test. Shank3 ⌬4-22 homozygous mice engaged in significantly more self-grooming and rotations relative to the other genotypes. A trend toward an increase amount of head stereotypies was also observed. B, Object preference and pattern of exploration in the repetitive novel object contact task. For each mouse, the time spent interacting with each object was measured and the objects were then ranked from the most (1) to less (4) preferred (left panel). No genotype differences were observed for the proportions of visits to each object. The pattern of object exploration was analyzed by recording specific sequential pattern of visits to three or four specific toys to identify the total number of three-object or four-object sequence investigations, the number of unique sequences, and the percentage of choices of the top, top two, or top three preferred sequences. All groups had identical percentage of their preferred three-object or four-object sequences choices over the total number of sequence choices. C, Cognitive flexibility measured by reversal learning in the Barnes maze. During initial learning (d1 to d4, each point represents the mean of traveled distance for four independent trials), improvement shown by reduction of the travel distance was faster in Shank3 ⌬4-22 wild-type and heterozygous mice than in homozygous animals; however, by day 4, the three groups were not different anymore and all of them had a strong preference for the escape hole quadrant during the initial probe test. During the reversal training (r1 to r4, each point represents the mean of travel distance for four independent trials), Shank3 ⌬4-22 homozygous mice initially traveled for longer distances but were still able to learn the new position and performed as well as their littermates on reversal days 2, 3, and 4. However, the reversal probe test at the end of the reversal training showed that while wild-type and heterozygous animals had a significant preference for the new target quadrant, the homozygous mice had a similar preference for the quadrants containing the initial and the reversal escape holes. WT, wild-type mice; Het, heterozygous mice; KO, homozygous knock-out mice. ‫:ء‬ WT versus KO; #: Het versus KO. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.1, ‫‪p‬ءءء‬ Ͻ 0.001.

Figure 9.
Learning and memory in Shank3 ⌬4-22 -deficient mice. A, Working memory in Y-maze measured by spontaneous alternation behavior. All genotypes showed comparable number of arm choices, percentage of correct choices (three-way alternation), type 1 error (three consecutive choices where the first and third choices are identical), or type 2 error (three consecutive choices where the second and third choices are identical). B, Contextual and cued fear conditioning in Shank3 mice. A higher percentage of freezing was observed in Shank3 ⌬4-22 homozygous mice compared to wild-type and heterozygous animals on day 1. While the difference was already present before the sound-shocks associations, it was strongly increased posttraining. No genotype differences were detected in freezing scores in the posttraining session on day 1. Opposite results were observed for contextual conditioning (day 2) and cued conditioning (day 3): verely. Difficulties in fine motor coordination have been described in ⌬4-9 and ⌬11 Shank3-deficient mice Drapeau et al., 2014;Vicidomini et al., 2017) and were confirmed in the current study. In addition, our homozygous mice were strongly impaired in the hanging test, the hindlimb placing test and the inverted screen and had small gait abnormalities. Hypersensitivity or hyposensitivity to sensory stimuli is frequently observed in PMS and ASD patients (Klintwall et al., 2011;Phelan and Betancur, 2011). However, little was known regarding the sensory abilities of Shank3deficient mice. No deficits were reported in ⌬4-9 or ⌬4-22 animals for either olfaction, audition, vision, neuromuscular reflexes or pain sensitivity (Bozdagi et al., 2010;Wang et al., 2011Wang et al., , 2016bYang et al., 2012). Normal pre-pulse inhibition was observed in many models including ⌬4-9, ⌬13, ⌬21, and ⌬4-22 Shank3-deficient mice Kouser et al., 2013;Wang et al., 2016b;Jaramillo et al., 2017) even if decreased pre-pulse inhibition was reported in in lines with point mutations in exon 21 . Here, we observed that Shank3 ⌬4 -22 homozygous mice have no strong visual deficits, and normal neuromuscular reflexes, but are hyper-reactive in response to handling and tactile stimuli. In addition, we observed a delay in the acquisition of the startle response in newborns and a decrease of the startle response in both heterozygous and homozygous adults. Since social behavior strongly relies on olfaction in rodents, we used different behavioral paradigms to evaluate our model. Interestingly, Shank3 ⌬4 -22 homozygous mice had a low interest for nonsocial olfactory stimuli as shown by deficits in the buried food test and by low amount of sniffing during the olfactory habituation/dishabituation paradigm. However, Shank3 ⌬4-22 -deficient mice were able to discriminate odors in the test for social transmission of food preference or to show interest for social stimuli during olfactory habituation/dishabituation, suggesting that they do not have anosmia but rather show reduced interest in nonsocial scents, which can be overcome when adding a social component.
One of the defining features of autism is the impairment of social interactions that can manifest by deficits in social approach, reciprocal social interactions and/or verbal and nonverbal communication. Mild social deficits have been reported, however with variability, in some of the previous studies of PMS mouse models (Fig. 11). In one of the most commonly used test, the three-chambered social approach test, no differences between the genotypes were reported in ⌬4-9, ⌬4-7, and ⌬9 models (Peça et al., 2011;Yang et al., 2012;Drapeau et al., 2014;Lee et al., 2015), while social deficits characterized by a lack of preference for a social stimulus were reported the models targeting ⌬11, ⌬13, or ⌬13-16 deletions (Peça et al., 2011;Duffney et al., 2015;Mei et al., 2016;Jaramillo et al., 2017;Luo et al., 2017;Vicidomini et al., 2017). Conflicting results were reported for ⌬21 models Duffney et al., 2015;Speed et al., 2015;Bidinosti et al., 2016;Zhou et al., 2016). Interestingly, consistent with Wang et al., 2016b and colleagues' study, we observed only minimal social deficit in our ⌬4-22 model. All genotypes had a similar preference for social stimulus in the three-chambered social approach test or the social transmission of food preference and only trends toward a decrease of interaction time and vocalization were found during male-female social interactions. Rodent social behavior is highly influenced by experimental conditions such as the animals' age, housing conditions, or animals handling and that can explain differences observed between cohorts of animals with identical or similar alterations of the Shank3 gene. While not representative of  When compared to wild-type and heterozygotes littermates, Shank3 ⌬4-22 homozygous mice displayed decreased rearing activity due to a decrease of wall rears rather than free standing rears. alternative splicing of Shank3 isoforms and even their subcellular distribution has been shown to be cell-type specific, activity dependent, and regionally and developmentally regulated , these differences also raise the possibility that different Shank3 isoforms could make distinct contributions to the phenotype of PMS and suggests that Shank3c and Shank3d (affected by deletions containing exons 11-16) could be particularly involved in the regulation of social behavior compared to isoforms Shank3a, Shank3b, or Shank3a/b that are disrupted by deletions of exons 4-9. The apparent absence of social deficit in the models with a complete deletion of Shank3 could be explained by the fact that those animals have a strong aversion for objects and be interpreted as an avoidance of the chamber containing the object rather than a real social preference. One of the strongest phenotype observed in the current study was indeed an active avoidance of inanimate objects. In the novel object recognition test, lack of preference for a novel object had previously been observed in two lines of ⌬4-9 mice Yang et al., 2012) but not in a third line (Jaramillo et al., 2016) nor in ⌬9 Shank3-deficient mice (Lee et al., 2015). However, in the present study, homozygous animals had very little interactions with both familiar and novel object making it impossible to properly compare novelty preference. Instead, they mostly spent their time in the corners of the open field away from the objects. Surprisingly, similar avoidance behavior was observed in the marble burying test and in the repetitive novel object contact task. We also observed a strong decrease of direct interactions with the applicator in the olfactory habituation/dishabituation test and a reduction of the quality of the nests build by Shank3 ⌬4 -22 -deficient animals with some mice even leaving the building material fully untouched. Some studies have reported that children with autism respond to novelty with avoidance behaviors and patients with PMS have enhanced reactivity to novel environments and reduced interest for objects. Decrease of marble burying has been consistently been described in other models of Shank3 deficiency as were nest building impairments (⌬11, ⌬13, ⌬21, and exon 21 point mutations; Kouser et al., 2013;Speed et al., 2015;Bidinosti et al., 2016;Jaramillo et al., 2017;Vicidomini et al., 2017). While we have shown that those animals are hypoactive and have significant motor deficits that could impact behavioral assays relying on exploratory locomotion, it is unlikely that this avoidance behavior is attributable to impaired motor activity or poor motivation as homozygous mice have normal pattern of investigation in an empty open field and actively avoid objects or even escape from the cages by jumping out while they will not escape from an empty cage or a cage containing an unfamiliar mouse. Furthermore, the number of escape attempts increased in relation with the number of objects present in the cage. In addition to this escape behavior, a high level of impulsivity was observed for adult homozygous mice in the beam walking test and for both newborn and adult homozygous mice in the negative geotaxis test.
Stereotypies, repetitive behaviors with restricted interests and resistance to change form the second set of core symptoms of ASD. Excessive grooming with or without development of skin lesions is the most commonly observed repetitive behavior in rodents. Repetitive/compulsive grooming has been reported in most of the previously published Shank3 mouse models (Fig. 11) while skin lesions where noticed only in some of them (⌬4-9, ⌬11, ⌬13-16, ⌬21, and point mutations in exon 21; Peça et al., Figure 11. Main features and comorbidities associated with Phelan-McDermid displayed by different mouse models with Shank3 deficits. Green indicates an absence of genotype difference. Blue indicates a decrease of the associated behavior in Shank3-deficient animals. Red indicates an increase of the associated behavior in Shank3-deficient animals. Gray indicates the behavior has not been studied in the corresponding article. Age column: d ϭ days, w ϭ weeks, m ϭ months, ‫ء‬ indicates that only the age at the beginning of the testing was provided. In conclusion, our complete Shank3 ⌬4 -22 mouse line provides a new and improved genetic model for studying mechanisms underlying ASD and PMS and is characterized both by better construct and face validities than previously reported lines of Shank3 mutants. Our in-depth behavioral characterization revealed behavioral features that reflect those observed in PMS and therefore suggest a greater potential as a translational model. Mice with a complete deletion of Shank3 are more severely affected than previously published mouse models with a partial deletion. Both sensory and motor disabilities were detected in neonate and adult mice. Shank3 ⌬4 -22 -deficient mice showed modest deficits in social behavior, reflected in reduced male to female anogenital sniffing and ultrasonic vocalization, but no major deficits in social preference in the three-chambered social interaction task. These findings are consistent with an independently generated mouse model (Wang et al., 2016b). Also in agreement with Wang's study, our Shank3 ⌬4 -22 mice showed increased anxiety and hyper-reactivity to novel stimuli, increased escape behaviors, and increased repetitive behaviors. Together with the increased freezing behavior in the cued fear conditioning, this suggest a dysregulation of amygdala circuitry that will require further investigation. In addition, our mice displayed impairments in several hippocampal-dependent learning and memory tests as well as cognitive inflexibility, thus recapitulating ID and insistence on sameness observed in autism and in the majority of patients with PMS. Although PMS patients are heterozygous for Shank3 mutations/deletions, most of the previous models have failed to demonstrate any relevant phenotype in heterozygous animals. Here, we were able to observe an intermediate phenotype for heterozygous mice in several of the parameters tested, notably in the open field, rotarod, startle response, escape behavior, reversal probe test, and elevated zero-maze. Heterozygous animals being less affected than their homozygous, we hypothesize that more challenging paradigms, for example by introducing a variable reward probability in tests such as the Barnes maze, would allow us to further highlight differences in heterozygous animals. Past studies have often failed to replicate behavioral phenotype even in models with very similar Shank3 disruption or in different cohorts from the same model. The concordant findings from two independently derived and analyzed Shank3 mouse models, including the comparison of two independent cohorts in our laboratory, demonstrate, for the first time, strong reproducibility and validity for a genetically modified mouse model of PMS, providing a valuable model for further investigations of the neurobiological basis of PMS and ASD.