All animal experiments were performed in compliance with the guidelines for the welfare of experimental animals issued by the Federal Government of Germany, the National Institute of Health or the Max Planck Society. The experiments in this study were approved by the review board of the Land Baden-Württemberg, permit numbers 0.103, 321/16 and 966/2016-PR, respectively. Shank2(−/−) mice (mus musculus) were generated as previously described (Schmeisser et al., 2012). Breeding was carried out as heterozygous breeding on a C57BL/6 background; animals were housed in standard laboratoryconditions (average temperature of 22°C with food and water available ad libitum, dark/light cycle as 12/12 rhythm).

Patient and control hiPSCs were reprogrammed from hair keratinocytes, following a published protocol (Aasen et al., 2008; Stockmann et al., 2013). Hair was obtained from Thomas Bourgeron, Paris, France. hiPSCs were grown on hESC-qualified matrigel (BD Biosciences) in mTeSR1 medium (Stemcell Technologies) at 37°C, 5% O₂, and 5% CO₂.

Human induced pluripotent stem cells were spontaneously differentiated in all three germlines by formation of embryoid bodies (EBs) in hESC medium (DMEM/F12 + GlutaMAX, 20% knockout serum replacement, 1% NEAA, 1% antibiotic-antimycotic [all Gibco), 1% β-mercaptoethanol (Millipore)] in low-attachment flasks. ROCK inhibitor was added the first 24 h. Cells were kept 7 days as free-floating EBs, plated on 35 mm μ-dishes (Ibidi) coated with poly-L-ornithine and laminin and grown further for 21 days. Cells were fixed and stained for ectoderm, mesoderm, and endoderm using the following primary antibodies: β-III-TUBULIN (chicken, 1:1,000, Millipore), α-Actinin (mouse, 1:500, Sigma-Aldrich) and alpha-fetoprotein (AFP, goat, 1:100, Santa-Cruz). hiPS cell lines were karyotyped to exclude cell lines-harboring chromosomal aberrations after reprogramming (Linta et al., 2012). hiPSCs were treated with 1.5 M Colchicine (Eurobio) diluted in mTeSR1 for 2 h at 37°C. Cells were trypsinized (TrypLE, Gibco), and metaphases were kindly analyzed by the Institute for Human Genetics, Ulm University.

Neural differentiation followed previously published protocols (Hu and Zhang, 2009; Stockmann et al., 2013). hiPSCs were lifted using dispase (Stemcell Technologies) and grown as EBs in low-attachment flasks (Corning Costar) using hESC medium. ROCK inhibitor (Ascent Scientific) was added the first 24 h. Medium was changed to neuronal basal-medium [DMEM/F12, 0.02 mg/ml insulin (SAFC), 24 nM sodium selenite, 16 nM progesterone, 0.08 mg/ml apotransferrin, 7.72 μg/ml putrescin, 50 mg/ml heparin (all Sigma-Aldrich), 1% NEAA, 1% antibiotic-antimycotic, 10 μg/ml BDNF, GDNF, and IGF-1 (all Gibco), 0.1 μMcAMP (Sigma-Aldrich), 50 mg/ml ascorbic acid (PeproTech)] at Day 3 of culture. At Day 7, EBs were plated on laminin (Sigma)-coated dishes. 0.1 μM retinoic acid was added from Day10 and B27 (Gibco) and 1 μM purmorphamine (Calbiochem) from Day 14, when neural rosettes were detached and grown in suspension again. At Day 28, EBs were plated on 35 mm μ-dish (Ibidi) coated with poly-L-ornithine (Sigma) and laminin. The medium was changed to 0.5 μM purmorphamine and 0.05 μM retinoic acid and changed one time per week. Neuronal cells were harvested or fixated at Days 21 and 42 after final plating.

Total RNA was isolated using the RNeasy Mini Kit (Qiagen), following the instructions of the manufacturers. Cell pellets were lysed in an RLT buffer. Quantitative real-time PCR was carried out using the one-step QuantiFast SYBR Green RT-PCR kit (Qiagen) and a Rotor-Gene-Q real-time PCR machine (Model 2-Plex HRM, Qiagen). All primers were obtained from Qiagen (QuantiTect Primer Assay). Amplification was carried out under the following conditions: 10 min at 55°C and 5 min at 95°C, followed by 40 cycles of 5 s at 95°C and 10 s at 60°C (one-step qRT-PCR). Rotor-Gene-Q software (version 2.0.2) calculated the cycle threshold (Ct) values. All measurements were run in technical duplicates and normalized to housekeeping genes hydroxymethylbilane synthase (HMBS) or neurofilament H (NFH).

Cell pellets were lysed in a RIPA buffer [150 mM NaCl (Sigma); 50 mM Tris/HCl (AppliChem); 1% NP-40; 0.5% Sodium deoxycholate (Merck); 0.1% SDS; Protease inhibitor (Roche); and Phosphatase inhibitor (Roche)] for 45 min on ice, sonicated (10 pulses), and cleared for 10 min at 13.000 rpm. Mouse brain tissue was homogenized in a lysis buffer (10 mM HEPES pH 7.4, 2 mM EDTA, 5 mM Sodium Orthovanadate, 30 mM Sodium Fluoride, 20 mM β-glycerol phosphate, complete (Roche). Protein concentration was quantified using a colorimetric measurement of protein in Bradford solution [0.1% Serva Blue Powder (Serva), 8.5% Phosphoric acid (VWR), 4.75% ethanol in H₂O]. Equal protein amount was boiled in an SDS-loading buffer [200 mMTris/HCl (pH6.8); 200 mM DTT (Sigma); 4% SDS (Roth); 4 mM EDTA (AppliChem); 40% Glycerol (AppliChem); 0.2% Bromphenol blue (Sigma)] and separated using SDS-PAGE. Blotting on nitrocellulose membranes was performed using Trans-Blot^® Turbo^TM RTA Midi Nitrocellulose Transfer Kit (BioRAD, 1704271). Membranes were blocked in 5% BSA (AppliChem) or 5% milk powder (Fluka) in 0.1% TBS-Tween 20 and incubated with primary antibodies. HRP-conjugated secondary antibodies were visualized with ECL Western Blot substrate (Pierce) and a MicroChemi 4.2 machine. Signals were quantified using Gelanalyzer Software and normalized against the corresponding loading control. The following primary antibodies were used: SHANK2 “ppI-SAMpabSA5192” (rabbit, 1:000 (Boeckers et al., 1999; Schmeisser et al., 2012), SHANK3/ProSAP2 (Fragment 1 + 2, Tier 2, rabbit, 1:500 (Schmeisser et al., 2012), mGluR5 (rabbit, 1:500, Abcam), vGLUT1 (guinea pig, 1:500, SYSY), GluA2 (mouse, 1:500, SYSY), GAD65 (mouse, 1:500, abcam ab26113), GABA_BR1 (rabbit, 1:200, SYSY), ERK1/2 (rb, 1:1,000, Cell Signaling, 4695), phospho-ERK1/2 (Thr202/Tyr204, rb, 1:2,000, Cell Signaling 4270), Akt (rb, 1:1,000, Cell Signaling 4691), phospho-AKT (Thr308, rb, 1:1,000, Cell Signaling 13038), NFH (mouse, 1:1,000, Convance), β-actin (mouse, 1:100,000, Sigma).

Caspase 3 Assay Kit (Colorimetric) was purchased from Abcam ab39401. Measurements were carried out following the instructions of the manufacturers. Cells were lysed in 50 μL of chilled Cell Lysis Buffer. Protein concentration was assessed by a Bradford assay as previously described. About 50 μg per 50 μL protein was measured after 1 h incubation at 37°C for 60 min. Output was measured at 405 nm on a Cytation^TM 3 Cell Imaging Multi-Mode Reader (BioTek).

Cells were fixed using 4% paraformaldehyde (Merck)/4% sucrose (Roth) in PBS at 37°C for 15 min and washed three times with PBS + Ca²⁺/Mg²⁺ (PAA). Fixed cells were permeabilized with 0.2% Triton-X100 (Roche) in PBS for 10 min and blocked in 5% fetal bovine serum (Gibco)/10% goat serum (Millipore) in PBS for at least 2 h. Primary antibodies were incubated at 4°C for 48 h in blocking solution. After three-time wash with PBS, secondary antibodies coupled to Alexa Fluor 488, 568, or 647 (Life Technologies) were incubated for 1 h at RT in blocking solution. Cells were mounted with ProLong^® Gold antifade reagent with DAPI (Thermo Fisher Scientific). The following primary antibodies were used: SHANK2 “ppI-SAMpabSA5192” [rabbit, 1:500, (Boeckers et al., 1999; Schmeisser et al., 2012)], SHANK3/ProSAP2 [Fragment 1 + 2, Tier 2, rabbit, 1:500 (Schmeisser et al., 2012)], Ki67 (rabbit, 1:500, Abcam), Active Caspase3 (rabbit, 1:500, R&D Systems), GluA2 (mouse, 1:500, SYSY), Bassoon (mouse, 1:500, SYSY), mGluR5 (rabbit, 1:500, Abcam), vGLUT1 (guinea pig, 1:500, SYSY), vGAT (rabbit, 1:500, SYSY), GABA_ARα1 (mouse, 1:200, NeuroMab), Homer1 (guinea pig, 1:500, SYSY), GluN1 (rabbit, 1:500 Sigma), ERK1/2 (rb, 1:200, Cell Signaling 4695), phospho-ERK1/2 (Thr202/Tyr204, rb, 1:1,000, Cell Signaling 4270), phospho-CREB (S133, rb:1:500, Abcam 32096), MAP2 (ck, 1:1,000, EnCor), NFH (chicken, 1:50,000, Antikörper online), and StemLite Pluripotency Kit (Cell Signaling, #9656). Fluorescent images were recorded using an Axioscope microscope with a Zeiss CCD camera (16 bits; 1,280 × 1,024 ppi) and Axiovision software (Zeiss).

The protein array [Full Moon Biosystems, ERK phospho Antibody Array (PEK208)] was performed according to the instructions of the manufacturer using three replicates per cell line and time point. Briefly, neurons were lysed in a 100 μl extraction buffer (Biosystems) supplemented with protease and phosphatase inhibitors (Roche). Protein concentration was determined performing and A280 assay. Blocking and coupling were performed according to the protocol of the manufacturer. Biotinylated samples were labeled with Cy5 Streptavidin (Invitrogen SA1011). The detection of antibody arrays was performed in a fluorescent slide scanner (Genepix 4000B microarray scanner, Molecular Devices). The 16-bit images were analyzed using the GenePix Pro 6.1 software. The data had to fulfill the requirements of a signal-noise ratio (SNR) above 3 in the 635 channels, and all flags needed to be zero. For quantification, the mean of six technical replicates per antibody was used and set relative to β-ACTIN.

R version 4.1.0 was used for all the following steps of the analysis. Data were grouped consecutively by batch, timepoint, and protein. Subgroups were normalized, respectively, on the mean of maternal and paternal measurements to return fold change values (FC). Pheatmap R package was used to create heatmaps and perform hierarchical clustering using Euclidean distance with a “complete” method. The heatmaps show Z-scored FC per protein. For labeled heatmaps, data were filtered for proteins with FC of either 0.45 greater or smaller for both a mother and a father in comparison to a patient, revealing changes in the patient between 0.65 and 1.45 relative to the mean of the parents. The FC was calculated using the following formula: [father < (patient - fc.threshold) and mother < (patient - fc.threshold)] or [father > (patient + fc.threshold) and mother > (patient + fc.threshold)].

Fluorescent images were analyzed using Axiovision and ImageJ software. For synaptic analysis, pictures were deconvoluted using Autoquant X3 Deconvolution software (Imaris). Intensity of foci along neurites was measured using the “Find Foci” plugin of ImageJ (Herbert et al., 2014) as previously published (Higelin et al., 2016). Neuronal soma size was measured manually in the NFH immunostaining channel using ImageJ. For analysis of growth cones size, a polygonal region of interest containing only one growth cone was selected using ImageJ. A threshold was set for the GAP-43 staining, and the respective size of the positive area was obtained. For analysis of Ki67 positive cells, the number of NFH and DAPI-positive neurons per picture was counted and set to 100%. Then, Ki67 positive nuclei in NFH-positive neurons were counted, and the percentage was calculated. Per experiment and cell line, nine pictures were analyzed. For analysis of ERK1/2 and phospho-ERK1/2 intensity analysis, the soma of MAP2 positive neurons was surrounded manually, and the mean gray value of ERK1/2 or phospho-ERK1/2 was measured in the selected area. For analysis of phospho-CREB intensity, the DAPI signal of MAP2 positive neurons was surrounded manually, and the intensity of phospho-CREB signal was measured within the selection.

Statistical analysis was performed using GraphPad Prism 6. Significance value was set to 0.05 with ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, and ^****p ≤ 0.0001. Values were tested using Student‘s t-test or one-way ANOVA, followed by post hoc analysis (Tukey’s multiple comparison test) as indicated.

The SHANK2 deletion patient analyzed in this study carries a heterozygous deletion of 1.8 Mb on chromosome 11, comprising SHANK2 (Figure 1A). Keratinocytes grown from hair roots of the patient, the father, and the mother were reprogrammed to hiPSCs, following previously published protocols (Supplementary Figure 1A and Linta et al., 2012). Generated hiPSCs expressed pluripotency markers SOX2, OCT4, NANOG, TRA-1-81, TRA-1-60, and SSEA4 in immunostaining (Supplementary Figure 1B) and SOX2, OCT4, NANOG, and KLF4 on the RNA level (Supplementary Figure 1C). They differentiated spontaneously into all three germlines as shown by immunostaining for α-ACTININ (mesoderm), alpha-fetoprotein (AFP, endoderm), and β-III-TUBULIN (ectoderm) (Supplementary Figure 1D). No genetic alterations were detectable by karyogram analysis in all three cell lines (Supplementary Figure 1E). We differentiated hiPSCs into neurons following previously established protocols (Hu and Zhang, 2009; Stockmann et al., 2013) and analyzed young neurons after 21 days (d21) and mature neurons after 42 days (d42) of final plating. Patient and parent neuronal cultures contained comparable numbers of neurofilament heavy chain (NFH)-positive neurons and equal NFH levels relative to housekeeping gene HMBS in d21 (Figure 1B) and d42 cultures (Figure 1C). About 40 to 50% of cells in mature d42 cultures were found to be NFH-expressing neurons, as previously reported for this protocol (Stockmann et al., 2013; Higelin et al., 2018). SHANK2 RNA levels were significantly decreased to approximately 50% in hiPSCs (Figure 1D), neurons d21 (Figure 1E), and neurons d42 (Figure 1F). Analysis of SHANK2 protein expression revealed similar reductions in hiPSCs (Figure 1G), neurons d21 (Figure 1H), and neurons d42 (Figure 1I), indicating haploinsufficiency in heterozygously SHANK2-deleted cells. SHANK2 localization in patient hiPSCs was not altered (Supplementary Figure 2A).

During neuronal differentiation from hiPSCs, the cells pass various stages summarized in Figure 2A. Forty-two days after plating, they reached a mature morphology characterized by expression of synaptic proteins, including SHANK2 and HOMER1 (Figure 2B). To address the early neuronal morphology, we analyzed the growth cone area in early 14-day-old cells (e14) and found reduced sizes in patient cells (Figure 2C), indicating an early developmental delay. Furthermore, the soma size was significantly increased in d21 patient cells, but not in d42 cells (Figures 2D,E). The number of primary neurites was not altered in patient d21 or d42 neurons (Supplementary Figures 2B,C). These changes during early neuronal development went along with significantly increased cell proliferation in these d21 neurons (Figure 2F). In addition, we found significantly decreased percentages of neurons with fragmented nuclei positive for cleaved caspase 3 in d21 (Figure 2G) and d42 neurons (Figure 2H). Assessment of cleaved caspase 3 levels in total protein lysate in a colorimetric assay showed the same trend (Supplementary Figure 2D). Thus, we conclude that SHANK2 deletion correlates with increased soma size, increased proliferation, and decreased apoptosis in immature d21 neurons during development. Decreased apoptosis seems to be a persistent phenotype that is also observed in mature d42 neurons.

Regarding the important role of SHANK molecules at synapses, we analyzed SHANK2 and SHANK3 expression directly at the synapse. The mean intensity of synaptic SHANK2 puncta along neurites was significantly decreased (Figures 3A,B), while the intensity of synaptic SHANK3 puncta was significantly increased in patient-derived cells (Figures 3A,B). No alterations of SHANK3 were found in cellular localization in hiPSCs (Supplementary Figure 2E) and in RNA and protein expression of total cell lysate (Supplementary Figures 2F–H). These data indicate a possible synaptic compensation of SHANK3 for the lacking SHANK2 protein.

To obtain a complete picture of glutamatergic signaling, we assessed pre- and postsynaptic proteins in d42 neurons. For the postsynaptic specialization, we analyzed HOMER1 as an interaction partner of both SHANK2 and SHANK3 and mGluR5 that is included in critical pathways contributing to ASD (Bear et al., 2004; Vicidomini et al., 2017). As reported previously, synaptic HOMER1 decreased (Vicidomini et al., 2017). We found decreased mGluR5 levels and increased synaptic AMPA receptor subunit GluA2 levels (Figures 3A,C), but no changes in GluN1 levels in SHANK2 deletion neurons. Assessment of the total protein intensity for all synaptic stainings revealed the same alterations between parents and patients compared with the mean intensity (Supplementary Figure 3A), stressing that, indeed, the protein concentrations are altered. Regarding the number of puncta per 30 μm of dendrite, mGluR5 was the only protein showing a decrease in the patient cells (Supplementary Figure 3B), suggesting a critical dysregulation of mGluR5.

In the pre-synapse, glutamate reuptake into synaptic vesicles is mediated by vesicular glutamate transporter 1 (vGluT1). We found a significantly increased mean intensity of vGLUT1 immunostaining in SHANK2 patient neurons (Figures 3D,E), while levels of vesicular GABA transporter (vGAT), the respective transporter at inhibitory synapses, were unchanged (Figures 3D,E). Total protein intensity revealed the same results (Supplementary Figure 3C), and the number of puncta per 30 μm was not changed (Supplementary Figure 3D). Indeed, vGluT1 levels were also increased in synapses, co-expressing SHANK2 and vGluT1, as shown by co-localization analysis of the two proteins (Figure 3F and Supplementary Figures 3E–G). In addition, increased vGluT1 levels were found in total RNA, but not in protein lysate (Figure 3G), indicating a specific neuronal synaptic alteration. Immunostaining for bassoon (BSN), a presynaptic scaffolding protein, revealed increased BSN levels in SHANK2 patient neurons (Figures 3D,E), indicating complex presynaptic alterations. However, protein levels in total cell lysate showed no changes between parents and patients (Supplementary Figure 3H), revealing the synapse as a critical point for dysregulation in SHANK2-related ASD.

Since we found decreased mGluR5 levels in SHANK2-deleted hiPSC-derived neurons, we analyzed extracellular signal-regulated kinase 1/2 (ERK1/2) and AKT as main mGluR5 downstream-signaling pathways. In Western Blot analysis of total protein lysate, we did not observe any changes in d21 neurons (Supplementary Figures 4A,B). In d42 neurons, ERK1/2 and phospho-ERK1/2 were significantly decreased (Supplementary Figure 4C) but not AKT or phospho-AKT expression (Supplementary Figure 4D). Western Blot analysis for CREB and phospho-CREB, a transcription factor downstream of the ERK-signaling cascade, did not show any changes in d21 or d42 (Supplementary Figures 4E,F). To analyze neurons only, we performed immunostainings and analyzed ERK1/2, phospho-ERK1/2, and phospho-CREB in NFH and MAP2 positive neurons. Indeed, d21 neurons showed decreased ERK1/2 (Figure 4A) and phospho-ERK1/2 (Figure 4B) expression. Phospho-CREB was not altered (Figure 4C). d42 neurons showed unchanged neuronal ERK1/2 expression (Figure 4D), but both phosphorylated ERK1/2, the active version of ERK1/2 (Figure 4E), and phospho-CREB (Figure 4F) were significantly lower in the patient, speaking for a differential activation of the ERK-signaling cascade under SHANK2 deletion.

To investigate if only ERK1/2 and phospho-CREB or also additional components of the MAPK-signaling cascade are affected by SHANK2 loss, we performed protein arrays, targeting the ERK1/2 pathway in d21 and d42 hiPSC-derived neurons of father, mother and patient. In d21 neurons, the data derived from the patient clustered differently than from the father and the mother, showing both up- and downregulated proteins in comparison to the two parents (Figure 5A). In d42 neurons, the changes in the patient in comparison to the father and the mother were even more obvious, since almost all proteins analyzed showed a drastically lower expression in the patient (Figure 5B). To identify proteins distinctly up- or downregulated between the patient and parents, we filtered only for those proteins that showed a fold change of 0.45 greater or smaller for the patient in comparison to both the mother and the father. In d21 neurons, we obtained a group of four up- and two downregulated proteins (Figure 5C), and, in d42 neurons, all proteins were downregulated, including phospho-ERK1/2 (p44/42 MAP kinase phospho-Tyr204) and phospho-CREB (Figure 5D) that we analyzed before. In the ERK1/2 pathway, a series of kinases is activated by consecutive phosphorylation, leading to the activation of a variety of transcription factors (Molina and Adjei, 2006). The identified proteins represented all levels of the ERK1/2-signaling hierarchy (Figure 5E), supporting our hypothesis of a general dysregulation of the ERK1/2-signaling cascade under SHANK2 deletion.

To confirm that SHANK2 deficiency affects ERK1/2 signaling in general, we performed Western Blot analysis of 7-day-(P7)-old Shank2(+/+), Shank2(+/−), and Shank2(−/−) mice. Hippocampal tissue lysates showed no changes in ERK1/2 or phospho-ERK1/2 expression (Figure 6A), but, in striatum, (Figure 6B) phospho-ERK1/2 was significantly decreased in Shank2(−/−) mice, while total ERK1/2 stayed constant. Interestingly, none of these changes were observed in 70-day-(P70)-old adult mice (Figures 6C,D), indicating a neurodevelopmental dysregulation of the ERK1/2 pathway in striatum. Comparable to the hiPSC neuron data, AKT and phospho-AKT (Supplementary Figure 4G) were not altered in P7 striatum. Completing the picture, mGluR5 was also significantly decreased in P7 striatum (Figure 6E), but not in hippocampus (Figure 6F). P7 striatum also showed a clear trend for decreased phospho-CREB (Figure 6G). These data support the hypothesis of developmental ERK1/2 dysregulation under SHANK2 deficiency during a specific window during neurogenesis that differentially affects specific brain regions.