Proteins of the ProSAP/Shank family act as major organizing scaffolding elements within the postsynaptic density of excitatory synapses. Deletions, mutations or the downregulation of these molecules has been linked to autism spectrum disorders, the related Phelan McDermid Syndrome or Alzheimer’s disease. ProSAP/Shank proteins are targeted to synapses depending on binding to zinc, which is a prerequisite for the assembly of the ProSAP/Shank scaffold. To gain insight into whether the previously reported assembly of ProSAP/Shank through zinc ions provides a crossing point between genetic forms of autism spectrum disorder and zinc deficiency as an environmental risk factor for autism spectrum disorder, we examined the interplay between zinc and ProSAP/Shank in vitro and in vivo using neurobiological approaches. Our data show that low postsynaptic zinc availability affects the activity dependent increase in ProSAP1/Shank2 and ProSAP2/Shank3 levels at the synapse in vitro and that a loss of synaptic ProSAP1/Shank2 and ProSAP2/Shank3 occurs in a mouse model for acute and prenatal zinc deficiency. Zinc-deficient animals displayed abnormalities in behaviour such as over-responsivity and hyperactivity-like behaviour (acute zinc deficiency) and autism spectrum disorder-related behaviour such as impairments in vocalization and social behaviour (prenatal zinc deficiency). Most importantly, a low zinc status seems to be associated with an increased incidence rate of seizures, hypotonia, and attention and hyperactivity issues in patients with Phelan-McDermid syndrome, which is caused by haploinsufficiency of ProSAP2/Shank3. We suggest that the molecular underpinning of prenatal zinc deficiency as a risk factor for autism spectrum disorder may unfold through the deregulation of zinc-binding ProSAP/Shank family members.
The three members of the ProSAP/Shank family of scaffolding proteins are key molecules for the spatial organization of the postsynaptic density of excitatory synapses (Boeckers et al., 2002). Recent studies report mutations in ProSAP1/Shank2, ProSAP2/Shank3 and Shank1 associated with autism (Durand et al., 2007; Moessner et al., 2007; Gauthier et al., 2009; Berkel et al., 2010; Pinto et al., 2010; Leblond et al., 2012; Sato et al., 2012). Moreover, the loss of one copy of ProSAP2/Shank3 in humans is correlated with Phelan-McDermid syndrome (22q13 deletion syndrome) (Bonaglia et al., 2001; Phelan et al., 2001; Wilson et al., 2003; Manning et al., 2004). This syndrome is characterized by moderate to profound mental retardation, neonatal hypotonia, global developmental delay, absent or severely impaired speech (Manning et al., 2004), occurrence of seizures and ‘autistic-like’ behaviour. Besides genetic factors, it has been suggested previously that a dysregulation of metal-ion homeostasis might contribute to the pathogenesis of autism spectrum disorders (Curtis and Patel, 2008). Intriguingly, the incidence rate of Zn2+-deficiency is as high as ∼50% for autistic patients between 0 and 3 years of age compared with <1% in a healthy control group (Yasuda et al., 2011).
Recent reports demonstrate a tight interplay between Zn2+-binding of ProSAP1/Shank2 and ProSAP2/Shank3, but not Shank1 (Baron et al., 2006; Gundelfinger et al., 2006; Grabrucker et al., 2011a) and their synaptic association and function (Grabrucker et al., 2011a). Zn2+ is enriched in cortical, striatal and hippocampal regions of mammals and selectively stored in, and co-released with glutamate from presynaptic vesicles (Lin et al., 2001; Wei et al., 2004). It has been argued that Zn2+ released from presynaptic vesicles might influx into the postsynaptic compartment through NMDA and AMPA receptors, and Ca2+-channels (Frederickson et al., 2005) and thereby provide a pool of free chelatable Zn2+. The postsynaptic density has an unusually high Zn2+-content (Jan et al., 2002) and postsynaptic Zn2+ is probably bound to metallothioneins (MTs) as well as to structural proteins such as ProSAP1/Shank2 and ProSAP2/Shank3 (Lee et al., 2003; Grabrucker et al., 2011a). Although indirect evidence for an activity-dependent increase of postsynaptic Zn2+-levels was reported (Bitanihirwe and Cunningham, 2009), it is currently unknown whether the postsynaptic Zn2+-pool is dynamic and whether free Zn2+ can contribute to the folding and synaptic association of ProSAP1/Shank2 and ProSAP2/Shank3. We therefore studied whether the postsynaptic density scaffold at the synapse is regulated by Zn2+-levels, and whether Zn2+-deficiency is accompanied by a dysregulation of the ProSAP/Shank postsynaptic density scaffold in vitro and in vivo. Moreover, we addressed the possibility that this mechanism might constitute a plausible interplay between an environmental risk factor and a known molecular pathway associated with autism spectrum disorder. To that end, we used acute and prenatal Zn2+-deficient animals and performed behavioural analyses testing for autism-related phenotypes. Finally, we investigated, whether Zn2+-deficiency augments specific symptoms observed in patients with Phelan-McDermid syndrome.
Materials and methods
ZnCl2, CaEDTA and TPEN [N,N,N′,N′-tetrakis(2-pyridylmethyl) ethylenediamine], Zinquin ethyl-ester and Zinpyr1 were purchased from Sigma-Aldrich as were all other chemicals unless indicated otherwise. Primary antibodies were purchased from Synaptic Systems (Gephyrin, Homer1, Homer1b/c, GluA1, GluA2, GluA3), Stressgen (bassoon), Abcam (PSD95) Novus Biological [GKAP/SAPAP, Shank1 (immunofluorescence, immunohistochemistry), mGluR5], Sigma [b-actin, GluN1, Shank1 (western blotting)], Santa Cruz (MT-3), Chemicon (MAP2, mGluR5) and Millipore (GluN2A, GluN2B). ProSAP1/Shank2 and ProSAP2/Shank3 antibodies have been described previously (Schmeisser et al., 2012). Secondary Alexa Fluor® conjugated antibodies were purchased from Invitrogen.
Expression constructs and transfection
The pEGFP (C1-3) vector system (Clontech) was used for MT-3 expression constructs and the pLVX vector system (Clontech) for MT-3 knockdown. Short hairpin RNA oligonucleotides were purchased from MWG-Eurofins. Two sequences were used for MT-3 knockdown: AAGGGCTGCAAATGCACGA and CCTGCCCCTGTCCTACTGG. Hippocampal cells were transfected at Day 10 and fixed on Day 14 in vitro using OptiFect™ (Invitrogen).
Cryosections were thawed for 20 min and fixed in paraformaldehyde for 20–30 min. After fixation, sections were washed 3 × 10 min with PBS. Sections were permeabilized with 0.2% Triton in PBS for 2 h and washed again for 3 × 10 min with PBS containing 0.05% Triton. Sections were then placed in 10% foetal calf serum in PBS for 2 h. After blocking, sections were incubated with the primary antibody at 37°C for 2 h before 3 × 10 min washing with PBS containing 0.05% Triton. Sections were incubated with the secondary antibody at 37°C for 1.5 h. After washing 3 × 15 min with PBS containing 0.05% Triton, sections were washed for 2 × 5 min with PBS containing DAPI, rinsed with distilled H2O and mounted with VectaMount™ (Vector Laboratories).
Hippocampal culture, stainings and treatments
Hippocampal cultures were prepared from rat (embryonic Day 18) as described previously (Grabrucker et al., 2009). For fluorescent Zn2+-staining, growth medium was discarded and cells were washed 3 × with Hank’s Balanced Salt Solution. Coverslips were incubated with a solution of 25 µM Zinquin ethyl ester or 5 µM Zinpyr1 in Hank’s Balanced Salt Solution for 40 min at 37°C (27). Immunofluorescence was performed as described previously with minor modifications (Grabrucker et al., 2011a) (Supplementary material). Growth media were supplemented with Zn2+-chelators (TPEN or CaEDTA) or ZnCl2 (10 µM). CNQX (10 µM) and AP5 (50 µM) were used to block synaptic activity. For stimulations, 10 µM ZnCl2, 50 mM KCl (HiK+) or 30 µM glutamate was used. Generation of nitric oxide (NO) was manipulated using the NO agonist spermine nonoate (1 µM) and antagonist N-nitro-l-arginine (100 µM).
Quantitative real-time polymerase chain reaction
Isolation of total RNA from three control and three Zn2+-deficient mice as well as 10 pups from control and Zn2+-deficient mice each was performed using the RNeasy® kit (Qiagen). First strand synthesis and quantitative real-time-PCR amplification were carried out in a one-step, single-tube format using the QuantiFast™ SYBR® Green RT-PCR kit (Qiagen). Thermal cycling and fluorescent detection were performed using the Rotor-Gene® Q real-time PCR machine (model 2-Plex HRM) (Qiagen). The SYBR® Green I reporter dye signal was measured against the internal passive reference dye (ROX) to normalize non-PCR-related fluctuations (Supplementary material). Resulting data were analysed using the hydroxymethylbilane synthase gene as an internal standard to normalize transcript levels. Cycle threshold (ct) values were calculated by the Rotor-Gene® Q Software (version 2.0.2). All quantitative real-time PCR reactions were run in technical triplicates and mean ct-values for each reaction were taken into account for calculations.
Subcellular fractions from mouse brain were isolated as described previously with minor modifications (Schmeisser et al., 2012) (Supplementary material). Proteins were separated by SDS-PAGE and blotted onto nitrocellulose membranes. Immunoreactivity was visualized using horseradish peroxidase-conjugated secondary antibodies and the SuperSignal® detection system (Pierce).
Intrinsic fluorescence spectroscopy
Tryptophan fluorescence emission spectra were recorded for the ProSAP2/Shank3-SAM domain (rat ProSAP2/Shank3 aa1464–1815, Genscript) on a Hitachi F7000 spectrofluorimeter. The quartz cuvette was cleaned with chromic acid and double-distilled H2O. EDTA (100 µM) solution was used to rinse the cuvette followed by double-distilled H2O. One millilitre of protein sample (0.1–0.3 mg/ml) in 50 mM Tris-HCl pH 7.4, 100 mM KCl was used for measurements. Excitation wavelength of 295 nm was used and spectra were recorded in the range of 300–450 nm by subtracting time-dependent reduction of fluorescence emission. All spectra were recorded at room temperature in corrected spectra mode using an excitation and emission band pass of 10 nm and 10 nm, respectively. The response-time was set to 2 s with a scan speed of 240 nm/min. The effects of Mg and Zn standard solutions (Fluka) were studied by titrating them into protein solution after 1–2 min incubation periods.
Ten-week-old mice were purchased from Janvier and housed in plastic cages under the standard laboratory conditions (average temperature of 22°C, food and water available ad libitum). Lights were automatically turned on/off in a 12 h rhythm (lights on at 7 am). After one week of acclimation, mice were divided into two groups, one group (12 females) was fed a Zn2+-deficient diet (4 ppm zinc, TestDiet) with distilled, demineralized drinking water, whereas the control group (12 females) was fed with standard laboratory food (35 ppm zinc) and tap water. To prevent zinc contamination, feeding jars, water bottles and plastic cages were rinsed with HCl and deionized water. After 5 weeks, females of the control and Zn2+-deficient group were mated. For behavioural tests, acute Zn2+-deficient animals were tested as well as their offspring (prenatal Zn2+-deficient) at the same age and compared with control animals. A detailed description of the behavioural tests performed can be found in the Supplementary material. All animal experiments were performed in compliance with the guidelines for the welfare of experimental animals issued by the Federal Government of Germany and the local ethics committee (Ulm University) ID Number: 0.103.
Blood samples from 37 participants diagnosed with Phelan-McDermid syndrome were obtained from Nelson Biological Laboratories at Rutgers University and atomic absorption spectroscopy performed at the ‘ZE klinische Chemie’ of Ulm University Hospital to quantify Cu2+ and Zn2+ levels. Blood samples were mostly from fasting blood donations and participants were evaluated regarding their status on dietary zinc or vitamin supplementations. All experiments were performed in compliance with the local ethics committee (Ulm University, ID: 282/12) and informed consent was obtained from all subjects. Clinical and developmental data of the participants were obtained through the Phelan-McDermid syndrome registry. A total of 26 parameters were evaluated based on the categorization of participants into three groups (no, mild and severe Zn2+-deficiency), including age, gender, diagnosis, birth weight, number of ear infections, possible signs of primary immunodeficiency, occurrence of: hypotonia, seizures, nose and throat problems, scoliosis, dysplastic finger/toenails, eczema, gastrointestinal problems, allergy-related problems, endocrine conditions, abnormal involuntary movements, sensitivity problems, other neurological conditions, abnormal MRI scans, abnormal CAT scans, feeding disorder, anxiety disorder, diagnosed language disorder, and attention deficit disorder/attention deficit hyperactivity disorder or other attention or hyperactivity issues. The overall occurrence of a parameter was calculated and the fraction of each Zn2+ status-group showing the parameter identified. Zn2+ and Cu2+ levels were compared with (whole blood) reference values for age matched healthy controls (age 0–1: 41–58 µmol/l; age 1–4: 40–50 µmol/l; age 4–6: 49–52 µmol/l; age 6–14: 51–60 µmol/l; age 14–18; 50–56 µmol/l; age 18 and up: 66–117 µmol/l). Adult Zn2+ concentrations show no significant correlation with age, although some studies report a decrease in very old age (Rükgauer et al., 1997).
Cell culture experiments
For cell culture experiments, 10 cells of each condition were imaged. For brain sections, three animals of each group (control, Zn2+-deficient) were used and each of the four brain regions (cerebellum, cortex, hippocampus, striatum) imaged by at least three optic fields. Fluorescence images were obtained with an upright Axioscope microscope equipped with a Zeiss CCD camera (16 bits; 1280 × 1024 ppi) using Axiovision software (Zeiss). The signal intensity of all signals within the optic fields was quantified using ImageJ 1.46p. Statistical analysis was performed using Microsoft Excel and tested for significance using t-tests followed by ANOVA.
Western blot data
Evaluation of western blot bands was performed using ImageJ 1.46p. Three independent experiments were performed and the integrated density of bands was measured. All bands were normalized to β-actin and the ratios averaged and tested for significance. The α level of significance was set at 0.05 (*P < 0.05; **P < 0.01; ***P < 0.001).
Normal distribution of data was determined by Shapiro-Wilk test. Two sets of data were compared with t-test, if normally distributed, or Wilcoxon-Mann-Whitney U-test if not normally distributed. Several data sets were compared with one-way ANOVA or ANOVA on ranks and Chi-square test. Although not all data are normally distributed, data are sometimes shown in figures as means ± SEM to facilitate comparisons. Statistical tests were performed using absolute values, despite some diagrams displaying percentages for better comparison. All tests were run by SPSS (version 19) and SigmaPlot software (version 2.0). Statistical tests were two-tailed with a significance level of α ≤ 0.05 (*P < 0.05; **P < 0.01; ***P < 0.001).
The local zinc concentration regulates ProSAP/Shank protein levels at the postsynaptic density
We first investigated whether alterations of Zn2+-concentrations regulate ProSAP1/Shank2 and ProSAP2/Shank3 levels at the synapse. Therefore, we supplemented hippocampal primary neurons for 1 h with 10 μM ZnCl2, which leads to an increase of postsynaptic Zn2+ concentration as evidenced by elevated Zinquin ethyl-ester fluorescence (Coyle et al., 1994) (Supplementary Fig. 1A). Supplementation of neurons with Zn2+ leads to a significant increase in synaptic ProSAP1/Shank2 and ProSAP2/Shank3 immunofluorescence levels (Supplementary Fig. 1B and Fig. 1A and C), which was blocked in the presence of AMPA and NMDA receptor antagonists (Supplementary Fig. 1C). Increasing neuronal activity with 2 min bath application of HiK+ (Fig. 1A–C) or 30 µM glutamate (Supplementary Fig. 1D) induces similar changes in synaptic ProSAP/Shank immunofluorescence, and protein levels in a synapse-enriched P2 fraction after 30 min without increase in overall protein levels (Supplementary Fig. 1F). Application of HiK+ increased synaptic Zn2+ measured by Zinquin (Supplementary Fig. 1E). HiK+ stimulation in presence of Zn2+ leads to a similar increase and does not act in an additive manner hinting towards a mechanism triggered by stimulation similar to those observed with increasing Zn2+ levels (Fig. 1C). To exclude toxic effects of HiK+ stimulation in the presence of Zn2+, we quantified the number of cells per optic field. Our data show no significant increase in cell death (Supplementary Fig. 1G), nor could we detect significantly more swelling or pinching off of dendrites. Most important, bath application of HiK+ in the presence of a Zn2+-chelator (CaEDTA or TPEN) was unable to elicit activity dependent increases in synaptic ProSAP1/Shank2 and ProSAP2/Shank3 levels and the application of CaEDTA or TPEN even led to a decrease in ProSAP1/Shank2 and ProSAP2/Shank3 levels (Fig. 1C). Although CaEDTA is cell impermeable, at the concentration used it will also deplete intracellular Zn2+ (Frederickson et al., 2002). Lower CaEDTA concentrations in contrast inhibited the activity-dependent upregulation of ProSAP1/Shank2 and ProSAP2/Shank3 to a lesser extent (Supplementary Fig. 1H).
An activity-dependent modification of the postsynaptic density by a Zn2+-dependent assembly of ProSAP/Shank proteins requires a modest Zn2+ binding affinity in a range that does not allow multimerization with cytoplasmic free Zn2+ levels, which are in the low picomolar range (Frederickson, 1989; Outten and O’Halloran, 2001; Krezel and Maret, 2006). Thus, to determine the dynamic range in which Zn2+-dependent conformational changes and oligomerization are apparent, we performed intrinsic fluorescence spectroscopy and Dynamic light scattering experiments with the ProSAP2/Shank3 SAM domain (Fig. 1E and F and Supplementary Fig. 1I–L). Taken together, we found that the Zn2+-dependent multimerization of ProSAP2/Shank3 occurs at nanomolar Zn2+-concentrations and it is therefore plausible that a cytoplasmic pool of ProSAP2/Shank3 exists that is not Zn2+ bound.
A regression correlation analysis revealed that Zn2+-supplementation or depletion had significantly stronger effects on the synaptic localization of ProSAP2/Shank3 than ProSAP1/Shank2 (Supplementary Fig. 2A). Given that dissociated hippocampal cell cultures are low in presynaptic vesicular Zn2+ and that application of Zn2+-chelators during 1 min HiK+ stimulation and subsequent re-supplementation with conditioned medium still results in an increase of synaptic ProSAP1/Shank2 and ProSAP2/Shank3 levels (Supplementary Fig. 2B), we also investigated whether releasable postsynaptic Zn2+ is a potential source for binding to ProSAP/Shank. Important intracellular Zn2+ buffers are MTs, in particular MT-3, which is prominently expressed in neurons (Lee et al., 2003). It has been reported that Ca2+ transients can rapidly and persistently release Zn2+ from MT through endogenously generated NO (Wang et al., 2008). As such a mechanism can conceivably also occur in synapses we therefore overexpressed MT-3 and applied a NO antagonist (N-nitro-l-arginine) as well as an agonist (spermine nonoate) (Fig. 1G and H). Stimulation of neurons in the presence of the NO antagonist inhibited an increase in synaptic ProSAP1/Shank2 and ProSAP2/Shank3 levels, whereas application of the NO agonist slightly increased ProSAP2/Shank3 platform formation (Fig. 1G). Moreover, analysis of MT-3 overexpressing neurons compared with control cells showed an increase in ProSAP2/Shank3 levels in transfected cells (Fig. 1H), indicating that releasable postsynaptic Zn2+ can also contribute to the assembly of the postsynaptic ProSAP/Shank scaffold. Stimulation with HiK+ further increased synaptic ProSAP1/Shank2 and ProSAP2/Shank3 levels in MT-3 overexpressing cells, whereas knockdown of MT-3 (Supplementary Fig. 2C) blocked activity-dependent increases (Fig. 1H).
Finally, the synaptic localization of Shank1 (which does not bind Zn2+ as well as Homer1b/c) PSD95, mGluR5 and gephyrin (a scaffold protein of inhibitory synapses that does not contain ProSAP/Shank proteins) were analysed. The results show that there is no general increase in postsynaptic density protein levels after Zn2+-supplementation. Shank1 concentrations at the postsynaptic density are not affected by either Zn2+ supplementation or HiK+ stimulation (Fig. 1D), whereas Homer1b/c seems to be negatively regulated by activity and Zn2+ supplementation, but unaffected from Zn2+ depletion. In contrast, synaptic immunofluorescence for PSD95 and mGluR5 receptors, was also elevated following Zn2+ supplementation or HiK+ stimulation (Fig. 1D), although only the increase in mGluR5, but not PSD95 was blocked by stimulation in the presence of CaEDTA.
Zinc deficiency leads to synaptic deficits associated with the loss of synaptic ProSAP2/Shank3 in vivo
Autism spectrum disorders are neurodevelopmental disorders and we therefore asked whether acute or chronic Zn2+-supplementation or depletion influence synapse maturation during development. Cultivation of hippocampal neurons under Zn2+-deficient conditions leads to a decrease in synapse density, whereas the number of inhibitory synapses was not altered (Supplementary Fig. 3). In contrast, we observed a significant upregulation of ProSAP1/Shank2 signal density in cells supplemented with ZnCl2 (Supplementary Fig. 3). Thus, chronic Zn2+-deficiency might contribute to defects that are observed in ProSAP/Shank associated brain disorders such as autism spectrum disorder. We therefore analysed the impact of Zn2+-deficiency on ProSAP/Shank scaffold formation in vivo by inducing Zn2+-deficiency in mice.
To this end, mice were fed a Zn2+-deficient diet for 8 weeks. Timm’s staining to visualize free Zn2+ ions revealed a significant reduction of brain Zn2+ levels in cortical, striatal and hippocampal regions (Supplementary Fig. 4A and B). The significant reduction in Zn2+ levels was further confirmed in brain sections using Zinpyr1, a Zn2+-staining fluorophore (Supplementary Fig. 4C and D). Given that levels of vesicular Zn2+ are generally low in cerebellum (Frederickson and Moncrieff, 1994), we failed to detect a significant reduction (Supplementary Fig. 4A–D). The general brain morphology and cell density assessed by 4',6-diamidino-2-phenylindole (DAPI) and Nissl staining was not altered in Zn2+-deficient animals (Supplementary Fig. 4E–H). Labelling of the pre- and postsynaptic marker proteins bassoon, homer1 and gephyrin in brain sections of Zn2+-deficient animals shows that compared with control animals, a general decrease in the number of homer1- and bassoon-positive signals (excitatory synapses), visible as a trend (P = 0.09) in the hippocampus but statistically significant in the cortex and striatum (Fig. 2A), can be detected. The number of gephyrin- and bassoon-positive signals (inhibitory synapses) was unchanged. Next, we performed immunohistochemistry to label ProSAP/Shank proteins. The immunofluorescence intensity correlating to protein levels of ProSAP2/Shank3 was significantly reduced in all four brain regions: striatum, hippocampus, cortex and cerebellum (Fig. 2B and C). Moreover, we found a significant reduction of ProSAP1/Shank2 immunoreactivity in cortex (Fig. 2C). The density and intensity of gephyrin immunofluorescence puncta used as a marker of inhibitory synapses was not affected in mice fed with a Zn2+ deficient diet (Fig. 2A and C).
Analysis of P2 fractions from striatum, hippocampus, cortex and cerebellum also revealed a significant reduction of protein levels of ProSAP1/Shank2 in cortex and striatum, ProSAP2/Shank3 in all four brain regions, and of Shank1 in hippocampus (Fig. 2D) in Zn2+-deficient mice. Along with this reduction we observed a shift from the insoluble P2 to the soluble S2 fraction for several synaptic proteins including ProSAP/Shanks (Supplementary Fig. 5A and B). Importantly, the loss of ProSAP1/Shank2 in the synapse enriched P2 fraction translates into a decrease of excitatory receptors in this fraction, including GluA1 (striatum and hippocampus), GluA2 (cerebellum and striatum), GluA3 (striatum), GluN1 (striatum and hippocampus) and GluN2B (cortex) (Fig. 2D). Interestingly, mGluR5 levels were significantly decreased in hippocampus but significantly increased in striatum (Fig. 2D).
We also prepared synaptic junction fractions of whole brain (without cerebellum) of Zn2+-deficient animals and compared protein levels to those of control mice (Fig. 2E). We detected a significant loss of ProSAP2/Shank3 proteins and GluA1 receptors in Zn2+-deficient animals. However, we could not see the decrease in ProSAP1/Shank2 and Shank1 (Fig. 2E). Quantitative real-time PCR confirmed that the observed changes occurred on protein rather than on messenger RNA level, since the transcriptional levels of ProSAP1/Shank2 and ProSAP2/Shank3 were not altered between control and Zn2+-deficient mice (Fig. 2F).
Analysing the pups of control and Zn2+-deficient animals on post-natal Day 3, we also detected a significant reduction in brain Zn2+-levels in pups from Zn2+-deficient mothers (Fig. 3A and B). However, all pups from and nursed by Zn2+-deficient mice died on post-natal Day 20, whereas pups from Zn2+-deficient mice but nursed by control mice survived. The brain Zn2+ levels of Zn2+-deficient pups nursed by control mice were no longer significantly different from those of control pups (Fig. 3C). Zn2+-deficient pups did not show significantly fewer synapses (Fig. 3D), but displayed a trend towards a decrease, and the transcriptional levels of ProSAP1/Shank2 and ProSAP2/Shank3 were not altered in Zn2+-deficient pups (Fig. 3E). However, protein biochemistry using P2 fractions from post-natal Day 3 pups shows that a significant loss of all ProSAP/Shank family members occurred in pups from Zn2+-deficient mothers (Fig. 3F). Along with this, we again detected a significant decrease in GluA1, GluN1 and GluN2B levels. In contrast, pups from Zn2+-deficient mothers that have been nursed by control animals (fed a normal diet) after birth show a significant rescue of ProSAP/Shank and GluA1, GluN1 and GluN2B protein levels. However, Shank1, GluA1 and GluN1 still were significantly reduced compared with control animals (Fig. 3F). Although the amount of mGluR5 was significantly decreased in Zn2+-deficient pups, a significant increase was detected in pups from Zn2+-deficient mothers that have been nursed by control animals (Fig. 3F). Using immunohistochemistry, similar results were obtained for ProSAP/Shank proteins (Fig. 3G). Although all synaptic concentrations of ProSAP/Shank family members are decreased in Zn2+-deficient pups (nursed by Zn2+-deficient mothers), the signal intensity of labelled ProSAP/Shank puncta in different brain regions of Zn2+-deficient pups nursed by control mothers was not significantly different for ProSAP1/Shank2 and ProSAP2/Shank3. In contrast, Shank1 levels were significantly decreased averaging all brain regions. Total gephyrin levels were again not altered. The number of signals per optic field was not significantly different for all ProSAP/Shank family members (Fig. 3H), whereas a slight decrease of gephyrin positive puncta in the cerebellum was measured (Fig. 3H).
Mouse models for acute and prenatal zinc deficiency are hyper-responsive, hyperactive and show impairments in social behaviour and ultrasonic vocalization
To investigate the influence of Zn2+-deficiency on behaviour, we generated acute and prenatal Zn2+-deficient mice. In a first set of experiments, we analysed the influence of acute Zn2+-deficiency in a test for maternal behaviour (response to wriggling calls) (Fig. 4A–E and Supplementary Fig. 6). Zn2+-deficient mice demonstrated increased activity and reactivity in the situation of maternal care compared to normal mothers (Fig. 4A–E and Supplementary Fig. 6D–I).
In a second set of experiments, we performed the same test for maternal behaviour, this time using mice that were exposed to a prenatal Zn2+-deficiency (Supplementary Fig. 7A–D). In contrast to acute Zn2+-deficient mothers, mothers that had experienced a prenatal Zn2+-deficiency during development showed no hyper-responsivity (Supplementary Fig. 7E–H). However, prenatal Zn2+-deficient mothers showed less maternal behaviour in response to calls (Fig. 4F).
This decrease of maternal behaviour prompted us to analyse prenatal Zn2+-deficient mice, performing behavioural tests (pup retrieval, analysis of ultrasonic vocalization and maternal resident intruder test) indicative for autism spectrum disorder-related behaviour. The results show that although we could not detect significant changes in the pup retrieval test between prenatal Zn2+-deficient and control mice regarding the latency to respond to the stimulus, the number of control walks performed and the number of pup retrievals (Supplementary Fig. 7I–K), prenatal Zn2+-deficient mice showed an impairment in auditory discrimination between a meaningful (50 kHz) and neutral (20 kHz) sound stimulus (Fig. 4G).
The most prominent impairments were observed evaluating parameters of ultrasonic vocalizations of adult males to female urine (Fig. 4H–L). Besides several parameters that were not altered in prenatal Zn2+-deficient mice compared with control mice (Supplementary Fig. 8A–H), we found that prenatal Zn2+-deficient mice vocalized significantly less calls per minute observation time (Fig. 4H), had a significantly reduced loudness of calls (Fig. 4I), a significantly increased latency to start calling (Fig. 4J) and a trend to a reduced average call duration compared to control animals (Fig. 4K). Moreover, the number of overtones and harmonics was significantly reduced in the calls (Fig. 4L).
Finally, we conducted a maternal resident intruder test. Besides parameters that were not altered (Fig. 4M and Supplementary Fig. 8I–L), the number of attacks performed by the mother was significantly increased in prenatal Zn2+-deficient mice (Fig. 4N).
Zinc deficiency in Phelan-McDermid syndrome is associated with a higher incidence rate of seizures, hypotonia and the occurrence of attention deficits and hyperactivity
Given that patients with Phelan-McDermid syndrome suffer from a haplo-insufficiency of ProSAP2/Shank3 we investigated whether Zn2+-deficiency in patients with Phelan-McDermid syndrome leads to a more pronounced phenotype. Using atomic absorption spectroscopy, we evaluated Cu2+ and Zn2+-levels in blood samples of 38 participants diagnosed with Phelan-McDermid syndrome. Similar to former studies with autistic patients (Faber et al., 2009; Lakshmi Priya and Geetha, 2011; Yasuda et al., 2011), we detected an increased incidence rate for Zn2+-deficiency in patients with Phelan-McDermid syndrome primarily in the age group <12 years (Supplementary Fig. 9A) as previously reported for autistic children (Yasuda et al., 2011), with the exception of two cases at 19 and 46 years of age. From 37 participants, 30 were found to have normal blood Zn2+-levels, four showed a mild Zn2+-deficiency (1–20% less than the lowest concentration of the age matched reference values for healthy controls) and three a severe Zn2+-deficiency (>20% less than the lowest concentration of the age matched reference values for healthy controls) (Fig. 5A and B). Zn2+-deficiency was not correlated with gender (Supplementary Fig. 9B) and the participants had no dietary problems (no underweight, no physical disability hampering food consumption, normal dietary habits), that might explain the deficiency.
Participants were evaluated based on 26 categories (see ‘Materials and methods’ section.). We found that 61% of participants suffering from seizures showed a Zn2+-deficiency (mild and severe combined) compared with only 12% of participants without seizures (Fig. 5C). Additionally, 60% of participants with attention deficit disorder/attention deficit hyperactivity disorder or other attention or hyperactivity issues were severely Zn2+-deficient and 27% of participants with hypotonia compared with 15% and 0%, respectively, of participants without the symptom (Fig. 5C). Additionally, the occurrence of Zn2+-deficiency was positively correlated with signs of immunodeficiency (Fig. 5C).
Recent first in vitro evidence indicated Zn2+ incorporation in large macromolecular platforms that are built by the SAM domains of the autism-associated postsynaptic density scaffolding molecules ProSAP2/Shank3 and ProSAP1/Shank2. Here, we propose a model where the synaptic assembly of ProSAP1/Shank2 and ProSAP2/Shank3 proteins is influenced by Zn2+-deficiency.
Zn2+-depletion from neuronal cultures or Zn2+-sequestration by amyloid-β decreased the synaptic levels of Zn2+-binding ProSAP/Shank family members as shown here and in former studies (Grabrucker et al., 2011a, b) and results in a reduced synapse density. Analysing Zn2+-deficient mice, here, we also detected a reduction in synapse number in the cortex and striatum. Along with this, we could show a significant decrease of ProSAP1/Shank2 (cortex) and ProSAP2/Shank3 (cortex, hippocampus, striatum, cerebellum) through immunofluorescence and a significant decrease of ProSAP1/Shank2 (cortex, striatum), ProSAP2/Shank3 (cortex, hippocampus, striatum, cerebellum) and Shank1 (hippocampus) through protein biochemistry in Zn2+-deficient mice. This loss of postsynaptic density scaffold proteins was accompanied by a downregulation of receptors, especially GluA1 and GluN1 (hippocampus and striatum) reminiscent of a similar downregulation of AMPA receptors reported in heterozygous and homozygous ProSAP2/Shank3αβ knockout mice (Bozdagi et al., 2010; Wang et al., 2011). ProSAP2/Shank3αβ knockout and ProSAP2/Shank3ΔC mice were also reported to have decreased GluN1 and GKAP levels (Bangash et al., 2011; Wang et al., 2011). Furthermore, ProSAP1/Shank2 knockout mice were reported to display a decrease in NMDA receptors (Won et al., 2012).
Previous studies have shown that especially nascent synapses are sensitive to Zn2+-depletion (Grabrucker et al., 2011a) and a model was proposed where ProSAP1/Shank2 is the first ProSAP/Shank family member found at forming synapses followed by ProSAP2/Shank3. Only if a sufficient stabilized synapse with preformed scaffold is established, Shank1 is targeted to postsynaptic densities, which leads to less Zn2+-sensitive mature synapses. The results of the present study show that although Zn2+-deficiency in adult mice only significantly changes Shank1 levels in hippocampus, a much more pronounced reduction of Shank1 is seen in pups from Zn2+-deficient mothers. One might speculate that in adult mice, synapses have already been stabilized by Shank1 and that loss of Shank1 because of impaired synapse maturation needs longer periods of Zn2+-deficiency in less plastic brain regions than the hippocampus. In contrast, in Zn2+-deficient pups upon initial synapse formation, postsynaptic densities might not be targeted by Shank1 from the beginning because of their weakened ProSAP1/Shank2 and ProSAP2/Shank3 scaffold.
Intriguingly, we observed most alterations in hippocampus and especially striatum, being the brain regions with the highest expression levels of ProSAP2/Shank3 compared with other ProSAP/Shank family members (Peça et al., 2011). Similar to a ProSAP2/Shank3 knockout mouse model (Peça et al., 2011), we detected a reduction in spine density in the striatum. Given that the striatum contains a low amount of presynaptic Zn2+ but has a high expression of ProSAP2/Shank3 scaffolds that can only be built in presence of Zn2+, indeed Zn2+ may be bound within the postsynaptic density in all brain regions regulated by postsynaptic Zn2+-stores. Thus, if postsynaptic stores are affected and ProSAP2/Shank3 has a predominant role in the striatum, it is likely that synapse density in this brain region is most affected by Zn2+-deficiency, independent from the presynaptic pool of Zn2+ ions. This might also explain the reduction of ProSAP2/Shank3 levels in cerebellum.
As it was also shown before that there is a rapid turnover of ProSAP/Shank proteins at the synapse (Tsuriel et al., 2006) and that ProSAP1/Shank2 and ProSAP2/Shank3 can be detected in the soluble (S2) as well as postsynaptic density-bound (P2) fraction of brain lysates, it is plausible that both proteins undergo a shift from P2 to S2 upon Zn2+-depletion (Grabrucker et al., 2011a). One can therefore assume that the soluble and postsynaptic density-bound pool exist in equilibrium. Synaptic activity and Zn2+-release may rapidly shift this equilibrium to a more postsynaptic density-bound pool of ProSAP/Shank (Grabrucker, 2013).
The present results raise a number of questions concerning the source of postsynaptic free Zn2+, the dynamics of the ProSAP/Shank scaffold and the pathomechanism of autism spectrum disorder and related disorders. Knockout mice for ZnT3, the major and possibly sole synaptic vesicular Zn2+-transporter, were not reported yet to display autism-like features, although they show an absence of free Zn2+ in synaptic vesicles (Cole et al., 1999). Therefore the question arises whether presynaptically released Zn2+ is the sole source of Zn2+ ions involved in the synaptic targeting of ProSAP/Shank. Alternatively, local elevation of Zn2+-concentrations could arise by activation of intracellular Zn2+-stores through NO. In the brain, NO has been shown to displace Zn2+ from protein-binding sites (Cuajungco and Lee, 1998; Aravindakumar et al., 1999), causing free or weakly bound Zn2+ to be present in the cytoplasm that could play a role in intracellular signalling (Bossy-Wetzel et al., 2004; Lin et al., 2007). Our in vitro experiments indicate that MT-3 might be a candidate for postsynaptic activity-dependent Zn2+-release due to activity-dependent nitrosylation or oxidation of the thiol ligands by NO (Maret, 2000). In line with this, NO antagonists prevented the stimulation-induced upregulation of Zn2+-binding ProSAP/Shank proteins at the synapse. Overexpression of MT-3, because of spontaneous synaptic activity or more so after HiK+ stimulation, increased synaptic ProSAP2/Shank3 levels, whereas MT-3 knockdown prevented an activity-dependent increase. The slight increase in ProSAP2/Shank3 by MT-3 knockdown in absence of stimulation might be because of a higher baseline availability of free Zn2+ that would have been bound to MT-3 otherwise. Intriguingly, mice lacking MT-3 show a reduction in brain Zn2+-concentrations as a result of the absence of Zn2+ bound to MT-3. However, the histochemically-reactive Zn2+-content is unaffected, providing evidence that MT-3 does not influence presynaptic vesicular Zn2+ (Erickson et al., 1995, 1997). MT-3 knockout mice show abnormalities in behaviour (Koumura et al., 2009) and have been suggested as suitable subjects to investigate psychological disorders (Koumura et al., 2009).
Interestingly, previous studies have reported a high rate of Zn2+-deficiency in autistic children (Walsh et al., 1997; Jen and Yan, 2010; Yasuda et al., 2011), suggesting that infantile Zn2+-deficiency might contribute to the pathogenesis of autism (Yasuda et al., 2011). Along with a reduction in synaptic ProSAP1/Shank2 and ProSAP2/Shank3 levels, Zn2+-deficient animals developed behavioural abnormalities.
Acute Zn2+-deficiency led to increased hyperactivity-like spontaneous behaviour and over-responsiveness to acoustic stimuli. This suggests lability of behavioural coordination in the highly affected striatum (see above) as changes both in sound perception, which could be related to cortical processing, and in goal-directed instinctive behaviour, which might reflect cerebellar action, were not observed. Because autism spectrum disorders are developmental disorders, induction of Zn2+-deficiency in adulthood is not expected to generate autism in mice. Nevertheless, the results show that Zn2+-deficiency might be responsible for co-morbidities seen in autism spectrum disorders such as increased anxiety and sensory over-responsiveness, which is underlined by the increased rate of hyperactivity issues in patients with Phelan-McDermid syndrome and Zn2+-deficiency. For instance, in addition to the core diagnostic features for autism spectrum disorder, atypical stimuli-evoked responses have been reported in up to 95% of children with autism spectrum disorder (Lane et al., 2012).
Intriguingly, our data from pups from Zn2+-deficient mice indicate that maternal Zn2+-deficiency is also able to induce changes in ProSAP/Shank levels in the offspring, similar to those reported in adult Zn2+-deficient mice here. Although Zn2+-deficiency led to the death of pups on post-natal Day 20 hinting towards a developmental milestone that Zn2+-deficient pups were unable to achieve, provision of adequate Zn2+ supply by control mothers interestingly led to survival and a rescue of the observed alterations in most, but not all postsynaptic density proteins. At the age of 17–22 days, pups usually start eating solid food and during this weaning period, the mothers begin to withdraw from their litter to reduce suckling attempts by pups (Koenig and Markl, 1987). Zn2+-deficiency is known to have an influence on motor development (Black, 1998) and a possible reason for the death of pups on post-natal Day 20 might be an inability to reach sufficient amounts of food and water at the top of the cage.
Prenatal Zn2+-deficient mice, in contrast to acute Zn2+-deficient mice, did not show signs of increased hyperactivity-like and over-responsive behaviour. We observed significantly decreased maternal behaviour along with impairment in auditory discrimination of sound stimuli, whereas sound perception per se was not altered. Most prominent, we could detect significant differences in the vocalization of prenatal Zn2+-deficient males. Intriguingly, a reduction in ultrasonic vocalizations has been reported previously for heterozygous ProSAP2/Shank3αβ (Bozdagi et al., 2010) and homozygous ProSAP1/Shank2 knockout mice (Schmeisser et al., 2012; Won et al., 2012). Additionally, prenatal Zn2+-deficient mice showed alterations in the social context of a maternal resident intruder test. We observed a significant increase in attacks towards the intruder. Increased aggressiveness has been reported in prenatal Zn2+-deficient animals (Sandstead et al., 1977). Taken together, prenatal Zn2+-deficient animals indeed display some behavioural features observed in mouse models for autism.
Patients with Phelan-McDermid syndrome already lack one functional copy of ProSAP2/Shank3, so that they should be more susceptible to changes in body Zn2+-levels. Although, given the low incidence rate of Phelan-McDermid syndrome with 2.5–10 per million births, the number of participants in our study was rather low and serum Zn2+-levels are only indicative of brain Zn2+-levels, the results complement those from our Zn2+-deficient mouse model with regard to increased rate of attention and hyperactivity issues compared with patients without these symptoms. Similarly, we found a higher proportion of Zn2+-deficient participants in the group of patients suffering from hypotonia and/or seizures and signs of primary immunodeficiency. Immune system abnormalities have been linked to Zn2+-deficiency before and both have been reported as risk factors for the development of autism (Grabrucker, 2012). Although an alternative explanation would be that patients with more severe clinical presentations are more susceptible to Zn2+-deficiency, it is likely that Zn2+-deficiency is influencing the severity of these symptoms.
In conclusion, our data predict that autistic children on Zn2+-therapy should improve regarding attention, hyperactivity, seizures, sound and tactile sensitivity, among others. Recent data confirm this therapeutic outlook (Russo and Devito, 2011).
The authors gratefully acknowledge the professional technical assistance of Elisabeth Pica and thank Geraldine Bliss, Megan O’Boyle and the Phelan McDermid Foundation for their help as well as the Phelan-McDermid syndrome families for participating in blood donations.
This work was supported by Baustein 3.2 (L.SBN.0083) and the DAAD (to A.M.G). T.B. was funded by the ANR (ANR-08-MNPS-037-01 – SynGen), Neuron-ERANET (EUHF-AUTISM), Fondation Orange and the Fondation FondaMentale. M.R.K by the DFG (SFB 779/TPB8; Kr1879/3-1). S.G. is a member of the International Graduate School in Molecular Medicine at Ulm University.
Supplementary material is available at Brain online.