Sleep phenotypes in male and female Shank3ΔC ASD model mice

We first tested whether Shank3ΔC ASD model mice exhibited sleep disruption during development. We measured total sleep amount and bout lengths in male and female Shank3WT/ΔC heterozygotes, Shank3ΔC/ΔC homozygotes, and wild-type (WT) littermates, at two developmental time points: juvenile (P25–P41) and adolescent (P42–P56). Our goals were to determine if developing Shank3ΔC exhibits sleep disruption, and to determine when this phenotype emerges. Wake and sleep behavior was monitored for an average of 8–10 uninterrupted days using PiezoSleep, a noninvasive piezoelectric home-cage recording system previously validated using simultaneous EEG/EMG and video recordings [29,30,31]. Average daily total sleep amount and bout lengths were calculated from repeated days of uninterrupted recording for each individual to yield a robust measurement of typical daily sleep behavior. PiezoSleep requires that mice be single housed during recording. Therefore, separate cohorts were generated for the juvenile and adolescent groups in order to mitigate the lasting negative effects of social isolation during development [35]. In juveniles and adolescents of both sexes we observed that Shank3ΔC/ΔC mice show a significant reduction in total daily sleep compared to WT littermates (1-way ANOVA followed by Tukey’s multiple comparisons test, WT: Shank3ΔC/ΔC, juvenile males p < 0.0001; juvenile females p = 0.0004; adolescent males p = 0.0021; adolescent females p = 0.009). Reduced sleep amount in juveniles was significant in the light and dark phases (Fig. 1A-D). Additionally, in adolescent Shank3ΔC/ΔC mice compared to WT, we observed reductions in sleep bout length in the light phase in the female group (female p = 0.0053; male closely approached significance at p = 0.0532) and both sexes in the dark phase (male p = 0.0199, female p = 0.0215) (Fig. 1C, D). Post hoc tests did not reveal any significant differences in sleep amount or bout length between WT littermates and Shank3WT/ΔC heterozygotes in either sex or age (Fig. 1A-D, complete reporting of statistical analysis is shown in Additional file 1). Overall, the data show that male and female Shank3ΔC/ΔC mice have reduced total sleep detected in juveniles as an early phenotype, and that sleep additionally becomes fragmented as the animals enter adolescence. These findings are in agreement with a recent study describing reduced sleep in developing homozygous Shank3 InsG3680 ASD model mice [10]. Additionally, the worsening (fragmentation) of sleep behavior during adolescence in Shank3ΔC/ΔC homozygous mice matches observations obtained from some Phelan–McDermid syndrome patients during the transition to adolescence [21, 27, 36]. Together with a previous report describing sleep disruption in male adult Shank3ΔC/ΔC mice [27], our findings show that sleep disruption in Shank3ΔC/ΔC mice has an early onset that persists throughout the lifespan.

Early life and post-adolescent sleep disruption (ELSD/PASD) in WT and Shank3 WT/ΔC heterozygotes

In contrast to homozygous Shank3 ASD mouse models, the clinically relevant genotype, Shank3 heterozygotes, has repeatedly been shown to have no or limited ASD-relevant phenotypes [22,23,24,25,26], in agreement with our sleep measurements (Fig. 1). We speculated that heterozygous mutation of Shank3 confers a clear genetic vulnerability in mice that may be sensitive to experimental perturbations of sleep during postnatal development. A major question in the ASD field is whether sleep disruption commonly seen in patients, contributes to other ASD phenotypes. Additionally, ASD is diagnosed with an ~ 4:1 male bias, although the basis for vulnerability in males or resilience in females is not understood [37, 38]. Therefore, we reasoned that Shank3 heterozygous mice are an ideal model to examine a potential causal role of developmental sleep disruption in lasting sex-specific changes in behavior as a novel gene x sex x environment interaction model for ASD susceptibility.

In order to test whether developmental sleep disruption contributes to altered adult behavior in our Shank3 ASD model, we utilized the experimental early life sleep disruption (ELSD) paradigm recently established by Jones and colleagues [7, 9]. In their recent publications, Jones et al. exposed developing prairie voles to ELSD from P14–P21, using an automated method in which the home cage is placed on an orbital shaker, and sleep is regularly disturbed by agitation at 110 RPM for 10 s every 109 s [10 s on, 99 s off]. In developing voles, this automated method produces a very mild reduction in total sleep amount, but a clear reduction in sleep bout lengths, indicating the method robustly produces sleep fragmentation. Because REM sleep is typically engaged from NREM sleep, one clear effect of this mild but frequent fragmentation method was a significant reduction in total REM sleep [7]. ELSD treatment in voles resulted in lasting changes in adult social and cognitive behavior, with a male bias, strongly suggesting that sleep disruption during sensitive periods of brain development is sufficient to drive lasting ASD-relevant phenotypes [7, 9]. Similar variations of this automated method have also been shown to cause a reproducible fragmentation of sleep behavior and lasting negative consequences with sustained treatment in adult mice [32, 33, 39, 40].

Here, we tested the lasting effects of this ELSD paradigm on WT and Shank3WT/ΔC mice (littermates), reproducing the method described by Jones et al. [7, 9] (see “Methods” section). To specifically investigate whether early life represents a unique period of vulnerability to sleep disruption, we also tested the effects of sleep disruption treatment delivered later in life from P56–P63, referred to as post-adolescent sleep disruption (PASD) (Fig. 2A). Control groups were placed onto identical orbital shakers that were left off. For ELSD treatments, litters of pre-weaned mouse pups were exposed to sleep disruption from P14–P21, together with their dam (WT) (Fig. 2A). For PASD treatments, separate cages of weaned males and females were exposed to sleep disruption from P56–P63 (Fig. 2A). All groups included WT and Shank3WT/ΔC heterozygous littermates of both sexes. Because the dam is also experiencing sleep disruption in the ELSD experiment, pups were examined and weighed daily to ensure that they were receiving sufficient maternal care. Control and ELSD pups maintained healthy appearance and showed comparable daily weight gain (2-way ANOVA: F[1, 74] = 1.128, p = 0.2916) (Fig. 2B), and no pups were excluded from further analysis due to any health concerns indicating that the treatment was well tolerated. Interestingly, male Shank3WT/ΔC mice in the ELSD group gained slightly more weight than control; however, the weight in this group normalized after weaning and the end of ELSD treatment (Additional file 3). The absence of any weight loss suggests that ELSD treatment did not impair maternal care. To test whether ELSD treatment induces high levels of stress, an additional cohort of mice underwent control or ELSD treatment followed immediately by killing at P21 and measurement of serum corticosterone by ELISA. We did not observe any changes in corticosterone levels between ELSD pups and control (unpaired t test, t = 0.8765, df = 48, p = 0.3851) (Fig. 2C, and Additional file 3), suggesting that pups acclimate to this treatment without sustained levels of stress. Finally, to confirm that our automated method is able to fragment sleep in developing mice, we used live-video recording to monitor sleep/wake activity of mouse pups undergoing control or ELSD treatment from P14–P21 (see “Methods” section). Our recordings show that mice treated by ELSD show a decrease in sleep bout lengths from P14 to P21 (Paired t test, control: t[7] = 1.484, p = 0.1814; ELSD: t[6] = 3.923, p = 0.0078) (Additional file 3: D–F). These data suggest that ELSD treatment successfully fragments juvenile sleep but does not overtly impair mouse pup growth and maternal care, or drive a sustained increase in stress, replicating the same conclusions reached from ELSD treatment in prairie voles [7].

Upon reaching P70, control, ELSD, and PASD mice underwent an extensive panel of behavioral testing that lasted 6–7 weeks (Fig. 2A). These behavior tests are divided into non-social (Figs. 3, 4), and social (Fig. 5); non-social tests: elevated plus maze, open-field behavior, pre-pulse inhibition of acoustic startle, accelerating rotarod, olfactory ability, and marble burying; social: sociability and social novelty preference in the 3-chamber assay. We asked whether ELSD/PASD treatment caused lasting changes in behavior, whether Shank3WT/ΔC heterozygotes were more vulnerable to these treatments than WT siblings, and whether distinct effects were seen between the sexes.

Sleep disruption during early life contributes to lasting sex-specific changes in specific non-social behaviors in genetically vulnerable Shank3 WT/ΔC heterozygotes

For the non-social assays, we did not observe any differences in behavior between WT and Shank3WT/ΔC littermates in the control treatment group (Figs. 3, 4, Additional file 4), consistent with several previous reports [22,23,24,25,26], suggesting that loss of a single copy of Shank3 alone minimally affects behavior in mice. Strikingly, we observed that ELSD treatment resulted in lasting and sex-specific changes in several non-social behaviors in genetically vulnerable Shank3WT/ΔC heterozygotes, while WT littermates were found to be resilient.

ELSD treatment had female-specific effect on behavior in the elevated plus maze, an assay of anxiety-like behavior. Shank3WT/ΔC females exposed to ELSD treatment showed a significant increase in time spent in the open arms of the plus maze in comparison with WT littermates (unpaired t test with Holm–Šídák correction: t[16] = 3.51, p = 0.0087), and a similar trend in the number of open arm entries (t[16] = 2.557, p = 0.062). No differences were seen in any male groups (Fig. 3A–D). We describe this female-specific phenotype as “decreased risk aversion.” PASD treatment resulted in a trend of increased time in the open arms of the elevated plus maze (unpaired t test with Holm–Šídák correction: t[22] = 2.15, p = 0.084) (Fig. 3C), suggesting that Shank3WT/ΔC females may have an extended period of vulnerability to the effects of sleep loss in this behavior.

Male-specific effects were seen in open-field behavior and pre-pulse inhibition of acoustic startle response (PPI). In comparison with sex-matched WT littermates, male Shank3WT/ΔC heterozygotes exposed to ELSD showed hypo-activity in the open field test as shown by a reduction in total distance traveled (unpaired t test with Holm–Šídák correction: t[19] = 2.99; p = 0.023) (Fig. 3E), whereas no changes were seen in females (Fig. 3F). We did not observe any differences in anxiety-like measures of rearing, or time in the center in the open field test (Additional file 4). Moreover, male Shank3WT/ΔC heterozygotes exposed to ELSD showed reduced PPI, a measure of sensory motor gating (2-way ANOVA: F[1, 19] = 5.85, P = 0.026, main effect of genotype) (Fig. 3G). ELSD treatment showed no effect on PPI behavior in females (Fig. 3H). Interestingly, a small difference in PPI was observed between genotypes in the PASD treated females, albeit at only a single pre-pulse sound (2-way ANOVA: F[1, 22] = 0.7215, p = 0.0423 at 74 dB, unpaired t test with Holm–Šídák correction) (Fig. 3H), suggesting a minor change in sensory motor gating.

All groups performed comparably on the accelerating rotarod assay of motor coordination and motor learning, in the marble burying assay for repetitive behavior, and in the hidden food assay of olfactory ability (Fig. 4). Therefore, decreased risk aversion in ELSD-Shank3WT/ΔC females and hypo-activity in open-field and reduced PPI in ELSD-Shank3WT/ΔC males are unlikely to result directly from altered motor behavior, olfaction, or repetitive behavior. In these assays, ELSD treatment had a greater effect on lasting behavior than PASD treatment, suggesting that Shank3WT/ΔC mice gain resilience to many of the effects of developmental sleep disruption with maturation (Fig. 3). Together, these results clearly show that ELSD treatment interacts with Shank3 mutation to potentiate lasting and sex-specific changes in multiple behaviors in genetically vulnerable Shank3WT/ΔC heterozygotes.

Differential effects of ELSD and PASD in social behaviors in genetically vulnerable Shank3 WT/ΔC heterozygotes

Alterations in social behavior are one of the core diagnostic criteria for ASD, and changes in certain aspects of social behavior are frequently reported in ASD model mice [41]. Here we tested the effects of Shank3WT/ΔC heterozygous mutation and the effects of our sleep disruption treatments on sociability and social novelty preference using the 3-chamber choice task previously described [26, 42, 43]. In the sociability phase of the assay, mice are evaluated for a preference to explore a non-social stimulus (N, a small empty enclosure) or a novel stranger mouse within a small enclosure (social stimulus: S/stranger 1). Social novelty preference is assessed in the subsequent phase by replacing the non-social stimulus with a novel stranger mouse (stranger 2), keeping the now familiar stranger 1 mouse in place. Typically developing mice are expected to spend more time engaging with the social stimulus vs. the non-social stimulus, and to spend more time engaging with novel stranger 2 vs. the already-investigated stranger 1, in the tests for sociability and social novelty preference, respectively. Sociability was recently reported to be intact in Shank3WT/ΔC heterozygotes of both sexes in comparison with WT littermates, whereas social novelty preference was reported to be impaired in both male and female Shank3WT/ΔC heterozygotes [26]. Moreover, a recent publication has shown that a period of adolescent sleep disruption can impair social novelty preference in WT C57BL/6J mice, while having no effect on sociability [10]. Therefore, these recent findings, together with previous work, have shown that sociability and social novelty preference are clearly distinct phenotypes, with different underlying neural circuits and separate genetic and environmental vulnerabilities [41].

Under control conditions, we observed the expected social approach behavior in WT mice of both sexes [42]. As previously reported, under control conditions Shank3WT/ΔC heterozygotes of both sexes showed intact sociability, comparable to WT littermates (Fig. 5A–C) [26, 44]. In response to ELSD or PASD treatments, all groups showed intact sociability, with the clear exception of male ELSD-Shank3WT/ΔC heterozygotes that exhibited a lack of preference for the social stimulus (S) in comparison with the non-social stimulus (N) (ELSD HET p = 0.125; ELSD WT p = 0.0006; CON HET p = 0.0384; CON WT p = 0.02, paired t test with Holm–Šídák correction for all social behavior results) (Fig. 5A-C). Importantly, sociability was intact in male PASD-Shank3WT/ΔC heterozygotes. These findings show that sociability is uniquely affected in genetically vulnerable males exposed to developmental sleep disruption, and that these individuals gain resilience to these negative effects with maturation. (Complete reporting of statistical analysis for social assays is found in Additional file 2.)

The subsequent test for social novelty preference was found to be influenced by sex, genotype, and PASD treatment, but not ELSD treatment. WT males showed intact social novelty preference under all conditions. Interestingly, where male ELSD-Shank3WT/ΔC heterozygotes showed impaired sociability (Fig. 5B), these same individuals displayed fully intact social novelty preference (ELSD HET p = 0.0003; ELSD WT p = 0.006) (Fig. 5D). In male PASD-Shank3WT/ΔC heterozygotes sociability was intact, but social novelty preference was impaired (Fig. 5D), albeit, these individuals still showed a trend toward preference for the novel stranger 2 (PASD HET p = 0.085; PASD WT p = 0.011), suggesting that social novelty behavior in genetically vulnerable males is sensitive to sleep disruption at later ages of development, consistent with recent reporting [10]. For the female groups, in our hands, we observed that only female Shank3WT/ΔC heterozygotes showed impaired social novelty preference under control conditions (Fig. 5D-E). This is in contrast to a recent publication reporting impaired social novelty preference in Shank3WT/ΔC heterozygotes of both sexes [26]. WT female littermates showed expected social novelty preference under control and ELSD conditions, but this behavior was found to be impaired in response to PASD treatment (PASD WT p = 0.298) (Fig. 5E), consistent with a vulnerability in this behavior to adolescent sleep loss [10]. Thus, in our hands, social novelty preference was found to be affected by sex, genotype, and PASD treatment. Shank3WT/ΔC females were the most vulnerable, showing baseline impairments. WT females and Shank3WT/ΔC males showed some selective vulnerability to PASD treatment. WT males were resilient to the treatments tested here. Overall, our findings show that developmental sleep disruption interacts with underlying genetic vulnerability and sex in Shank3WT/ΔC heterozygotes to drive lasting and sex-specific changes in specific aspects of non-social and social behaviors.

Fig. 5

Social approach and social novelty preference behaviors are differentially sensitive to sex, Shank3 genotype, and sleep disruption treatments. A Examples of individual WT and Shank3WT/ΔC heterozygotes (HET) from control, ELSD, and PASD treatment groups performing the 3-chamber social approach task. Heat map indicates time in location. Mice can freely move between 3-chambers and choose to spend time engaging with the non-social stimulus (N) or with a social stimulus (S). B, C Social approach behavior: time spent in proximity with the non-social (N) and social stimulus (S) for WT and HET males (B) or females (C). ELSD HET males show no preference for the social stimulus. D, E Social novelty preference: time spent in proximity with familiar stranger 1 (S1) and novel stranger 2 (S2) for WT and HET males (D) or females (E). PASD HET males and WT females show no preference for S2. HET females show now preference for S2 under any condition tested. *P < 0.05, **P < 0.01, **P < 0.001 (paired t test with Holm–Šídák correction). N = 8–12 per treatment/sex/genotype. Summary of statistical analysis is shown in Additional file 2