Dietary zinc supplementation rescues fear-based learning and synaptic function in the Tbr1+/− mouse model of autism spectrum disorders

Background Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterised by a dyad of behavioural symptoms—social and communication deficits and repetitive behaviours. Multiple aetiological genetic and environmental factors have been identified as causing or increasing the likelihood of ASD, including serum zinc deficiency. Our previous studies revealed that dietary zinc supplementation can normalise impaired social behaviours, excessive grooming, and heightened anxiety in a Shank3 mouse model of ASD, as well as the amelioration of synapse dysfunction. Here, we have examined the efficacy and breadth of dietary zinc supplementation as an effective therapeutic strategy utilising a non-Shank-related mouse model of ASD—mice with Tbr1 haploinsufficiency. Methods We performed behavioural assays, amygdalar slice whole-cell patch-clamp electrophysiology, and immunohistochemistry to characterise the synaptic mechanisms underlying the ASD-associated behavioural deficits observed in Tbr1+/− mice and the therapeutic potential of dietary zinc supplementation. Two-way analysis of variance (ANOVA) with Šídák's post hoc test and one-way ANOVA with Tukey’s post hoc multiple comparisons were performed for statistical analysis. Results Our data show that dietary zinc supplementation prevents impairments in auditory fear memory and social interaction, but not social novelty, in the Tbr1+/− mice. Tbr1 haploinsufficiency did not induce excessive grooming nor elevate anxiety in mice. At the synaptic level, dietary zinc supplementation reversed α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) and N-methyl-d-aspartate receptor (NMDAR) hypofunction and normalised presynaptic function at thalamic-lateral amygdala (LA) synapses that are crucial for auditory fear memory. In addition, the zinc supplemented diet significantly restored the synaptic puncta density of the GluN1 subunit essential for functional NMDARs as well as SHANK3 expression in both the basal and lateral amygdala (BLA) of Tbr1+/− mice. Limitations The therapeutic effect of dietary zinc supplementation observed in rodent models may not reproduce the same effects in human patients. The effect of dietary zinc supplementation on synaptic function in other brain structures affected by Tbr1 haploinsufficiency including olfactory bulb and anterior commissure will also need to be examined. Conclusions Our data further the understanding of the molecular mechanisms underlying the effect of dietary zinc supplementation and verify the efficacy and breadth of its application as a potential treatment strategy for ASD.


Background
Autism spectrum disorder (ASD) is a prevalent neurodevelopmental disorder characterised by social and communication deficits, excessive repetitive behaviours, sensory abnormalities, and cognitive dysfunction [1,2]. Although the full understanding of the aetiology of ASD has not yet been established, a diverse range of genetic aberrations and environmental factors are thought to contribute to the heterogeneity of behavioural traits observed in children with ASD [3]. In recent years, nextgeneration whole-genome sequencing has identified hundreds of deleterious variants in 10-30% of ASD patients [4,5]. Among the recurrent ASD-associated high-confidence risk genes, T-brain-1 (TBR1) has been of particular interest because it encodes a neuron-specific transcription regulator of the T-box family that modulates cortical development and brain wiring [6]. To date, genetic aberrations in one allele of TBR1 have been discovered in ASD patients who commonly display socialisation impairments, defective language skills, and intellectual disability [7][8][9][10][11][12][13]. Consistently, mouse models of Tbr1 haploinsufficiency (Tbr1 +/− ) or human TBR1 mutation knock-in (Tbr1 +/K228E ) also display ASD-like behavioural deficits [14][15][16], and thus, serve as good models to further our understanding of the pathogenic mechanisms of ASD induced via de novo TBR1 mutations.
d-cycloserine, a partial agonist of NMDARs, has been examined as a therapeutic agent in Tbr1 +/− mice to correct the reduced neuronal activation and NMDAR hypofunction that are proposed to be the pathogenic cellular mechanisms underpinning ASD-like behaviours [14,15]. Systemic application or local infusion of d-cycloserine effectively restored neuronal activation in both amygdalae and olfactory bulb and reversed ASDlike behaviours induced by Tbr1 haploinsufficiency [14,15]. Interestingly, trans-synaptic zinc mobilisation via clioquinol, a zinc chelator and ionophore, has also been shown to successfully rescue social interaction deficits and NMDAR hypofunction in Tbr1 +/− mice [37]. Zinc is a prevalent metal in the brain that modulates synaptic function and plasticity via interaction with neuronal ion channels, transporters, receptors, and synaptic structural proteins [40]. Increased zinc levels have been shown to promote synaptic protein scaffold formation, facilitating synapse development and maturation [41][42][43][44]. We have previously shown that dietary zinc supplementation for 6 weeks post-weaning can prevent ASD-associated behavioural deficits and adjust the structure and function of glutamatergic cortico-striatal synapses in Shank3 deficient mice [45]. Here, we have evaluated the therapeutic breadth of dietary zinc supplementation beyond ASDlinked Shank3 mutations by examining its effectiveness in Tbr1 +/− mice. Specifically, we show that dietary zinc elevation can restore social interaction, fear memory, and functional thalamic-amygdalar synapses in the Tbr1 +/− mice, thereby broadening the applicability of dietary zinc supplementation for ASD treatment.

Animals
All animal manipulations and experiments were performed under regulations approved by the University of Auckland Animal Ethics Committee and in adherence to the ARRIVE guidelines. The Tbr +/− mice were provided by Y.-P. Hsueh (Institute of Molecular Biology, Academia Sinica, Taipei, Taiwan) [14,15] with permission of Dr. John Rubenstein (University of California, San Francisco) and maintained at the Vernon Jansen Unit animal facility at the University of Auckland, Auckland, New Zealand. Wildtype Tbr +/+ (WT) and heterozygous Tbr +/− mice were generated from Tbr +/− male x WT female breeding pairs. All experimental animals were housed under a standard 12/12-h light-dark cycle. Offspring were earpunched for genotyping and identification at postnatal day 10, and were kept with the dam until weaning at postnatal day 21. After weaning, animals were housed in groups of 2-4 per individually ventilated cage with mixed genotypes. Food and water were available ad libitum.

Experimental animals
At weaning, animals were randomly assigned to a control zinc diet (30 ppm [parts per million] zinc; D19410B; Research Diets, Brunswick, NJ, USA) or a zinc supplemented diet (150 ppm zinc; D06041101; Research Diets, Brunswick, NJ, USA), for 6-8 weeks. These egg white-based diets were identical in composition except for their zinc levels. As we have previously shown in Shank3 ex13-16−/− mice [45,46], which have the same C57BL/6/J background, no adverse effects on animal health, weight, or development were evident on either diet. No significant differences were observed in body weight for all Tbr1 genotypes: Male mice at 9 weeks: WT 30ppm , 22 Tbr +/− : immunohistochemistry-5 males and 5 females; electrophysiology-5 males and 9 females; behavioural experiments-25 males and 23 females). The animals were randomly grouped and were used independently for behavioural, electrophysiological, or imaging experiments. All experiments and analyses were conducted with the experimenter blinded to genotype and zinc diet by independent animal coding with a unique identification number at weaning.

Behavioural tests and analysis
All behavioural tests were performed in an isolated room under the light cycle, and the animals were habituated in the room for at least one hour before testing. Except for the auditory fear conditioning test, a ceiling camera DFK21AF04 (SDR Scientific) was used for recording and monitoring animal behaviour. EzTrack (Open source) [47] was used for generating heat maps and tracking animal movement. The grooming test arena, dark-light chamber, and three chamber apparatus were washed with 70% ethanol and then 3% acetic acid between each behavioural trial. The same cohort of animals underwent the grooming, light and dark, social interaction, and social novelty tests, respectively, over a period of 1-2 weeks. No two behavioural tests were performed on the same day. A separate cohort was used for the auditory fear conditioning test so that social behaviour was not influenced by fear conditioning.

Auditory fear conditioning test
Auditory fear conditioning was performed in operant chambers (17 × 17 × 25 cm 3 , Ugo Basile) within soundattenuating boxes (Ugo Basile) illuminated by a 7.5 W white light and constant white noise of 60 dB playing in the background through built-in speakers. Animal motion was recorded with the built-in camera, and the freezing behaviour was analysed by EthoVision XT Version 12.0 (Noldus). The chambers were cleaned with 3% acetic acid (training days and day 0) or 70% ethanol (day 1). Each mouse was placed in the operant chamber and habituated for 8 min on training day. On day 0, each mouse was placed into the operant chamber for 4 min, and then 3 auditory conditioned trials were performed, comprising of a tone (2 kHz; 80 dB; 18 s) followed by an electric foot shock (0.6 mA; 2 s) in 1 min intervals. Freezing responses of the mice immediately after the third electric foot shock over 60 s were measured as "after shock". Mice were returned to their home cages after the training. On day 1, mice were placed in a novel operant chamber, and forty tones (80 dB; 2 kHz; 20 s) were applied at 5 s intervals. Time spent freezing was averaged over the first ten auditory stimulations [14].

Grooming behaviour
Under red light conditions [15 lx], each mouse was placed in a cylindrical arena [17 cm radius] and habituated for 10 min. The behaviour of the mouse was then recorded, and the grooming behaviour was analysed over 30 min. Grooming behaviour included licking, wiping, and rubbing any body parts [45,46].

Dark-light emergence test
Each mouse was placed in a dark chamber (001 lx) for 5 min. The door between the dark and light chambers was opened enabling the mouse to freely explore the chambers for 10 min. The latency to enter the light chamber (275 lx), transition number to enter the light chamber, and the time spent in the light chamber were measured.

Three-chamber social interaction test
The three chamber apparatus was used under low light conditions (15 lx) and conducted as previously described [14,37,45,46]. Each mouse was placed in the middle chamber, and a mesh container was placed in both the left and right chambers. The doors between the chambers were then opened to enable the mouse to explore all three chambers. After 10 min, the mouse was gently guided to the centre chamber, and the doors were closed. In phase 2, a sex and age-matched stranger mouse (Stranger 1: S1) was placed in one of the containers, while the other chamber's container remained empty (Empty cup: E). The test mouse was then allowed to explore freely for 10 min. The mouse was again gently guided to the centre chamber, and the doors were closed. In the last phase of the test, another sex and age-matched stranger mouse (Stranger 2: S2) was introduced in the previously empty container. The mouse was then allowed to explore for 10 min in all three chambers. The social interaction reference index was calculated by the time spent in close interaction with S1 subtracted by time spent in close interaction with the empty cup in the chamber. The social novelty reference index was calculated by the time spent in close interaction with S2 subtracted by time spent in close interaction with S1.

Whole-cell patch-clamp recordings
Only brain slices with clearly defined thalamic afferents crossing the dorsolateral region of the LA, a converging location of auditory and somatosensory inputs, were utilised for whole-cell patch-clamp experiments. Slices were transferred to the recording chamber, superfused with carbogenated recording aCSF: (in mM) 119 NaCl, 2.5 KCl, 1 Na 2 HPO 4 , 1.3 MgSO 4 , 26.2 NaHCO 3 , 11 D-( +)-glucose, 2.5 CaCl 2 , pH 7.4, 305-310 mOsm at RT at 2-3 mL/minute. The principal neurons of the LA were visualised using infrared differential interference contrast (IR-DIC) optics with a 40× water immersion objective lens mounted on an Olympus BX-51 microscope (Olympus Corporation, Japan). A platinum-iridium concentric bipolar stimulating electrode was placed onto the thalamic afferent pathway to induce presynaptic stimulation of the principal neurons of the LA. Stimulation was performed with a constant isolated current stimulator (Model DS3; Digitimer, USA) with a duration of 500 μs at 0.05 Hz frequency. Whole-cell patch-clamp recordings of LA principal neurons were acquired with glass recording electrodes (borosilicate tubing filamented glass electrodes, BF150-86-7.5; Sutter Instrument Company, USA) of 5-7 MΩ resistance pulled by a vertical electrode puller (PC-10; Narishige, Japan), and filled with an internal solution: (in mM) 120 Cs gluconate, 40 HEPES, 5 MgCl 2 , 2 NaATP, 0.3 NaGTP and 5 QX314, pH 7.2, 298 mOsm. The distance between the stimulating electrode and recording electrode was approximately 200 μm. Electrophysiological signals were amplified (Multiclamp 700B; Axon Instruments, USA) and digitised (Digidata 1550B; Axon Instruments, USA). Events were sampled at 20 kHz and low-pass filtered at 2 kHz. All data were obtained and analysed using pClamp 10 acquisition software and Clampfit 10, respectively (Axon Instruments, USA). Series resistance (Rs) was < 25 MΩ. Rs was monitored before and after each recording paradigm, and any recordings with Rs variation greater than 20% were discarded from the analysis. Overall, we observed no differences across experimental groups in membrane capacitance ( To measure α-amino-3-hydroxy-5-methyl-4isoxazolepropionic acid receptor (AMPAR)-mediated excitatory postsynaptic currents (EPSCs), LA principal neurons were voltage-clamped at -70 mV and stimulated with an intensity that was adjusted to provide a half-maximal amplitude. AMPAR-mediated EPSCs were measured for 30 sweeps at 0.05 Hz. Only the peak amplitudes of monosynaptic EPSCs relative to the prestimulation baseline were used for data analysis. For paired-pulse ratio (PPR) and NMDAR/AMPAR ratio experiments, the stimulation intensity was modified to the level that stably evoked 200-350 pA AMPAR-mediated EPSCs. For paired-pulse stimulation, two consecutive stimuli separated by 50 ms were delivered to LA principal neurons (voltage-clamped at − 70 mV, 30 sweeps, 0.05 Hz). Only recordings that displayed stable evoked first EPSCs were included for data analysis. PPR was determined by dividing the peak amplitude of the second AMPAR-mediated EPSC response by that of the first peak current amplitude. For NMDAR/AMPAR ratio experiments, LA principal neurons were first voltage-clamped at -70 mV, and 30 consecutive AMPARmediated EPSCs at a stimulus intensity that stably evoked 200-350 pA current responses were recorded (0.05 Hz) to ensure consistent presynaptic stimulation and AMPAR-mediated responses when measuring NMDAR EPSCs. Then, 10 μM CNQX was bath applied to block AMPAR-mediated current responses, and the holding potential was switched to + 40 mV to record NMDAR-mediated EPSCs (30 sweeps, 0.05 Hz). The NMDAR/AMPAR ratio was calculated as the average peak amplitude of NMDAR-mediated EPSCs divided by the average peak amplitude of AMPARmediated EPSCs. Picrotoxin (100 µM) was included in the recording aCSF to block GABA A receptor-mediated inhibitory currents during stimulation. Decay kinetics of the NMDAR-mediated responses were measured by fitting a double exponential function to the decay phase of normalised NMDAR-mediated EPSCs (Clampfit 10) [46,48]. To compare decay times between different experimental groups directly, we calculated a weighted mean decay time constant: , where I f and I s represent the amplitudes of the fast and slow decay components, respectively, and τ f and τ s represent fast and slow decay time constants, respectively.

Luxol fast blue staining
After dietary supplementation with normal (30 ppm) and increased (150 ppm) zinc for 6-8 weeks, WT and Tbr1 +/− mice were euthanised with CO 2 and transcardially perfused with 20 mL of ice-cold 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (pH 7.4). The brains were collected and postfixed in 4% PFA for 24 h at 4 °C and then incubated in 30% sucrose in 0.1 M phosphate buffer for 72 h at 4 °C. Coronal brain sections were prepared at 50 µm thickness. Brain sections from 7 animals were used for each group (i.e. WT 30ppm vs. Tbr +/− 30 ppm vs. WT 150ppm vs. Tbr +/− 150 ppm ). Brain sections were placed in alcohol/chloroform solution and then rinsed with 70% ethanol, 5 min × 3 times. The sections were then placed in 0.1% Luxol fast blue solution overnight at 56 °C and were destained with 0.05% lithium carbonate solution. Each whole brain section was imaged with a Leica MZ6 stereomicroscope using a Leica Flexacam C1 camera at 1.6X magnification.

Confocal imaging and image analysis
GluN1, Shank2 or Shank3 and synapsin1/2 immunostaining in the LA and basal amygdala (BA) was imaged via confocal microscopy (FV1000; Olympus Corporation, Japan) at 63X magnification (UPLSAPO, 1.35 NA) with 3X digital zoom using FluoView 4.0 image acquisition software, yielding images with a size of 512 × 512 pixels (0.138 µm/pixel, both x and y). Images were taken sequentially, and two counts of the line Kalman integration method were applied. For each section, z-stacks were obtained (30-35 images taken at 0.3 µm apart) from a minimum of two regions in the LA and BA per brain slice, and a minimum of three brain slices were imaged per brain. Imaging parameters including laser power, amplifier gain, and offset, were optimised for each antibody and kept consistent for all subsequent imaging to enable direct comparisons between images of different genotype and zinc diet groups. ImageJ software (NIH, USA) was used for puncta-by-puncta and colocalisation analysis. The postsynaptic GluN1, Shank2 or Shank3, and presynaptic synapsin1/2 immunolabelling was corrected for background intensity variation by subtracting a Gaussian (σ = 3 pixels) blurred version of the substack. The 3D Objects Counter tool in ImageJ was then utilised to analyse the number and intensity of GluN1, Shank2 or Shank3, and synapsin1/2 puncta and identify their colocalisation in a 3-dimensional space captured by the z-stack. Only GluN1, Shank2 or Shank3, and synapsin1/2 puncta colocalised in each z-plane were counted as colocalised. GluN1, Shank2 or Shank3 puncta that colocalised with synapsin1/2 were defined as synaptic to ensure data are directly relevant to changes occurring at synapses. For each data set, the image analysis criteria including intensity threshold and detection voxel size range were kept consistent for each experimental set, regardless of the genotype and diet manipulation. The intensity and puncta density (number of puncta per 10,000 µm 3 ) measurements were compared between different groups by normalising to the average intensity and puncta density values, respectively, calculated from images from the WT 30 ppm dietary zinc group.

Statistical analysis
All data are presented as mean ± standard error of the mean (SEM). Statistical analyses were conducted using Graphpad Prism 9.0, with a p value < 0.05 considered significant. The D' Agostino and Pearson normality test was performed to assess normality and homogeneity of variances in the data. Statistical significance was determined by two-way analysis of variance (ANOVA) with Šídák's post hoc test, one-way ANOVA with Tukey's post hoc multiple comparisons, or nonparametric Kruskal-Wallis test, depending on the normality of data distribution and data comparison. Significant results are marked with * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001. Details of each statistical test for each data set, experimental 'n' number of neurons, images, and animals are provided in the figure legends.

Results
To examine the effects of dietary zinc supplementation on ASD-related behaviours and synaptic deficits beyond Shank3-ASD mutations, we fed wildtype Tbr1 +/+ (WT) and heterozygous mutant Tbr +/− mice control (30 ppm) or high (150 ppm) dietary zinc in their normal chow for 6-8 weeks after weaning [45]. This was followed by behavioural testing as well as glutamatergic excitatory synaptic transmission and immunocytochemical analysis in the amygdala at 9-11 weeks of age.

Auditory fear memory deficits in Tbr1 +/− mice are rescued by dietary zinc supplementation
Previously it has been shown that one of the profound behavioural deficits observed in mice with haploinsufficiency in Tbr1 (i.e. Tbr1 +/− ) is the impairment in fear conditioning [14]. Therefore, we performed an auditory fear conditioning test to examine the efficacy of our zinc supplementation strategy on this ASD phenotype (Fig. 1A). As expected [14], auditory fear conditioning behaviour was significantly altered in Tbr1 +/− mice. Specifically, Tbr1 +/− mice fed the control 30 ppm zinc diet showed a significant deficit in the percentage of time freezing in response to the auditory tone compared with WT mice fed the control 30 ppm diet (WT 30ppm , 64.27 ± 4.70%; Tbr +/− 30 ppm , 46.67 ± 3.64%; WT 30ppm vs. Tbr +/− 30 ppm , p value = 0.0060; Fig. 1B), reflecting a deficit in fear memory. Dietary supplementation with 150 ppm zinc in WT mice did not induce any significant change in fear memory (WT 150ppm , 69.23 ± 7.48%; Fig. 1C). However, in Tbr1 +/− mice fed 150 ppm dietary zinc, fear memory was significantly improved, resulting in no significant difference between WT and Tbr1 +/− mice in fear memory (Tbr +/− 150 ppm ,56.43 ± 6.44%; WT 150ppm vs. Tbr +/− 150 ppm , p value = 0.2867; Fig. 1C). Increasing dietary zinc can therefore remove the deficit in fear memory in mice expressing the ASD-associated Tbr1 mutation.

Fig. 2
High dietary zinc rescues social interaction deficits in Tbr1 +/− mice. A Heat maps representing example movements of animals during the social interaction test. B Both WTmice and heterozygous mutant Tbr1 +/− mice spend significantly more time closely interacting with stranger 1 mouse (S1) than with the empty cup in the empty chamber (Empty: E), whether fed with the normal zinc diet (30 parts per million) or the high zinc diet. C Tbr1 +/− mice fed the normal zinc diet show a significant deficit in preference index (measured as the difference in time spent between S1 and Empty). The preference index is restored to that of WT mice when the Tbr1 +/− mice are fed with the high zinc diet. D Heat maps representing the example movements of animals during the social novelty test. E Dietary zinc supplementation did not rescue social novelty recognition deficit in Tbr1 +/− mice, measured by the time spent in closely interacting with familiar stranger 1 (S1) mice or novel stranger 2 (S2) mice. F The social preference index derived from these results, is not different between WT mice and Tbr1 +/− mice and this is not changed by dietary zinc supplementation. Each point represents individual mice in each experimental group (Sample size: WT 30ppm = 11, Tbr +/− 30 ppm = 11, WT 150ppm = 9, Tbr +/− 150 ppm = 11). All data represent mean ± standard error of the mean, analysed using two-way analysis of variance (ANOVA) with Šídák's post hoc test and one-way ANOVA with Tukey's post hoc test. ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001

Tbr1 +/− mice do not display ASD-related anxiety behaviours nor excessive grooming
Next, we examined other ASD-associated behaviours such as heightened anxiety traits and excessive grooming in Tbr1 +/− mice (Fig. 3). The dark-light emergence anxiety test revealed no significant differences between the WT controls and Tbr1 +/− mice in either the latency to enter the bright arena nor time spent in the bright arena ( Fig. 3D), and this was also not significantly altered with the high zinc diet (WT 150ppm , 8.32 ± 1.25%; Tbr +/− 150 ppm , 5.94 ± 0.56%; WT 150ppm vs. Tbr +/− 150 ppm , p value = 0.5467; Fig. 3D). These results were not surprising as previous work has also described unaltered grooming behaviours and normal anxiety phenotypes measured by open field or elevated plus maze test, respectively, in Tbr1 +/− mice, when compared to the WT mice [14].

High zinc diet supplementation restores presynaptic function at glutamatergic thalamic-LA synapses in Tbr1 +/− mice
We also examined AMPAR-mediated paired-pulse ratio (PPR) in principle neurons of the LA to assess presynaptic function in WT and Tbr1 +/− mice. The ratio of the first to second AMPAR EPSC amplitudes (separated by 50 ms) was measured in all four experimental groups. In the control dietary zinc group, we observed a significant increase in PPR in Tbr1 +/− mice compared with WT mice (WT 30ppm , 0.7725 ± 0.0257; Tbr +/− 30 ppm , 0.9348 ± 0.0388; WT 30ppm vs. Tbr +/− 30 ppm , p value = 0.0039; Fig. 5A and B), supporting a lower release probability occurring at thalamic-LA synapses in these ASD mutant mice. Interestingly in mice fed the high zinc diet, there was no significant difference in PPR between Tbr1 +/− mice compared with WT mice (WT 150ppm , 0.8486 ± 0.0340; Tbr +/− 150 ppm , 0.8370 ± 0.0278; WT 150ppm vs. Tbr +/− 150 ppm , p value = 0.9953; Fig. 5A and B), showing that increased dietary zinc levels return the PPR to control level, suggesting a return to normal presynaptic function.

High dietary zinc restores synaptic GluN1 and Shank3, but not Shank2, puncta density in the amygdala of Tbr1 +/− mice
To examine whether dietary zinc supplementation normalises thalamic-LA synapse function in Tbr1 +/− mice via restoration of inter-amygdalar axonal projections, we performed Luxol fast blue staining of brain sections in WT and Tbr1 +/− mice (n = 7/group; example in Fig. 6). Luxol fast blue staining clearly showed the significant reduction in inter-amygdalar axonal projections in Tbr1 +/− mice, however we observed no improvement in these impaired axonal projections in Tbr1 +/− mice with the increase in dietary zinc (Fig. 6). Next, we performed immunocytochemical analysis of the GluN1 subunit of the NMDAR in the LA to validate whether changes in the synaptic expression of NMDARs contribute to our behavioural and electrophysiological differences observed in WT and Tbr1 +/− mice (Fig. 7A-E). Supporting our electrophysiological data (Fig. 4), we observed that synaptic NMDARs, as measured by colocalised GluN1 and synapsin1/2 puncta, were significantly decreased in the LA of Tbr1 +/− mice compared with WT mice fed with normal 30 ppm zinc diet (WT 30ppm , 1.0000 ± 0.1304; Tbr +/− 30 ppm , 0.5776 ± 0.0413; WT 30ppm vs. Tbr +/− 30 ppm , p value = 0.0144; Fig. 7B and C). Moreover, high dietary zinc restored the deficits in synaptic GluN1 density in the LA of Tbr1 +/− mice, such that the synaptic GluN1 densities measured from Tbr1 +/− mice fed 150 ppm zinc were comparable with WT mice fed with the normal zinc diet (WT 150ppm , 0.6610 ± 0.1014; Tbr +/− 150 ppm , 0.9583 ± 0.0858; Tbr +/− 30 ppm vs. Tbr +/− 150 ppm , p value = 0.0041; WT 30ppm vs. Tbr +/− 150 ppm , p value > 0.9999; Fig. 7B and C). We also conducted immunocytochemical analysis of the GluN1 subunit of the NMDAR in the basal nuclei of the amygdala (BA), the major target for LA pyramidal neurons that are involved in the expression and extinction of conditioned fear memory [52]. As observed in the LA, the synaptic density of GluN1 in mice fed normal dietary zinc levels was significantly reduced in the BA of Tbr1 +/− mice compared with the WTmice (WT 30ppm , 1.0000 ± 0.0661; Tbr +/− 30 ppm , 0.5206 ± 0.0432; WT 30ppm vs. Tbr +/− 30 ppm , p value = 0.0015; Fig. 7D and E). In the BA, dietary zinc supplementation also reversed the reduced synaptic GluN1 density in Tbr1 +/− mice (WT 150ppm , 0.8009 ± 0.0920; Tbr +/−  Fig. 8C and D). However, interestingly we did observe a significant increase in Shank3 synaptic expression in both the LA and BA in Tbr1 +/− mice fed high dietary zinc, such that Shank3 levels were returned to WT levels (Shank3 LA:  Each point represents an individual region of interest image in each experimental group. Scale bar = 5 µm. All data represent mean ± standard error of the mean, analysed using one-way ANOVA with Tukey's post hoc test. ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 these data show that a deficit in synaptic NMDAR expression occurs in both the basal and lateral nuclei of the amygdala (BLA) in Tbr1 +/− mice which likely underpins the decrease in NMDAR-mediated currents and the NMDAR/AMPAR ratio, and that this deficit can be rescued by dietary zinc supplementation in Tbr1 +/− mice, in part, through Shank3-related mechanism.

Discussion
Here, we show that dietary zinc supplementation prevents impairments in auditory fear memory and social interaction, but not social novelty, in the Tbr1 +/− mouse model of ASD. At the synaptic level, dietary zinc supplementation also reversed AMPAR and NMDAR hypofunction and normalised presynaptic function at glutamatergic thalamic-LA synapses that are crucial for auditory fear memory [49,50], and restored synaptic density of the GluN1 subunit essential for functional NMDARs. Together, our data further the understanding of the molecular mechanisms underlying the effect of dietary zinc supplementation and verify the efficacy and breadth of its application as a potential treatment strategy for ASD.
Zinc supplementation is a viable therapeutic approach beyond Shank mutations for ASD phenotypes stemming from the amygdala.
The concept of zinc supplementation as a potential ASD treatment has been based on observations that (1) reduced serum zinc levels have been identified in individuals with ASD [53][54][55][56][57], and (2) prenatal zinc-deficient animals display ASD-like behaviours including increased anxiety, social interaction deficits and altered ultrasonic vocalisation [58,59]. The efficacy and mechanistic underpinnings of zinc supplementation as a treatment strategy have so far been primarily examined in rodent models carrying ASD-associated SHANK2 or SHANK3 mutations, but also in the maternal immune activation model of ASD [37,41,44,45,[59][60][61][62]. A focus on SHANK models is likely due to (1) the high prevalence of SHANK2 and SHANK3 mutations in people affected by ASD [63]; (2) SHANK2 and SHANK3, which are excitatory glutamatergic postsynaptic proteins, contain a SAM domain that directly binds zinc [41,43,64]; (3) synaptic localisation and expression of SHANK2 and SHANK3 are highly dependent on zinc [42][43][44][64][65][66]; and (4) zinc deficiency dysregulates the postsynaptic SHANK scaffold, which is thought to contribute to the synaptic mechanisms underlying behavioural deficits in ASD [41,44,58]. The data described in the current study is one of the few that have identified zinc supplementation as a viable therapeutic approach beyond ASD-associated Shank mutations and provided potential mechanisms underlying its beneficial effect [37,67], specifically in the amygdala. Together with previous studies [37,41,44,45,[59][60][61][62], this shows impairment of zinc sensitive pathways may be a shared biological substrate for ASD behaviours across several ASD models.
The amygdala forms intricate neural circuits with other brain regions such as the prefrontal cortex, thalamus, hypothalamus, hippocampus, and striatum, and plays a principal role in emotional memory and social behaviours [68,69]. Defective inter-and intra-amygdalar axonal projections and impaired neuronal activation in the amygdalae are major phenotypes observed in Tbr1 +/− mice, and consistent with this, amygdala-dependent behaviours such as social interaction, social novelty recognition, and auditory fear memory are impaired in Tbr1 +/− mice [14,37]. In particular, thalamic-LA synapse function is significantly impaired, likely contributing to auditory fear memory deficits observed in Tbr1 +/− mice. We observed that the amplitudes of both AMPAR-and NMDAR-mediated EPSCs were decreased, and that presynaptic function (as measured by PPR) was altered at thalamic-LA synapses in Tbr1 +/− mice. Also, the synaptic puncta density of GluN1 and Shank3, was reduced in Tbr1 +/− amygdalae, indicating that diminished synaptic expression of (See figure on next page.) All data represent mean ± standard error of the mean, analysed using nonparametric Kruskal-Wallis test. ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.001 Shank3 proteins and NMDARs contributes to NMDAR hypofunction at Tbr1 +/− thalamic-LA synapses. A critical result is that dietary zinc supplementation can restore excitatory glutamatergic pre-and postsynaptic function at Tbr1 +/− thalamic-LA synapses, as well as the synaptic GluN1 and Shank3 density, comparable to that of WT mice. Together our data suggest that normalisation of thalamic-LA synapse function in Tbr1 +/− mice is a central mechanism behind the dietary zinc-induced prevention of auditory fear memory impairment.

Potential zinc-related pathways involved
The majority of zinc ions are extensively bound within proteins as structural or catalytic cofactors in the brain, while a pool of free (or chelatable) zinc is highly localised within synaptic vesicles, accumulated by the vesicular zinc transporter ZnT3, at glutamatergic nerve terminals [40,70,71]. In the amygdala, only the excitatory cortical-amygdalar synapses implicated in auditory fear conditioning contain vesicular zinc. In contrast, thalamic projections to the LA involved in auditory fear conditioning lack free zinc, as demonstrated by minimal expression of the ZnT3 gene [72]. Therefore, the high dietary zinc-induced restoration of glutamatergic thalamic-LA synapse function in the Tbr1 +/− mice is unlikely to be caused by the enrichment of free zinc, but rather by changes in the availability of protein-bound zinc at these synapses. Further analysis measuring total zinc levels by inductively coupled plasma mass spectrometry, and free or intracellular zinc with zinc indicators (e.g. TFL-Zn or ZnAF-2DA, respectively) [37] would be required in the amygdala to assess this possibility.
At excitatory glutamatergic synapses, bound zinc maintains the organisation of a complex assembly of postsynaptic density proteins where it associates with SAP102, SHANK2, and SHANK3 [43,64,73]. The exact role of the zinc-binding motif of SAP102 remains elusive [73], although it may link to its trafficking of AMPARs and NMDARs during synaptogenesis to regulate synaptic organisation and plasticity [74][75][76]. Conversely, much more is known about zinc's role in driving synaptic localisation and expression of SHANK2 and SHANK3, and subsequently AMPARs and NMDARs [41,42,44,45,59,77,78]. In Cttnbp2 −/− ASD mice (independent of Shank ASD mutations), zinc supplementation induced a significant increase in synaptic expression of SHANK2 and SHANK3, as well as NMDAR (GluN1 and GluN2B) [67]. Our observed zinc-induced rescue of glutamatergic receptor function and GluN1 puncta density in Tbr1 +/− mice appears to be SHANK3, but not SHANK2, related, suggesting that dietary zinc-induced increases in SHANK3 can drive NMDAR expression at synapses in Tbr1 +/− mice. Moreover, SHANK3 can regulate glutamatergic presynaptic function via trans-synaptic signalling through the neurexin-neuroligin cell adhesion molecules to recruit presynaptic structural proteins and increase transmitter release, which could further contribute to our observed dietary zinc-related changes in presynaptic and postsynaptic function in Tbr1 +/− mice [79]. In addition, zinc-binding proteins are located in the presynapse, including Piccolo, Bassoon, Munc13, and RIM, that control the priming, tethering, and exocytosis of synaptic vesicles, and thereby determine presynaptic release probability [80][81][82]. Dietary zinc supplementation may drive zinc-binding to presynaptic proteins together with zinc-dependent recruitment of SHANK protein complexes pre-and postsynaptically to modify vesicular release probability and prevent abnormal PPR observed in the Tbr1 +/− mice. Altogether, the supplemented zinc may exert its therapeutic effect through association with zinc-responsive proteins at both pre-and postsynapses that synergistically work together to repair glutamatergic thalamic-LA synapse structure and function.

Timing and duration are critical in dietary zinc supplementation.
In Tbr1 +/− mice, dietary zinc supplementation prevents impairments in auditory fear memory and social interaction, but not social novelty recognition impairments. This selective therapeutic effect of dietary zinc supplementation may be attributed to (1) differences in brain circuitry and regions involved in distinct social behaviours [83,84], (2) discrete zinc responsiveness in different neural pathways [72,85], and (3) timing of zinc supplementation [86]. The detrimental behavioural phenotypes induced by Tbr1 haploinsufficiency likely manifest from early stages in brain development [17,20,[24][25][26]. Indeed we have recently shown that both social interaction and social novelty deficits are present in Shank3 −/− mice as early as 3 weeks of age, indicating that the neural circuitry responsible for social behaviours is established before weaning [46]. In the current study, dietary zinc supplementation was performed post-weaning and our data suggest this is after the "window" to modulate this circuitry has closed. Indeed, our Luxol fast blue staining supports that some structural deficits, such as in the anterior commissure, are not rescued by dietary zinc supplementation. Similarly adult restoration of Shank3 expression in 2-to 4.5-month-old Shank3 −/− mice only partially prevented ASD-associated behavioural phenotypes [87], and clioquinol-induced increases in free zinc in 2-to 4-month-old Shank2 −/− mice recovered social interaction deficits but did not restore defective social novelty recognition behaviour nor normalise heightened anxiety [37]. Together with our data, these studies suggest that specific ASD behaviours are more difficult to alter as development and maturation progresses, and therefore the timing of treatment is an essential factor. The underpinning neural pathways do appear to have a window of modulation during early development however, as supported by the intriguing observation that dietary maternal zinc supplementation during pregnancy and lactation can prevent ASD-associated social behaviour deficits developing in 3-week-old Shank3 −/− offspring, and this persists into adulthood [46]. Timing therefore appears to be critical for zinc-based therapies for ASD-associated mutations in genes such as Tbr1 and Shanks that encode proteins involved in early brain development.
Our dietary zinc treatment strategy, in which animals receive dietary zinc in their food for a minimum of 6 weeks [45,46], induces not only functional but also structural changes at the synaptic level, and effects on ASD behaviours can be observed 2-3 months after zinc supplementation [46]. In contrast, zinc supplementation for a shorter period through the drinking water for 7 days in Cttnbp2-deficient mouse does not restore decreased spine density in the hippocampus, and the initial rescue of social interaction behaviours deteriorated 1 week after discontinuation of zinc [67]. Therefore, not only the timing but also the duration of the zinc treatment could be pivotal in the persistence of synaptic remodelling and subsequent behavioural outcomes. Another notable feature of the dietary zinc supplementation approach in this study is that when used at a physiological level (30 ppm or 150 ppm) [88,89], the increased zinc levels did not influence behaviours in the WT mice. This is consistent with previous findings that WT mice fed with elevated dietary zinc displayed normal behaviours regardless of the timing, duration, or method of delivery [45,46,67], even in the presence of small changes at the molecular or synaptic levels.

Rationale and translational potential of dietary zinc supplementation.
The supplemented dietary zinc level of 150 ppm was used based on our previous studies examining the effects of supplemented zinc diets to ensure a direct comparison was possible across ASD model genotype effects [45,46]. We have previously identified the large variance in zinc levels in rodent chow ranging from 25 ppm to 120 ppm [45], resulting in a large "normal" range. How this translates to human dosage is difficult to directly compare. ASD children show lower serum zinc levels, and a recent study showed an increase in cognitive-motor performance following daily zinc supplementation for 12 weeks in ASD children with daily zinc dose to each child equal to their body weight in kg plus 15-20 mg [90]. Nevertheless, a conservative approach is required to discover the effective dosage for ASD treatment. The therapeutic window of zinc concentration is narrow, and zinc can influence the metabolism of other metal ions, such as copper [91][92][93][94][95]. Moreover, zinc does not cross the blood-brain barrier (BBB) freely [61], yet the recent development of nano-particle carriers to aid BBB crossover may prove to be encouraging [96,97]. Together, dietary zinc supplementation has a great therapeutic potential as an ASD treatment strategy, but additional research is required to bring forward safe and effective clinical use.

Limitations
The therapeutic effects of zinc supplementation have only been evaluated in the mouse models of ASD [37,45,46,64], including Tbr1 +/− mice in the current study. Whether zinc-based treatment strategies can induce the same beneficial outcomes in human-derived models and ASD patients, that is reversal of ASD-related behaviours through the restoration of synapse function, requires further examination. An additional challenge will be determining whether ASD diagnosis can precede the therapeutic window in humans. Also, the effects of dietary zinc supplementation on neural circuits other than the thalamic-LA projection pathway such as the olfactory system and anterior commissure-related interhemispheric connectivity that have previously been shown to be dysfunctional in Tbr1 +/− mice [15,38,39], or those that may underlie ASD-associated behaviours observed in Tbr1 +/− mice (i.e. prefrontal cortex and ventral hippocampus for social memory and behaviours) require examination in future studies. We also recognise the current study focussed on the synaptic axis for the therapeutic role of dietary zinc, and it is possible that given the many roles of zinc, that its therapeutic actions are also indirect, for example via gut-brain signalling.

Conclusion
In summary, here we demonstrate the effectiveness of dietary zinc supplementation in the reversal of ASDassociated synaptic and behavioural deficits in the Tbr1 +/− ASD rodent model, highlighting the breadth of this treatment strategy beyond Shank ASD mutations. Zinc supplementation can restore impaired glutamatergic synapse function and density, particularly in the thalamic-LA projection pathway, and prevent ASD-associated behaviours in Tbr1 +/− mice. These data improve our understanding of the mechanisms and the potential of dietary zinc supplementation as a treatment strategy for ASD.