Background
Since the initial discovery by Thomas Bourgeron and coworkers of
SHANK3 mutations in three cases of autism spectrum disorder (ASD) in 2007, many more cases have been reported [
1‐
14].
SHANK3 deficiency causes a monogenic form of ASD with a frequency of 0.5–1% of ASD cases [
7]. Deletion in the
SHANK3 gene is also central to the cause of the rare neurodevelopmental disorder, Phelan McDermid Syndrome (PMS)[
9]. The prevalence of
SHANK3 mutations has motivated the use of animal models with corresponding
Shank3 mutations to understand the underlying pathophysiology in cases of ASD, which harbor a
SHANK3 mutation, cases of PMS, and idiopathic ASD more broadly, with the goal of developing targeted pharmacological therapies.
Shank3, a scaffolding protein involved in the strengthening and stabilizing of synapses, is expressed in postsynaptic densities, a site of functional convergence of many ASD-related genes, rendering
Shank3 mutation a representative model of synaptopathy in ASD. A variety of mouse models have been generated with mutations in the
Shank3 gene, which include exon deletions affecting the ankyrin domain (
Shank3A, [
15‐
18], PDZ domain (
Shank3B, [
16,
19]), Homer domain (
Shank3ΔC, [
20]), and complete knockout of all isoforms [
21]. Reduced social behaviors, elevated repetitive behaviors, cognitive impairments, abnormalities in dendritic spines, and aberrant in vitro electrophysiological measures of synaptic plasticity have been reported to various degrees in these models [
15‐
17,
20‐
29]. Independent replications of these original reports, to confirm the strength of the various findings when conducted by other laboratories, have been conducted in only some cases. Further, to fully utilize these models for the development of novel therapies to treat ASD and PMS, quantitative and replicable biomarkers of neural dysfunction are needed in
Shank3 mutant mouse models.
Robustness and reproducibility of ASD-relevant phenotypes is essential to establish before an animal model can be effectively employed as a preclinical tool for therapeutic discovery. We therefore quantified seizure susceptibility and EEG power in the gamma frequency band in two cohorts of
Shank3B null mutant mice. To evaluate the reproducibility of the previously reported social deficits and repetitive behaviors of this
Shank3B mutant line [
16], we investigated a wide range of behavioral phenotypes in two independent cohorts of
Shank3B mice and their WT littermate controls. Behavioral testing followed a protocol of standardized methods and a precise order of testing at specified ages, and employed two or more corroborative tests within each behavioral domain, representing the rigorous experimental design developed by our collaborative Autism Speaks Preclinical Autism Consortium for Therapeutics (PACT).
EEG abnormalities, including seizures and subclinical epileptiform activity, are prevalent in both PMS and idiopathic ASD, consistent with the hypothesis that excitatory-inhibitory balance is widely disrupted in ASD [
30‐
32]. Importantly, EEG can be similarly measured in both rodent models and in human patients and thus EEG phenotypes have great translational relevance [
33]. To evaluate the utility of EEG as a quantitative biomarker, we characterized seizure propensity and oscillatory activity in the
Shank3B mutant mice. The stability of the EEG phenotype was assessed in two independent cohorts to evaluate reproducibility of the phenotype.
Shank3B mutation resulted in a dramatic resistance to seizure induction and an enhancement of gamma band oscillatory activity indicative of enhanced inhibitory tone, in both mouse cohorts. Behavioral phenotypes including elevated levels of repetitive self-grooming and parameters of male-female reciprocal social interactions were replicated in both Shank3B cohorts. Thus, demonstration of the replicability of behavioral phenotypes and the identification of a novel, translational EEG phenotype in the commercially available Shank3B line from The Jackson Laboratory (JAX) provides evidence of a stable model that can be utilized consistently and reliably across independent laboratories. Detailed methods are provided for the generation of both the behavioral and the electrophysiological phenotypes, to enable the use of this model for both mechanistic and treatment studies broadly within the field.
Discussion
Reliable, objective, and quantitative biological markers that translate across species remain an unmet need in ASD therapeutic drug development. Our report is the first comparison of such markers, both electrophysiological and behavioral, in separate cohorts of a genetic ASD mouse strain. Encouragingly, we find that most metrics that define the Shank3B KO are replicated in two mutant mouse cohorts.
Quantitative abnormalities in neurophysiology have been reported in subsets of individuals with ASD, as well as in some animal models of ASD, suggesting that EEG signatures may be promising markers for ASD patient stratification and treatment response monitoring [
56‐
59]. For example, enhanced EEG gamma oscillatory power has been reported across a number of genetic mouse ASD models, consistent with resting state EEG findings in individuals with neurodevelopmental disorders [
31,
60,
61]. Such examples include increased gamma power in both
Mecp2 and
Pten mutant mouse models [
62,
63]. In the present study, we found that male
Shank3B KO mice have a decreased susceptibility to PTZ induction of all forms of seizures and an enhancement of power in the EEG gamma oscillatory band preceding seizure induction. In line with our objective to identify reliable biomarkers, this finding was replicated in two independent
Shank3B KO cohorts. This phenotype is consistent with the initial Feng Laboratory assessment that stress-induced seizures were rarely observed in the
Shank3B KO mouse and consistently spontaneous seizures were never observed [
16].
EEG abnormalities, including seizures and subclinical epileptiform activity, are prevalent in both PMS and idiopathic ASD, consistent with the hypothesis that excitatory-inhibitory balance is widely disrupted in ASD [
30‐
32]. Importantly, EEG can be similarly measured in both rodents and humans, and thus EEG phenotypes have realistic translational relevance [
33]. Our EEG findings contrast to some degree with the clinical picture of patients with
SHANK3 mutations, as some patients with loss of function
SHANK3 mutations have epilepsy and 67% have some EEG abnormality [
31]. Yet, patients with
SHANK3 mutations exhibit considerable heterogeneity in seizure frequency, and a subset may, potentially, resemble the mouse phenotype described above. A prospective natural history study of EEG and epilepsy in individuals with
SHANK3 mutations will help determine to what degree the EEG abnormalities we observed in this mouse model relate to the clinical population.
The electrophysiological phenotype uncovered in the present study suggests the enhancement of inhibitory tone in
Shank3B KO mice. Notably, in contrast to findings in our null mutant mouse population, EEG of mutant mice with
SHANK3 duplication showed hyperexcitability discharges and electrographic seizures as compared to WT littermates [
64]. Thus,
SHANK3 levels—both too little and too much—appear to affect inhibitory neurotransmission. At a cellular level, EEG gamma-oscillations are likely generated by networks of parvalbumin (PV) cells, the most abundant subtype of GABAergic interneurons that contributes to perisomatic inhibition of glutamatergic principal cells [
65,
66]. Thus, EEG power in the gamma frequency band reflects the integrity of PV circuitry [
67,
68]. PV cell loss in the hippocampus and neocortex is associated with progression to spontaneous seizures after status epilepticus as well as other epileptogenic injuries [
69]. Inborn PV cell deficiency also increases seizure susceptibility [
70]. Recent meta-analysis indicates that the function and absolute number of PV cells are deficient in mouse ASD models [
71], and from this one can hypothesize that gamma EEG power would be lower in the mutant mouse strains. Yet, in our
Shank3B KO cohorts, we found enhanced gamma EEG power and reduced seizure susceptibility.
The above findings are consistent with mitigation for seizure risk by enhanced cortical inhibition as reflected in the EEG gamma band. Alternatively, the reduced seizure susceptibility in the Shank3B strain may reflect reduced excitatory tone, rather than enhanced inhibitory tone. The high power in the gamma frequency band of the Shank3B KO mouse model is also a plausible readout of high PV cell network activity responsible for heightened seizure threshold. Perhaps, this reflects a compensatory over-activation of the PV inhibitory system in the setting of increased seizure vulnerability in ASD. Yet, independent of the specific mechanism, our data raise prospects for some patients with SHANK3 mutations to also have a mild or absent epilepsy phenotype. Further, while beyond the scope of this report, such finding of seizure protection in a Shank3B KO mutant indicates a potential to manipulate the SHANK3 gene or protein product as a means to stop or prevent epileptic seizures.
ASD is diagnosed and defined by behavioral symptoms in the domains of (a) social interaction and communication and (b) stereotyped, repetitive behaviors with restricted interests. However, considerable heterogeneity characterizes the broad range of diagnostic and associated symptoms across individuals [
56,
72]. Similarly, mouse models of ASD with mutations in risk genes for ASD vary considerably in their expression of social and repetitive abnormalities, and in phenotypes relevant to the cognitive, anxiety, sensory, and motor abnormalities associated with many cases of ASD [
50]. One issue in the current literature is the extent to which variability in observed behavioral phenotypes in a mouse model of ASD may arise from procedural or environmental issues, such as differences in housing conditions or animal handling practices. Other issues include the use of non-standard methods and incorrect interpretations of behavioral results. In most cases, the behavioral phenotypes reported in the first publication of a mutant line of mice have not yet been repeated in follow-up cohorts, either by the same or other laboratories. In cases in which follow-up studies were conducted, findings were replicated in many cases [
43,
44,
49,
73,
74]. However, findings have not replicated in other cases, e.g. [
45], and anecdotal reports suggest that failures to replicate have not yet been published. Similar issues may arise in the future as more laboratories engage in assaying physiological parameters in rodent models of ASD.
To improve the utility of genetic mouse models of ASD pathogenesis, as a part of our collaborative Autism Speaks Preclinical Autism Consortium for Therapeutics (PACT), we are investigating behavioral and translational physiological phenotypes in replication cohorts of mice with published mutations in risk genes for ASD and related conditions. To this end, we conducted in-depth phenotyping of the
Shank3B null mutant mouse model originally generated by Guoping Feng and coworkers at Duke University [
16]. These mice harbor the
Shank3 mutation in the PDZ domain and were reported to display remarkably high levels of repetitive self-grooming and social deficits [
16]. Of note, similar results were reported by Craig Powell and coworkers using a different genetic manipulation of the
Shank3 gene [
19]. Our larger goal for the present report and for related PACT studies is to identify behavioral phenotypes that replicate consistently in independent cohorts of mice within and between laboratories, in order to strengthen the use of preclinical mouse models of ASD, (a) for understanding the mechanistic underpinnings of ASD-relevant phenotypes, and (b) for preclinical translation in evaluating the therapeutic potential of novel treatments.
In the present studies, we employed a highly-standardized set of behavioral testing methods to identically evaluate two separately bred cohorts of Shank3B KO and their WT littermates, both males and females, in a fixed sequence of behavioral assays. Overall strategy, techniques and methods, testing batteries, and test sequence were developed as a component of PACT, in close collaboration among Drs. Crawley and Sahin and senior leadership of the Autism Speaks scientific research team. In addition to translational neurophysiological markers, the PACT preclinical platform was designed to evaluate behaviors relevant to the diagnostic and associated symptoms of autism, including social, repetitive, cognitive, anxiety-related, sensory, and motor traits, in multiple lines of genetic mouse models, across two independent cohorts and at least two corroborative tests within each behavioral domain. Two cohorts, each designed to yield Ns of 10 per genotype and sex, were independently bred and tested, to include a comparison of sex as a biological variable, for optimal experimental design in testing potential therapeutics.
Similar results were obtained between the two cohorts on most, but not all, of the ASD-relevant social and repetitive behavior assays conducted. The strongest ASD-relevant phenotype in
Shank3B KO mice was repetitive self-grooming, as originally reported [
16]. Grooming scores were almost twice as high in
Shank3B KO as compared to their WT littermates. Time spent self-grooming was significant for both male and female KO and their combined scores in cohort 1, and for females and the combination of males and females in cohort 2. These findings support the interpretation that repetitive self-grooming is a robust and reproducible feature of
Shank3B KO mice. Repetitive behaviors are a common manifestation in individuals with Phelan-McDermid Syndrome [
75], seen in over half the patients. Thus,
Shank3B KO mice offer a robust animal model for future studies to develop treatments for repetitive behaviors.
In the social domain, reciprocal social interactions in same-sex dyads at juvenile age 22–28 days old were not significantly different between genotypes, in both cohorts 1 and 2. Three-chambered social approach, an assay developed by our group in 2004 [
36,
37], revealed normal sociability in
Shank3B KO males in cohort 1 and in
Shank3B KO females in cohorts 1 and 2, but absence of sociability in
Shank3B KO males in cohort 2. The more sensitive, nuanced assay of reciprocal social interactions between freely moving subject males and WT estrous females showed strong genotype differences on some parameters in both cohorts, including ultrasonic vocalizations emitted. Other parameters of reciprocal social interactions did not differ between genotypes. For behavioral assays with multiple outcome parameters, interpretations may best be based on the preponderance of significant genotype differences across several related parameters. The preponderance of significantly less sniffing and vocalizing during male-female reciprocal social interactions in two cohorts, but some normal scores on three-chambered social approach, emphasizes the need to conduct more than one behavioral assay and to select the strongest outcome measures of the mutation to employ in therapeutic discovery.
Results that differ between cohorts are the most difficult to interpret. For example, if interpretations had been based on three-chambered sociability results from only the second cohort, we would have concluded that sociability was impaired in Shank3B KO males but not females, and focused the discussion on an interesting sexual dimorphism relevant to the higher prevalence of ASD in boys than girls. However, since sexual dimorphism appeared in only one of two cohorts, interpretations must be more cautious. Further caution would be extended to the use of our simpler automated three-chambered social approach task as the sole social assay for preclinical therapeutic discovery in the Shank3B KO line of mice. Social assays with higher sensitivity, such as male-female interactions, and new assays with greater face validity to the types of social abnormalities that characterize autism, may improve translational success.
Open field exploration was reduced in Shank3B KO in males in cohort 1, and in both sexes in cohort 2, representing a relatively strong and reproducible phenotype. Lower scores on several open field parameters in Shank3B KO mice, detected in the Crawley lab, is consistent with the Shank3B KO displaying significantly less activity in circadian home cages in the Sahin lab. While reduced activity introduced a potential confound in interpreting light↔dark anxiety-related behavior and fear conditioning, as described below, the magnitude of locomotor differences did not appear to be large enough to impact performance on other assays, as measured by internal activity parameters during performance in other tests including three-chambered social approach and novel object recognition assays.
Anxiety-related tests produced variable results. On the elevated plus-maze, no genotype differences were detected in either cohort on either of the two anxiety-related parameters, percent open arm time and number of open arm entries. However, total number of entries, the control measure for locomotion during elevated plus-maze testing, showed less exploratory activity in Shank3B KO males, but not in Shank3B KO females, and in Cohort 1 only, suggesting overall normal performance on this anxiety-related test. Light↔dark transitions were significantly lower in both male and female Shank3B KO in Cohort 1, and time in the dark was higher in Cohort 1 males. However, in Cohort 2, all parameters of light↔dark anxiety-like behavior and locomotion showed no genotype differences in either sex. An anxiety-like phenotype in one cohort but not in the other cohort is best interpreted as a minor finding, of insufficient replicability to use in a treatment study. It is interesting to note that comparison by sex revealed behaviors in which similar phenotypes were detected for both males and females, and behaviors in which genotype differences were significant in only one sex, either male or female in the anxiety-related domain and in other behavioral assays. These results highlight the usefulness of displaying results by sex as well as a combined total, when confirming replicability.
Marble burying was lower in
Shank3B KO, rather than higher as predicted from the assumption that marble burying is a repetitive behavior. Both male and female
Shank3B KO buried fewer marbles as compared to same-sex WT. This unexpected result adds to an existing question in the behavioral neuroscience field about whether marble burying represents a repetitive behavior, an anxiety-related behavior, an artifact of digging in deep litter, or something else [
76]. It seems possible that a competing behavior such as self-grooming was responsible for less marble burying in
Shank3B KO. However, as marble burying was not conducted in cohort 1, but added as a corroborative task in cohort 2, findings will require replication in a third cohort.
Sensory abnormalities appeared on some assays. Reduced acoustic startle to loud decibel level white noise stimuli indicate somewhat reduced hearing or reduced perception of startle stimuli. In contrast, no genotype differences in prepulse inhibition were detected in either sex in either cohort. Olfactory abilities in the olfactory habituation/dishabituation assay were normal for Shank3B KO females but showed some impairments in Shank3B KO males, a finding that will require replication in another cohort. Hot plate response latencies did not differ between genotypes, indicating normal nociception on this gross measure of pain sensitivity.
Cognitive abilities appeared to be mostly intact in Shank3B KO mice on novel object recognition, contextual and cued fear conditioning, and Morris water maze acquisition. Female Shank3B KO displayed borderline deficits on novel object recognition and failed the water maze probe trial. These variable findings justify future studies to repeat the cognitive tests in a future cohort of Shank3B KO and WT mice. While intellectual impairments, anxiety, hyperactivity, and unusual responses to sensory stimuli are associated symptoms rather than diagnostic symptoms of ASD, and therefore were not the primary focus of our PACT experimental design, strong phenotypes in an associated symptom domain could provide valuable additions to a mouse model of ASD when evaluating potential therapeutics.