Altered sensory processing is a pervasive but poorly understood symptom in individuals with autism spectrum disorders (ASD)1. Sensory symptoms manifest as muted or excessive responses to light, sound, and/or touch. Because of variability in both the presence and presentation of sensory symptoms, gaining a mechanistic understanding of these sensory processing deficits remains a challenge. In contrast to ASD as a whole, genetically defined forms of ASD share similar sensory deficits. For instance, individuals with Phelan McDermid Syndrome (PMS), a syndromic form of ASD, show low sensitivity to pain and reduced responses to auditory and visual stimuli 2, 3. PMS is caused by the loss of function of one copy of the SHANK3 gene, due to either terminal deletions of chromosome 22 4 or SHANK3 point mutations 2. Here we identify the neurobiological basis of sensory hyporeactivity in shank3 loss-of-function zebrafish models of PMS.
Several animal models of PMS recapitulate muted responses to diverse sensory stimuli: pain in Shank3 mutant mice 5, sound in Shank3 mutant rats 6, and both touch and light in shank3ab mutant zebrafish 7, 8; nonetheless, a brain-wide understanding of these muted responses is lacking. Hyporeactivity in PMS could reflect functional changes that either span the entire brain or are localized to specific brain regions and/or muscle 9. Zebrafish allow unique experimental approaches to identifying underlying mechanisms because, within the first week of life, larvae have fully functional sensory-motor circuits and produce robust, stereotyped responses to calibrated sensory stimuli 10 These larval zebrafish have transparent vertebrate brains composed of only ~100,000 neurons, allowing unbiased functional approaches to map brain-wide neuronal activity. Moreover, embryonic transplantation can be used to make wildtype-mutant chimeras to test for brain-region-specific functional rescue 11, 12. Here, we use brain-wide activity mapping and transplants to identify and functionally validate brain regions that underlie sensory hyporeactivity to changes in light in zebrafish shank3 mutant models.
In contrast to the single SHANK3 gene in people, the shank3 gene is duplicated in zebrafish; therefore, to generate zebrafish models of PMS we used CRISPR/Cas9 to mutate both the shank3a and shank3b (shank3ab) gene paralogs. Shank3 proteins are large, ~200 kD, with multiple isoforms that can be differentially impacted by mutations in different parts of the gene13. To capture this complexity, we generated two zebrafish PMS models, shank3abDN with mutations truncating both the Shank3 a and b proteins in the ankyrin repeat domains and shank3abDC with mutations truncating both the Shank3 a and b proteins near the proline-rich domain 13 (Fig. 1a; Supplementary Fig. S1, Supplementary Tables 1&2). These models mimic the most common types of SHANK3 mutations found in people with PMS and, by having two models, we control for genetic background. In mice and humans, Shank3 protein is expressed in glutamatergic granule cells of the cerebellum, colocalizing with the scaffolding protein PSD-95. Likewise, in wildtype zebrafish, we show that Shank3 protein colocalizes with PSD-95 in the cerebellum and along ventral neural tracts of the brainstem (Fig. 1b; Supplementary Fig. S2). In contrast, in both shank3abDN-/- and shank3abDC-/- PMS models Shank3 staining is lacking despite intact PSD-95 synaptic puncta (Fig. 1b). These data indicate that the four alleles that underlie the two shank3abDN-/- and shank3abDC-/- models are loss-of-function mutations. Hereafter, we refer to shank3abDN-/- and shank3abDC-/- models are as shank3ab-/- mutant models except in cases that the results differ between the models.
The sensory reactivity of zebrafish shank3ab-/- models was measured by quantifying behavioral changes to a light-based stimulus using the well-established visual motor response (VMR;14). The VMR is characterized by dramatic increases in movement in response to sudden transitions from light to darkness (Fig. 1c). Both shank3ab-/- mutant models exhibited reduced VMR responses as quantified by comparing the distance traveled in the thirty seconds before and after the transition from lights-on to lights-off conditions (Fig. 1c & d; Supplementary Tables 3-8). Muted VMR responses were more pronounced in homozygous shank3ab-/- larvae (p<0.001) than in heterozygous shank3ab+/- larvae (p<0.05). We used the pronounced VMR deficits shank3ab-/- mutants as the basis of all subsequent experiments to determine the mechanistic underpinnings of these altered sensorimotor integration phenotypes.
To identify the neural circuits underlying hyporeactivity in shank3ab mutant models, we used an unbiased brain-wide Mitogen Activated Protein (MAP)-mapping12 approach, based on phosphorylation of extracellular signal-regulated kinase (pERK). Because ERK phosphorylation increases when calcium is elevated during action potentials, staining for pERK provides a proxy for neuronal activity (Fig. 2a & b). Brain regions differentially active between light-on and lights-off conditions were identified by statistically comparing relative ERK signals (pERK/total ERK) in two groups of 15-21 larvae per group (p<105; Fig. 2c & d, Supplementary Figs. S3 & S6). In response to the lights-on stimulus, wild type (WT) and shank3ab-/- models showed similarly elevated pERK staining in the optic tectum (green) that receives input from retinal ganglion cells. In response to the lights-off stimulus, WT showed elevated pERK staining in the pineal, the telencephalic pallium and subpallium, the torus semicircularis of the midbrain, brainstem, and spinal cord (magenta). While shank3ab-/- mutant models showed similarly elevated pERK staining in the pineal, they showed little or no elevated pERK other brain regions. These VMR brain activity maps in shank3ab-/- models show that sensory brain regions including the pineal, retina, and optic tectum detect changes in light normally, but that downstream brain regions fail to integrate and respond to dark transitions consistent with muted lights-off behavioral responses.
Next we explored whether restoring Shank3 function would be sufficient to rescue hyporeactivity in both shank3abDN-/- and DC-/- mutant models. We generated genetically mosaic larvae by transplanting WT cells into otherwise shank3ab mutant embryos at the late gastrula shield stage, ~ six hours post-fertilization (Fig. 3a, Supplementary Fig. S5). WT donor cells were deposited in the region of the shank3ab-/- embryo fated to become brainstem. To track the fate of transplanted cells, WT donor Zebrabow embryos expressing dTomato under a ubiquition promoter 15 were used as the source of WT cells, referred to as ZbT for Zebrabow transplants (Fig. 3). Remarkably, when tested as six-day-old larvae, transplanted ZbT cells were sufficient to rescue shank3ab-/- mutant lights-off reactivity in the VMR assay (Fig. 3 b-d, Supplementary Fig. S5; Supplementary Tables S9-16). To determine ZbT brain regions in common among behaviorally rescued shank3abD:ZbT larvae, we registered shank3abD:ZbT larvae to the Z-brain atlas. We found that the majority of rescued shank3abD:ZbT larvae had integrated ZbT cells in a rostral dorsal glutamatergic brainstem nucleus known in zebrafish brain atlases as vGluT cluster 2 (90.5%; n=19/21; Supplementary Fig. S6; Supplementary Table S17). Previous studies using whole-brain gCaMP have identified this brainstem nucleus as important in transforming sensory inputs to behavioral responses 16, 17. To control for non-specific transplantation effects, we performed within genotype transplants. WT donor to WT recipient chimeras and shank3abDN-/- donor to shank3abDN-/- recipient chimeras had no effects on VMR behaviors compared to unmanipulated larvae of the corresponding genotype. shank3abDC-/- donor to shank3abDC-/- recipient chimeras had more severe hyporeactivity compared to unmanipulated larvae of the same genotype (Supplementary Figs. S7, Supplementary Tables S18-23). Consistent with the MAP-mapping experiments, these results indicate Shank3ab function in rostral brainstem is sufficient for WT levels of light-evoked activity.
Previous work applying whole brain imaging in zebrafish larvae has highlighted the roles of the cerebellum and rostral brainstem as regions that receive inputs from sensory centers to coordinate the appropriate motor output 16-18. Brainstem deficits in shank3abDN-/- and DC-/- mutants could be due to synaptic deficits and/or altered development. In support of a synaptic role, loss of Shank3 protein in mammalian models is known to decrease glutamate receptor expression, disrupt post-synaptic density composition, and reduce synaptic transmission 19. Weaker excitatory synaptic responses could therefore explain the failure of sensory brain regions to evoke responses at the levels of both other brain regions and motor behaviors in shank3ab mutant PMS models. Functional deficits could also be due to altered development that could disrupt functional connectivity. Supporting this possibility, global developmental delay has previously been reported in shank3ab zebrafish models 7, 8. Moreover, altered brainstem development has been suggested as the likely basis for multisensory integration and sensory-motor deficits more generally in ASD20, 21. Such developmental deficits in brainstem regions could help explain the efficacy of rostral sensorimotor brainstem transplants in rescuing VMR behaviors in shank3ab-/- mutants in this study. With the recent inclusion of sensory deficits, more clinical research is needed to determine links between changes in the brainstem function and sensory deficits in individuals with autism.