Research Article

Autism-associated SHANK3 haploinsufficiency causes Ih channelopathy in human neurons

See allHide authors and affiliations

Science  06 May 2016:
Vol. 352, Issue 6286, aaf2669
DOI: 10.1126/science.aaf2669

Faulty channels, not faulty synapses

SHANK3 is a widely expressed scaffolding protein that is enriched in postsynaptic specializations. In mutant mice, SHANK3 mutations cause autism-like behavioral changes and exhibit alterations in synaptic transmission. Yi et al. produced human neurons lacking SHANK3 but not other genes that are also involved in the autism-like disease Phelan-McDermid syndrome. Instead of affecting synapses, SHANK3 mutations primarily caused a channelopathy, with the major phenotype consisting of a specific impairment of HCN channels. Chronic impairment of membrane currents through channelopathy could account for the phenotypes observed in Phelan-McDermid neurons, such as alterations in cognitive functions and the predisposition to epilepsy.

Science, this issue p. 10.1126/science.aaf2669

Structured Abstract


SHANK3 is a scaffolding protein that is enriched in postsynaptic densities of excitatory synapses but ubiquitously expressed in most cells. SHANK3 gene mutations are significantly associated with autism spectrum disorders (ASDs), and deletion of SHANK3 is thought to cause the major symptoms of Phelan-McDermid syndrome. Moreover, increasing evidence links SHANK3 mutations to schizophrenia. Because SHANK3 is a synaptic protein, SHANK3 mutations are thought to predispose to neuropsychiatric disorders by impairing synaptic function. How SHANK3 mutations are pathogenic, however, remains unclear.


Human neurons derived from Phelan-McDermid syndrome patients display complex abnormalities, including synaptic deficits and altered intrinsic electrical properties. Although some of these abnormalities are reversed by SHANK3 reexpression, the altered electrical properties are difficult to reconcile with a primarily synaptic impairment. Moreover, in mice, Shank3 deletions produce behavioral changes and synaptic transmission deficits, although no cellular phenotype has been identified. Here, we explored the pathogenetic mechanism of human SHANK3 mutations with a conditional genetic approach in human neurons and correlated the results with those obtained in Shank3-mutant mouse neurons. We introduced conditional SHANK3 deletions into human embryonic stem cells and examined isogenic control and heterozygous and homozygous SHANK3-mutant neurons derived from these conditionally mutant cells. In addition, we analyzed developing mouse Shank3-mutant neurons and compared their phenotype with that of human SHANK3-mutant neurons.


Heterozygous and homozygous SHANK3-mutant human neurons displayed diverse abnormalities, ranging from a massive increase in input resistance to increased excitability, modest impairments in dendritic arborization, and decreases in synaptic transmission. Because the increased input resistance suggested an altered channel conductance as a primary impairment, we tested various conductances. We found that the SHANK3 mutations caused a profound impairment in hyperpolarization-activated cation (Ih) currents, which are mediated by hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. This impairment produced the increased input resistance; moreover, chronic pharmacological inhibition of Ih currents in wild-type human neurons impaired dendritic arborization and synaptic transmission similar to the SHANK3 mutations. Mechanistically, we detected a direct interaction of HCN channels with SHANK3 protein and observed a decrease in HCN-channel proteins in SHANK3-mutant neurons. Finally, we found that developing hippocampal neurons cultured from heterozygous and homozygous Shank3-mutant mice also exhibited an increased input resistance, reduced Ih currents, and an increased excitability similar to SHANK3-mutant human neurons.


Using human neurons with conditional SHANK3 mutations, we found that SHANK3 mutations impair Ih-channel function, thereby increasing neuronal input resistance and enhancing neuronal excitability. This impairment in intrinsic electrical properties accounts, at least in part, for the decreased dendritic arborization and synaptic transmission of SHANK3-mutant neurons. The reduced Ih-current phenotype manifests early in neuronal development and is similarly observed in immature Shank3-mutant mouse neurons. We propose that, in addition to having a specifically postsynaptic function, SHANK3 protein may perform a general role during neurodevelopment by scaffolding HCN channels that mediate Ih currents in neurons and nonneuronal cells consistent with the ubiquitous expression of SHANK3. Thus, we hypothesize that SHANK3 mutations induce an Ih channelopathy that contributes to ASD pathogenesis and may be amenable to pharmacological intervention.

Conditional SHANK3 deletion in human neurons impairs Ih channel.

Comparison of isogenic control and SHANK3-deficient human neurons reveals that heterozygous and homozygous SHANK3 mutations dramatically decrease Ih-channel function, resulting in multifarious secondary impairments, including a decrease in dendritic arborization and synaptic responses and an increase in input resistance and neuronal excitability.


Heterozygous SHANK3 mutations are associated with idiopathic autism and Phelan-McDermid syndrome. SHANK3 is a ubiquitously expressed scaffolding protein that is enriched in postsynaptic excitatory synapses. Here, we used engineered conditional mutations in human neurons and found that heterozygous and homozygous SHANK3 mutations severely and specifically impaired hyperpolarization-activated cation (Ih) channels. SHANK3 mutations caused alterations in neuronal morphology and synaptic connectivity; chronic pharmacological blockage of Ih channels reproduced these phenotypes, suggesting that they may be secondary to Ih-channel impairment. Moreover, mouse Shank3-deficient neurons also exhibited severe decreases in Ih currents. SHANK3 protein interacted with hyperpolarization-activated cyclic nucleotide-gated channel proteins (HCN proteins) that form Ih channels, indicating that SHANK3 functions to organize HCN channels. Our data suggest that SHANK3 mutations predispose to autism, at least partially, by inducing an Ih channelopathy that may be amenable to pharmacological intervention.

Haploinsufficiency of SHANK3 constitutes one of the more frequent single-gene mutations in autism spectrum disorders (ASDs) (15). Moreover, SHANK3 is among several genes deleted in Phelan-McDermid syndrome (22q13.3 deletion syndrome), and SHANK3 haploinsufficiency may be the most important contributing factor to Phelan-McDermid syndrome pathology (15). In addition, emerging evidence links SHANK3 mutations to schizophrenia (6, 7), and SHANK3 overexpression has also been implicated in neuropsychiatric disorders (8). Analysis of human neurons differentiated from induced pluripotent stem cells (iPSC) from patients with Phelan-McDermid syndrome have revealed two apparently disparate phenotypes, an impairment in synaptic transmission and an increase in input resistance (9). The synaptic impairments of Phelan-McDermid neurons are rescued with SHANK3 (9), consistent with the presence of SHANK3 in postsynaptic specializations where SHANK3 is thought to function as a scaffolding protein that organizes receptor signaling (1012). Thus, Phelan-McDermid syndrome may result from a synaptic impairment caused by SHANK3 haploinsufficiency. The dramatically increased input resistance of Phelan-McDermid neurons, however, remains an enigma (9). Because human neurons with sole SHANK3 mutations have not been analyzed, the role of SHANK3 in the phenotypes of Phelan-McDermid neurons remains incompletely understood, as does the functional effect of a pure SHANK3 haploinsufficiency on human neurons. In mice, heterozygous and homozygous Shank3 mutations cause a behavioral phenotype resembling autism and/or obsessive-compulsive disorders and produce changes in synaptic transmission (1321). However, the underlying molecular mechanisms and possible contributory nonsynaptic effects are unclear, as are the cellular effects of murine Shank3 mutations.

To address these issues that are crucial for progress toward understanding ASDs and Phelan-McDermid syndrome, we generated human neurons with conditional heterozygous and homozygous SHANK3 loss-of-function mutations. Unexpectedly, we found that, besides causing synaptic impairments, loss of SHANK3 function selectively and severely impaired hyperpolarization-activated cation (Ih) currents. We observed that SHANK3 protein directly bound to hyperpolarization-activated cyclic nucleotide-gated channel (HCN) proteins mediating Ih currents and that at least some of the synaptic impairments in SHANK3-mutant human neurons are an indirect result of the Ih channelopathy produced by the SHANK3 mutations during neuronal development. Finally, we observed a similar phenotype in Shank3-deficient mouse neurons, suggesting a general function for SHANK3 in scaffolding HCN channels.

Generation of heterozygous SHANK3 conditional knockout (cKO) mutations

To investigate the role of SHANK3 mutations in human neurons, we constructed conditional mutations in the SHANK3 gene in human H1 embryonic stem cells (ES cells) using homologous recombination (Fig. 1, A and B, and fig. S1) (22). We chose this approach because it enables generation of matching control and mutant neurons from the same ES cell clones, thus eliminating subclone-to-subclone variability. The SHANK3 locus expresses multiple transcripts that encode different SHANK3 isoforms with distinct protein interaction domains (Fig. 1A and fig. S2A). To conditionally delete major SHANK3 isoforms, we targeted exon 13 of the SHANK3 gene whose deletion causes a frameshift in all major SHANK3 mRNAs.

Fig. 1 SHANK3 haploinsufficiency impairs dendritic development of human neurons.

(A) Diagram of the SHANK3 gene and of the three major SHANK3 transcripts that are blocked by conditional deletion of exon 13 (yellow). (B and C) SHANK3 targeting strategy in human ES cells. Homologous recombination was mediated using recombinant adeno-associated virus (AAV) (B) and confirmed by PCR (fig. S1B). The PGK-puromycin resistance cassette (brown box) was excised by Flp-recombinase to generate the cKO allele (SHANK3+/cKO) (C). SHANK3+/cKO ES cells were converted into human SHANK3+/+ or SHANK3+/− neurons by coexpression of Ngn2 with either mutant inactive Cre-recombinase (ΔCre) or active Cre-recombinase (Cre). (D) Reduction of SHANK3 protein levels in human SHANK3+/− neurons derived from two independent SHANK3+/cKO ES cell clones. (E) Representative images of SHANK3+/+ and SHANK3+/− neurons (day 21) labeled by double immunofluorescence for MAP2 (red) and synapsin (green). (F) Representative dendritic arborization analyses by MetaMorph software of SHANK3+/+ and SHANK3+/− neurons (sparsely transfected with EGFP). (G) SHANK3 haploinsufficiency impairs dendritic arborization (summary graphs of indicated parameters measured in matching SHANK3+/+ and SHANK3+/− neurons derived from independent SHANK3+/cKO ES cell clones; normalized to SHANK3+/+ controls). (H) Representative images of SHANK3+/+ and SHANK3+/− dendrites stained for MAP2 (red) and synapsin (green) for analysis of synaptic puncta by MetaMorph software. (I) Summary graphs of dendritic synaptic puncta density and size in SHANK3+/+ and SHANK3+/− neurons derived from two independent SHANK3+/cKO ES cell clones. Data in (G) and (I) are means ± SEM. Numbers of cells/independent cultures analyzed are shown in the bars. Statistical significance was evaluated by Student’s t test (*P < 0.05; **P < 0.01). For additional data, see figs. S1 and S2.

We converted heterozygous conditionally mutant SHANK3+/cKO ES cells into neurons by forced expression of the transcription factor Ngn2, which generates a homogenous population of glutamatergic excitatory neurons with abundant synapse formation (23). We coexpressed active (Cre) or mutant Cre-recombinase (ΔCre) with Ngn2 during neuronal differentiation to produce precisely matching wild-type [ΔCre (SHANK3+/+)] and heterozygous mutant neurons [Cre (SHANK3+/−)] (24, 25), using two independently targeted heterozygous clones of SHANK3+/cKO ES cells to control for clonal variation (Fig. 1C). Immunoblotting and polymerase chain reaction (PCR) demonstrated that the heterozygous SHANK3 KO decreased SHANK3 expression but left SHANK1 and SHANK2 expression unchanged (Fig. 1D and figs. S1, D and E, S2B, and S3). No significant changes in other synaptic proteins were observed except for a decrease in PSD95 that is known to bind to SHANKs (fig. S2, C and D) (1012).

SHANK3 haploinsufficiency impairs dendritic arborization, intrinsic electrical properties, and synaptic transmission in human neurons

Human SHANK3+/− neurons exhibited a typical neuronal morphology with abundant synapse formation (Fig. 1E). When we quantified the morphological features of matching SHANK3+/+ and SHANK3+/− neurons derived from independently derived ES cell clones, we observed that SHANK3 haploinsufficiency significantly decreased the length and branching of neurites but had no significant effect on the density or size of synapsin-positive synapses (Fig. 1, F to I).

Next, we performed whole-cell patch-clamp recordings from matching SHANK3+/+ and SHANK3+/− neurons. SHANK3 haploinsufficiency caused a large increase (~25 to 33%) in input resistance without a change in capacitance (Fig. 2A). Moreover, SHANK3+/− neurons exhibited a major decrease in evoked excitatory postsynaptic currents (EPSCs) (~40 to 50% decrease) and in the amplitude of spontaneous miniature EPSCs (mEPSCs) (~25% decrease) but not in other mEPSC parameters (Fig. 2, B and C, and fig. S2E). Overall, these results resemble those obtained with Phelan-McDermid neurons (9), supporting the notion that despite the large genomic deletion present in Phelan-McDermid neurons, SHANK3 haploinsufficiency may account for most Phelan-McDermid syndrome phenotypes.

Fig. 2 SHANK3 haploinsufficiency increases electrical input resistance of human neurons and decreases overall synaptic strength.

(A) Input resistance (Rin) but not capacitance (Cm) is increased in SHANK3+/− neurons (summary graphs from matching human SHANK3+/+ and SHANK3+/− neurons derived from two independent SHANK3+/cKO ES cell clones). (B) Evoked synaptic transmission is decreased in SHANK3+/− human neurons [representative EPSC traces (left) and EPSC amplitude summary graphs (right) for independent sets of neurons]. (C) Amplitudes but not frequencies of spontaneous mEPSCs (monitored in 1 μM tetrodotoxin) are impaired in SHANK3+/− neurons [top, representative traces; bottom, cumulative plots and summary graphs of the mEPSC frequency (left) and amplitude (right)]. Data are means ± SEM. Numbers of cells/cultures analyzed are shown in bars. Statistical significance was evaluated by either Student’s t test (bar graphs) or Kolmogorov-Smirnov test (cumulative probability plots) (**P < 0.01; ***P < 0.001).

Homozygous SHANK3 deletion aggravates SHANK3 haploinsufficiency phenotype

To inquire whether homozygous SHANK3 mutations produce phenotypes similar to those of heterozygous SHANK3 mutations, we generated mutant H1 ES cells with homozygous conditional SHANK3 mutations (22) and converted them into matching SHANK3+/+ or SHANK3−/− neurons by coexpression of Ngn2 and inactive or active Cre-recombinase, respectively (Fig. 3A and fig. S4). Immunoblotting confirmed that the major SHANK3 protein isoforms were no longer expressed in SHANK3−/− neurons (Fig. 3B and fig. S4E). Quantitative analyses uncovered a similar, but more severe, phenotype in SHANK3−/− neurons than in SHANK3+/− neurons, with a significant decrease in dendritic arborization and in synapse density (Fig. 3, C to F, and fig. S5). Electrophysiologically, SHANK3−/− neurons also exhibited a large increase in input resistance, a massive decrease in evoked EPSC amplitude, and a significant decrease in mEPSC amplitude similar to SHANK3+/− neurons (Fig. 3, G to I). In addition, SHANK3−/− neurons displayed a decrease in capacitance and a notable decline in mEPSC frequency (Fig. 3, G and I).

Fig. 3 Homozygous SHANK3 deletion aggravates SHANK3 haploinsufficiency phenotype.

(A) Diagram of homozygous SHANK3cKO/cKO alleles (see supplementary materials for details). (B) SHANK3−/− neurons lack major SHANK3 proteins. Images depict immunoblots of matching SHANK3+/+ and SHANK3−/− neurons derived from two independent SHANK3cKO/cKO ES cell clones. (C) Representative images of matching SHANK3+/+ and SHANK3−/− neurons (day 21) stained by double immunofluorescence for MAP2 (red) and synapsin (green). (D) Homozygous SHANK3 deletion severely impairs dendritic arborization [summary graphs of indicated parameter (normalized to SHANK3+/+ controls) measured in matching SHANK3+/+ and SHANK3−/− neurons derived from two independent SHANK3cKO/cKO ES cell clones]. (E) Representative images of dendrites stained for MAP2 (red) and synapsin (green) for analysis of synaptic puncta in SHANK3+/+ and SHANK3−/− neurons. (F) Homozygous SHANK3 deletion reduces synapse density (summary graphs of synaptic puncta density and size on proximal dendrites of isogenic SHANK3+/+ and SHANK3−/− neurons from two independent SHANK3 cKO/cKO ES cell clones). (G) Homozygous SHANK3 deletion increases neuronal Rin (top) and decreases Cm (bottom). (H) Homozygous SHANK3 deletion reduces evoked EPSC amplitudes (left, representative traces; right, summary graphs of EPSC amplitudes). (I) Homozygous SHANK3 deletion decreases the frequency and amplitude of mEPSCs [top, representative traces; bottom, cumulative plots and summary graphs of mEPSC interevent intervals and frequency (bottom left) or mEPSC amplitudes (bottom right)]. Data in bar diagrams are means ± SEM. Numbers of cells/cultures analyzed are shown in bars. Statistical significance was evaluated by Student’s t test (bar graphs) or Kolmogorov-Smirnov test (cumulative probability plots) (*P < 0.05; **P < 0.01; ***P < 0.001). For additional data, see figs. S3 to S5.

Viewed together, our results suggest that conditional heterozygous and homozygous deletions of SHANK3 produce similar phenotypes in human neurons. Although the SHANK3-mutant neurons clearly exhibit synaptic impairments, their increased input resistance and decreased dendritic arborization cannot be readily explained in terms of a synaptic change, prompting us to search for other pathogenetic mechanisms.

Heterozygous and homozygous SHANK3 mutations impair Ih currents

The input resistance of neurons depends at least in part on their ionic conductances and ion channels. Thus, we first asked whether changes in voltage-gated Na+ channels or K+ channels (delayed rectifier), which generate action potentials and determine neuronal excitability, could be responsible for the increased input resistance of SHANK3-mutant neurons, prompted in part by binding of SHANK3 to K+ channels (26). However, we detected no changes in Na+ or K+currents in SHANK3-mutant neurons (fig. S6). We next investigated whether decreased ionotropic glutamate receptor activation caused by impaired synaptic transmission in SHANK3-mutant neurons might be responsible. However, blocking ionotropic glutamate receptors had no effect on input resistance in SHANK3+/+ or SHANK3+/− neurons (Fig. 4A).

Fig. 4 SHANK3 deletions impair Ih currents in human neurons, and SHANK3 proteins interact with HCN channels.

(A) Neuronal silencing by blocking glutamate receptors with CNQX (20 μM) and AP5 (50 μM) does not alter neuronal Rin. (B) Blocking Ih currents with ZD7288 (100 μM) increases the Rin of wild-type neurons, abolishing the difference between wild-type and mutant neurons. (C) SHANK3+/− and SHANK3−/− neurons exhibit a more negative Vrest than SHANK3+/+ neurons; inhibition of Ih currents in SHANK3+/+ neurons with ZD7288 (5 μM) abolishes the difference. (D and E) SHANK3+/− neurons (D) and SHANK3−/− neurons (E) exhibit decreased Ih-current amplitudes compared with matching SHANK3+/+ neurons (left, experimental protocol and sample traces; right, current/voltage relation of Ih currents, which are validated by inhibition with ZD7288). (F) Kinetics of Ih-current activation is impaired in SHANK3+/− and SHANK3−/− neurons [left, representative capacity- and leak current–subtracted Ih-current recordings fitted with a double-exponential function (yellow line superimposed on traces); center and right, summary graphs of time constants (τ) and amplitudes of fast and slow components of Ih-current activation at a test potential of –120 mV]. (G) SHANK3 protein binds to HCN channels mediating Ih currents. Representative immunoblots document coimmunoprecipitation of Myc-tagged SHANK3 with HA-tagged HCN1, HCN2, and HCN3 proteins coexpressed in transfected HEK293T cells; cells expressing only one or the other protein serve as negative controls. (H) Levels of endogenous HCN3 and HCN4 proteins are decreased in SHANK3−/− neurons, as determined by quantitative immunoblotting (for representative blots, see fig. S10). Data are means ± SEM. Numbers of cells/cultures analyzed are shown in bars or parentheses. Statistical significance was evaluated by Student’s t test [bar graphs in (F) and (H)] or two-way analysis of variance (ANOVA) [bar graphs in (A) to (C)] and two-way repeated measure ANOVA [I-V plots in (D) and (E)] followed by Bonferroni’s post hoc test (*P < 0.05; **P < 0.01; ***P < 0.001).

A third candidate for an impaired membrane conductance as a cause of the increased input resistance is Ih currents that are mediated by HCN channels. In mammals, HCN channels are encoded by four genes (HCN1 to HCN4) and are expressed, like SHANKs, in neuronal and nonneuronal cells (2730) (fig. S7). HCN channels mediate hyperpolarization-activated Ih currents that depolarize membranes toward the action-potential threshold and reduce membrane resistance. Ih currents control neuronal excitability, membrane resting potentials, dendritic integration of synaptic potentials, and rhythmic oscillation of neurons (29, 30). Given their multifaceted functions, impairments of Ih currents can have profound consequences for neuronal network activity [e.g., see (3133)].

Strikingly, we found that the Ih-current inhibitor ZD7288 (34, 35) increased the input resistance of wild-type neurons dramatically but had only a small effect on SHANK3-mutant neurons, thereby abolishing the difference in input resistance between wild-type and SHANK3-mutant neurons (Fig. 4B). Moreover, SHANK3-deficient neurons displayed an increased resting membrane potential, and the addition of ZD7288 to wild-type neurons increased their resting potential to that of SHANK3-deficient neurons (Fig. 4C).

Because these results suggest that the changed electrical properties of SHANK3-mutant neurons may be due to an impairment of Ih currents, we next directly measured Ih currents in precisely matching SHANK3+/+ and SHANK3+/− or SHANK3−/− neurons using whole-cell recordings (Fig. 4, D and E, and fig. S8). Ih currents were readily activated by brief (2 s) hyperpolarizing voltage steps, exhibited a reversal potential of –32 mV, and were inhibited by extracellular Cs+ and ZD7288 (fig. S8). Compared with SHANK3+/+ neurons, both SHANK3+/− and SHANK3−/− neurons exhibited a severe decrease in Ih-current density (Fig. 4, D and E) and a deceleration of Ih-current activation (Fig. 4F). However, other basic properties of Ih currents, such as their half-maximal activation potential (V50), were not significantly altered (fig. S8). Furthermore, depolarizing voltage sag responses that are evoked by brief injections of hyperpolarizing currents in current-clamp mode (36, 37) were also significantly impaired in SHANK3+/− and SHANK3−/− neurons; these impairments again were occluded by ZD7288, suggesting that these phenotypes are also due to a specific alteration in Ih-channel function (fig. S8).

Viewed together, these data suggest that SHANK3 mutations cause Ih-channel dysfunction in human neurons. To investigate the possibility that SHANK3 interacts with HCN channels, we coexpressed SHANK3 with HCN1, HCN2, or HCN3 channels in human embryonic kidney (HEK) 293T cells. Coimmunoprecipitations revealed that SHANK3 bound to all three isoforms of HCN channels in this assay (Fig. 4G). Mapping of the interaction domains using glutathione S-transferase (GST) pull-downs suggested that the SHANK3 ankyrin repeats directly bind to HCN channels (fig. S9). Finally, measurements of the levels of two human HCN isoforms to which antibodies were available, HCN3 and HCN4, demonstrated that the SHANK3 deletion significantly decreased the levels of endogenous HCN proteins in human neurons, consistent with a direct interaction (Fig. 4H and fig. S10).

Because Ih currents are major determinants of neuronal excitability (29, 30), we examined the effects of SHANK3 mutations on action potential generation (Fig. 5). Consistent with the increased input resistance, SHANK3+/− and SHANK3−/− neurons fired significantly more action potentials than SHANK3+/+ neurons in response to depolarizing current injections; this phenotype was abolished by the addition of ZD7288 and thus was Ih-current-dependent (Fig. 5). Moreover, because sustained rhythmic oscillations are a hallmark of neuronal circuits in various brain regions that overlap with highly enriched regions of SHANK3 expression (14, 29, 38), we monitored the spontaneous spiking activity of SHANK3-mutant neurons. When neurons were held at the resting membrane potential, a large proportion of wild-type cells (~60%) fired action potentials spontaneously and regularly (fig. S11). In contrast, SHANK3-mutant neurons fired fewer action potentials; again, this phenotype was reversed by addition of the Ih-current inhibitor ZD7288 (fig. S11).

Fig. 5 SHANK3 mutations render neurons hyperexcitable by impairing Ih currents.

(A) SHANK3+/− mutant neurons reach action potential (AP) firing threshold earlier than matching SHANK3+/+ human neurons and exhibit a steeper input-output relationship, as assessed by the number of APs elicited by increasing current injections (from –10 to +50 pA, 1-s, 5-pA increments) during current-clamp recordings. Ih-channel inhibition with ZD7288 abolishes the difference (left, experimental protocol and representative traces; right, plots of the mean AP number versus injected current). (B) Summary graphs of active electrical properties of matching SHANK3+/+ and SHANK3+/− neurons (measured without or with ZD7288 application). From left to right: AP firing threshold, AP amplitude, and AP after-hyperpolarization amplitude (AHP). (C) Same as (A), but for matching SHANK3−/− and SHANK3+/+ neurons. (D) Same as (B), but for matching SHANK3−/− and SHANK3+/+ neurons. Data are means ± SEM. Numbers of cells/cultures analyzed are shown in bars or parentheses. Statistical significance was evaluated by two-way ANOVA (bar graphs) or two-way repeated measure ANOVA followed by Bonferroni’s post hoc test (input-output plots) (n.s., not significant; *P < 0.05; **P < 0.01; ***P < 0.001).

Shank3 deletions impair Ih currents in developing mouse neurons

To determine whether Shank3-mutant mice also exhibit an impairment in Ih currents, we cultured hippocampal neurons from newborn littermate Shank3+/+, Shank3+/−, and Shank3−/− mice. In the Shank3-mutant mice, exons encoding the PDZ domain of SHANK3 are deleted, similar to our conditionally SHANK3-mutant human neurons (14). We observed an overall very similar phenotype in developing mouse neurons as in human neurons (Fig. 6 and figs. S12 and S13). Specifically, quantitative imaging revealed that at 8 days in vitro (DIV8), homo- but not heterozygous Shank3 mutations caused a significant impairment in dendritic arborization and synapse density similar to human SHANK3 deletions (Fig. 6, A and B, and fig. S12). Electrophysiological recordings showed that although the total input resistance of mouse neurons was lower than that of human neurons, both hetero- and homozygous Shank3 mutations produced a large increase in input resistance (Fig. 6C). Homozygous Shank3 deletions also significantly decreased cell capacitance and increased the resting membrane potential, similar to human mutations. We then directly measured Ih currents and detected a massive impairment in both hetero- and homozygous Shank3-deficient mouse neurons (Fig. 6, D and E, and fig. S12). The Ih-current amplitude was decreased more than twofold by Shank3 mutations, the Ih-current voltage sag was reduced, and the Ih-current activation kinetics was significantly decelerated. In all of these phenotypes, the homozygous mutation was more deleterious than the heterozygous mutation. These changes closely resemble those observed in human SHANK3-deficient neurons and also led to a markedly increased excitability in mouse Shank3-deficient neurons (Fig. 6F).

Fig. 6 Developing hippocampal neurons from Shank3 knockout mice reproduce phenotype of human SHANK3-mutant neurons.

(A and B) Dendritic arborization (A) and synapse formation (B) are impaired in developing Shank3-deficient mouse neurons [top, representative images of hippocampal neurons (A) and dendritic segments (B); bottom, summary graphs of the indicated dendritic, cellular, and synaptic parameters]. Neurons cultured from littermate Shank3+/+, Shank3+/−, or Shank3−/− mice were stained at DIV8 for MAP2 (red) and synapsin (green). (C) Hetero- and homozygous Shank3 deletions increase neuronal input resistance (top), decrease cell capacitance (center), and enhance the resting membrane potential (bottom) in hippocampal neurons cultured from littermate Shank3+/+, Shank3+/−, and Shank3−/− mice and analyzed at DIV8-9. (D) Hetero- and homozygous Shank3 deletions impair neuronal Ih currents in hippocampal neurons at DIV8-9 (left, experimental protocol and sample traces; right, summary graph of the current/voltage relation). (E) Hetero- and homozygous Shank3 deletions decelerate Ih-current activation [left top, experimental protocol; left bottom, capacitance- and leak current-subtracted representative traces fitted with the sum of two exponential functions shown superimposed on the traces in yellow; right, summary graphs of activation time constants (τ) and component amplitudes obtained at a test potential of –120 mV]. (F) Hetero- and homozygous Shank3 deletions render hippocampal neurons hyperexcitable (left, experimental protocol and representative traces of stepwise depolarizing current injections; right, summary plots of the AP number versus injected current during current-clamp recordings of neurons analyzed at DIV8-9). Data are means ± SEM. Numbers of cells/cultures analyzed are shown in bars or parentheses. Statistical significance was evaluated by one-way ANOVA followed by Tukey’s post hoc test (bar graphs) or two-way repeated measure ANOVA followed by Bonferroni’s post hoc test (I-V plot) (*P < 0.05; **P < 0.01; ***P < 0.001). For additional data, see figs. S12 and S13.

Chronic Ih-current inhibition impairs neuronal development similar to SHANK3 mutations

A decrease in Ih currents by SHANK3 mutations likely accounts for the electrophysiological changes of SHANK3-mutant neurons, but can it also account, at least in part, for their morphological and synaptic changes? To investigate this question, we chronically inhibited Ih channels in wild-type neurons by continuous application of low concentrations of ZD7288 (1 and 5 μM) during neuronal differentiation.

Chronic inhibition of Ih channels dramatically impaired neuronal morphology in a manner indistinguishable from that of the SHANK3 haploinsufficiency (Fig. 7A and fig. S14). Specifically, chronic application of ZD7288 caused a dose-dependent decrease in neurite outgrowth, number of primary processes, and dendritic branching. Moreover, ZD7288 decreased the synapse density, again suggesting that major phenotypes of SHANK3 mutations may represent indirect effects of an Ih-current impairment (Fig. 7B). In addition, chronic inhibition of Ih channels in human neurons increased the neuronal input resistance and impaired spontaneous synaptic transmission, measured after washout of the ZD7288 used to inhibit Ih currents (Fig. 7, C and D). These results suggest that long-term inhibition of Ih currents has dramatic effects on multiple facets of neuronal development.

Fig. 7 Chronic inhibition of Ih currents in human neurons mimics SHANK3 haploinsufficiency phenotype.

(A) Chronic partial inhibition of Ih currents in wild-type SHANK3+/+ neurons causes dendritic arborization defects similar to SHANK3+/− haploinsufficiency. Matching SHANK3+/+ and SHANK3+/− neurons were differentiated from SHANK3+/cKO ES cells; SHANK3+/+ neurons were treated with low-dose ZD7288 (1 μM or 5 μM) from days 3 to 21 of neural induction (top, representative images of neurons double-labeled for MAP2 and synapsin; bottom, summary graphs of indicated parameters normalized to the untreated SHANK3+/+ control). (B) Chronic inhibition of Ih currents in SHANK3+/+ neurons decreases synapse numbers. Experiments were performed as described for (A) (left, representative images of dendritic segments double labeled for MAP2 and synapsin; right, summary graphs of density and size of synaptic puncta). (C) Chronic inhibition of Ih currents in SHANK3+/+ neurons significantly increases Rin (left) and decreases cell Cm (right). Experiments were performed as described for (A); recordings were performed after washout of ZD7288. (D) Chronic inhibition of Ih currents in SHANK3+/+ neurons with ZD7288 (1 μM) reduces the frequency and amplitude of spontaneous mEPSCs (top, representative traces; bottom left, summary plots of the interevent interval and summary graph of the mEPSC frequency; bottom right, same for the mEPSC amplitude). Experiments were performed as described for (A); recordings were performed after washout of ZD7288. Data are means ± SEM. Numbers of cells/cultures analyzed are shown in the bars. Statistical significance was evaluated by Student’s t test [bar graphs in (C) and (D)], one-way ANOVA followed by Tukey’s post hoc test [bar graphs in (B)], or two-way ANOVA followed by Bonferroni’s post hoc test [bar graphs in (A)] or Kolmogorov-Smirnov test [cumulative probability plots in (D)] (*P < 0.05; **P < 0.01; ***P < 0.001).


Many mutations are thought to predispose to idiopathic ASDs by causing primary impairments in synaptic transmission (15). Our data show that SHANK3 haploinsufficiency impairs synaptic function but also demonstrate that SHANK3 haploinsufficiency decreases Ih-channel function as a primary impairment, which in turn produces major changes in intrinsic neuronal properties and secondarily affects synaptic function. Indeed, the fact that several salient phenotypes of hetero- and homozygous SHANK3-mutant human neurons were reproduced by pharmacologic inhibition of Ih channels suggests that Ih-channel dysfunction is a major effect of SHANK3 haploinsufficiency. Moreover, the very similar phenotypes produced by Shank3 mutations in developing mouse neurons, which also severely impaired Ih currents, indicates a general role for SHANK3 in scaffolding HCN channels during neuronal development at a period coinciding with the manifestation of ASDs. A role for SHANK3 as a scaffolding protein for HCN channels is plausible given the broad expression of SHANK3 in nonneuronal cells that also express HCN channels, and SHANK3 may function to enrich HCN channels at postsynaptic sites together with other proteins (31, 32). Thus, we would like to suggest that an Ih-current impairment is a major pathogenetic force of SHANK3 mutations in predisposing to ASDs and in Phelan-McDermid syndrome. Changes in Ih-channel function can conceivably be influenced pharmacologically, suggesting that pharmacological manipulation of Ih channels may be therapeutically beneficial.

HCN-channel mutations have been linked clinically with human neurological disorders, including epilepsy, sleep disorder, and impaired learning, which are commonly associated with enhanced neuron firing and/or aberrant neuronal firing patterns and are also observed in mice with HCN-channel mutations (29, 30, 3841). The symptoms resulting from impaired HCN channels agree well with the hypothesized involvement of SHANK3 deletions in ASDs and Phelan-McDermid syndrome that are also commonly associated with intellectual disability, impaired learning and memory, and epilepsy (15). Therefore, our data collectively suggest that impairment of HCN-channel function may contribute to the manifestations of ASDs in patients with SHANK3 mutations and Phelan-McDermid syndrome.

Methods summary

See the supplementary materials for full details of the materials and methods (22).

Generation and analysis of human ES cells with heterozygous and homozygous SHANK3 cKO alleles

SHANK3-mutant ES cells were generated from H1 ES cells (passage 40; WiCell, using recombinant adenoassociated virus (AAV)–mediated homologous recombination (Fig. 1A and fig. S1A) (24, 25). AAVs used for gene targeting contained a puromycin-resistance cassette surrounded by FRT sites (for Flp-recombinase mediated deletion). The cassette was inserted into the 3′ intron of exon 13 (76 base pairs) of the human SHANK3 gene, exon 13 was flanked by LoxP sites (for Cre-recombinase mediated deletion), and 5′ and 3′ DNA homology arms were added. The same gene-targeting template was also used in the generation of homozygous SHANK3 cKO ES cells, except that a CRISPR/Cas9 targeting method was used to specifically target the second wild-type allele of the SHANK3 gene and not the already-targeted first SHANK3 cKO allele. Heterozygous and homozygous SHANK3 cKO ES cell clone mutations were confirmed by PCR, and the puromycin-resistance cassette was removed by Flp-recombinase. To generate isogenic control and SHANK3-mutant human neurons, hetero- or homozygous SHANK3 cKO ES cells were transdifferentiated to neurons using forced expression of Ngn2, as described (23). Lentiviruses encoding active (Cre) or inactive Cre-recombinase (ΔCre) were applied at 1 day before induction of neuron differentiation. Neurons were analyzed at day 21 to 23 in most experiments. For the experiments of chronic Ih-current inhibition with low-dose ZD7288 (Fig. 7), wild-type human neurons were treated with 1 μM or 5 μM ZD7288 from day 3 to day 21.

Generation and analysis of Shank3-mutant hippocampal mouse neurons

Breedings of heterozygous Shank3-mutant mice with deletion of the PDZ domain (exons 13 to 16) (14) (B6.129-Shank3tm2Gfng/J; Jackson Labs stock No. 017688) were used to produce littermate wild-type, heterozygous, and homozygous Shank3-mutant mice. Hippocampal neurons were cultured from newborn mice (42, 43) and analyzed as immature developing neurons at DIV8-9. Mouse genotyping was performed by PCR using the Jackson Laboratory protocol.

Morphological analyses

Dendritic arborizations were analyzed in neurons that were sparsely transfected with an enhanced green fluorescent protein (EGFP) to obtain fluorescent images of individual neurons. Confocal images were analyzed unbiasedly using the “neurite outgrowth” application on MetaMorph software. For synapse morphology analyses, fixed neurons were stained by double immunofluorescence with antibodies to mitogen-activated protein kinase kinase (MAP2) (to stain for dendrites) and synapsin (to label presynaptic terminals) or HOMER1 (to label postsynaptic specializations), and images were again quantified using MetaMorph software (Molecular Devices) (44).

Electrophysiological recordings

Whole-cell patch-clamp recordings were performed essentially as described (23, 42). EPSCs were pharmacologically isolated with 50 μM picrotoxin (PTX) and recorded at a –70 mV holding potential in voltage-clamp mode in response to extracellular stimulation with a concentric bipolar electrode (43). Spontaneous mEPSCs were monitored in the presence of tetrodotoxin (1 μM). Recordings of the intrinsic and active membrane properties were generally recorded in human neurons in the presence of 50 μM PTX (unless otherwise stated), and in hippocampal mouse neurons in the presence of CNQX (6-cyano-7-nitroquinoxaline-2,3-dione) (20 μM), AP5 (2-amino-5-phosphonopentanoic acid) (50 μM), and PTX (50 μM).

Input resistance (Rin) was calculated as the slope of linear fits of current-voltage plots generated from a series of increasing current injection steps in current-clamp mode. Ih-channel activity was measured in voltage-clamp mode as the amplitude of the slowly activating inward current component elicited by 2-s voltage steps from –50 to –120 mV in 10-mV increments from a holding potential of –40 mV with 2 mM 4-aminopyridine (4-AP) and 0.5 mM BaCl2 in the bath solution. Depolarizing voltage-sag responses were evoked by brief injections of hyperpolarizing currents in current-clamp mode. Voltage-dependent Na+ (INa) and K+ (IKD) currents were recorded in voltage-clamp mode at a holding potential of –70 mV in the presence of 2-mM 4-AP; voltage steps ranging from –90 to +40 mV were delivered at 10-mV increments. Intrinsic action potential firing properties of neurons were recorded in current-clamp mode. To assess neuronal excitability, first minimal currents were introduced to hold membrane potential around −65 to −70 mV, then increasing amounts of depolarizing currents were injected for 1 s in stepwise manner. For spontaneous AP firing, cells were held at their resting membrane potential (Vrest); no current was injected. The Vrest obtained during spontaneous firing was determined as the mean steady-state voltage recorded during interspike intervals. All experiments were performed at room temperature.

Protein-protein interaction analyses

Protein-protein interaction analyses were performed by coimmunoprecipitation of Myc-tagged full-length mouse Shank3 protein with hemagglutinin (HA)–tagged full-length human HCN proteins expressed in HEK293T cells. Coimmunoprecipitations of truncated Shank3 proteins with full-length HCN proteins were also performed to map the responsible interaction domain on Shank3. To further demonstrate the interaction between Shank3 and HCN proteins, a GST pull-down assay was performed using purified Shank3 ankyrin repeats domain and full-length HCN1 protein.


All immunoblots were visualized by fluorescently labeled secondary antibodies and quantified on Odyssey CLx Infrared Imager and Odyssey software (LI-COR Biosciences). Signals were normalized to human guanine nucleotide dissociation inhibitor (GDI) as the neuronal loading control.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

References (4560)

References and Notes

  1. For experimental procedures, see the supplementary materials and methods on Science Online.
Acknowledgments: We thank Y. Zhang, S. Maxeiner, and S. J. Lee for advice, G. Feng (MIT) for reagents, and V. Sebastiano (Stanford) for sharing instruments. This work was supported by grants from NIH (MH092931 to M.W.; NS077906 to T.C.S.; and U19MH104172 to M.W. and T.C.S.) and a postdoctoral fellowship from Vetenskapsradet, Sweden (to S.C.B.). The data are included in the main manuscript and in the supplementary materials. Author contributions: F.Y. performed the ES cell, molecular biology, expression, and cell-biology experiments, T.D. the electrophysiological experiments, S.C.B. the protein chemistry experiments, C.P. the immunoblotting experiments, and C.H.P. the gene-expression experiments. All authors planned the experiments, analyzed data, and edited the paper written by T.C.S.
View Abstract

Stay Connected to Science

Navigate This Article