CLK2 inhibition ameliorates autistic features associated with SHANK3 deficiency

See allHide authors and affiliations

Science  11 Mar 2016:
Vol. 351, Issue 6278, pp. 1199-1203
DOI: 10.1126/science.aad5487

Signal problems in autism spectrum disorder

Autism spectrum disorders have many causes. Bidinosti et al. studied Phelan-McDermid syndrome (PMDS), one of the symptoms of which can be autism (see the Perspective by Burbach). The authors used neurons derived from these patients, as well as from mice, with the culprit gene disrupted and found that a chain of intracellular signals becomes imbalanced. Signaling and behavioral symptoms could be improved by a small-molecule therapeutic that inhibits a key kinase.

Science, this issue p. 1199; see also p. 1153


SH3 and multiple ankyrin repeat domains 3 (SHANK3) haploinsufficiency is causative for the neurological features of Phelan-McDermid syndrome (PMDS), including a high risk of autism spectrum disorder (ASD). We used unbiased, quantitative proteomics to identify changes in the phosphoproteome of Shank3-deficient neurons. Down-regulation of protein kinase B (PKB/Akt)–mammalian target of rapamycin complex 1 (mTORC1) signaling resulted from enhanced phosphorylation and activation of serine/threonine protein phosphatase 2A (PP2A) regulatory subunit, B56β, due to increased steady-state levels of its kinase, Cdc2-like kinase 2 (CLK2). Pharmacological and genetic activation of Akt or inhibition of CLK2 relieved synaptic deficits in Shank3-deficient and PMDS patient–derived neurons. CLK2 inhibition also restored normal sociability in a Shank3-deficient mouse model. Our study thereby provides a novel mechanistic and potentially therapeutic understanding of deregulated signaling downstream of Shank3 deficiency.

Chromosomal aberrations at 22q13 that delete or inactivate one SH3 and multiple ankyrin repeat domains 3 (SHANK3) allele are genetic hallmarks of Phelan-McDermid syndrome (PMDS). De novo mutations in SHANK3 are also associated with nonsyndromic autism spectrum disorder (ASD) and intellectual disability (14). Genetic ablation of Shank3 in mice yields ASD-like behavioral phenotypes and synaptic dysfunction (510), the latter of which has also been observed in PMDS patient–derived neurons or after Shank3 knockdown in primary rodent neurons (11, 12). Shank3 is a large protein that serves as a scaffold to organize excitatory postsynaptic densities through protein-protein interactions [reviewed in (13)]. Reduced expression of Shank3 results in decreased dendritic spine density (6, 7, 11, 14), whereas its overexpression enhances spine number or induces spine formation in aspiny neurons (14, 15).

PMDS patients treated with insulin-like growth factor–1 (IGF-1) showed improvements in social and stereotyped behaviors (16). IGF-1 also alleviated deficits in Shank3-deficient mice (17) and PMDS neurons (12). Here, we used phosphoproteomics to investigate the signaling pathways that link Shank3 and IGF-1.

We used rat cortical neurons and reduced Shank3 expression by knockdown. We treated these neurons with brain-derived neurotrophic factor (BDNF) to elicit synaptic plasticity-dependent changes and then analyzed the phosphoproteome (fig. S1). Several phosphopeptides originating from protein substrates in the protein kinase B (PKB or Akt)–mammalian target of rapamycin complex 1 (mTORC1) pathway showed reduced phosphorylation levels in Shank3 knockdown neurons (Fig. 1A, table S1, and database S1). In particular, phosphorylation on GSK3β and ribosomal protein S6 (rpS6) was less than normal on Akt and mTORC1-dependent sites, respectively (Fig. 1A and table S1). Phosphorylation on B56β (Ppp2r5b), a PP2A phosphatase regulatory subunit (18), was elevated on a peptide sequence known (19) to regulate dephosphorylation of Akt (Fig. 1, A and B). No significant changes were observed in extracellular–signal regulated kinase (ERK) signaling (Mapk1/3) (Fig. 1A and table S1). Consistent with this, we also observed reduced pathway activity in synaptosomal fractions from a newly produced Shank3-deficient mouse model (table S2, database S2, and fig. S9). In a second method of proteomic analysis, we enriched less mTOR from Shank3 knockdown neurons on the basis of affinity for pan-kinase inhibitor conjugated beads (Kinomatrix beads) (20) (fig. S1 and database S3). The Akt-mTORC1 signaling pathway is linked to ASD (21, 22) and is regulated by IGF-1, which is currently being explored as a therapeutic for PMDS (12, 16, 17). We therefore explored whether Shank3 deficiency in PMDS leads to reduced Akt-mTORC1 signaling by way of enhanced PP2A-mediated inactivation of Akt (fig. S1).

Fig. 1 Shank3 deficiency impairs Akt-mTORC1 signaling.

(A) Relative quantitation of phosphopeptide abundance between Shank3 knockdown and control neuron samples. Down-regulation of phosphopeptides from several proteins targeted by Akt-mTORC1 signaling were detected, as was up-regulation of a phosphopeptide from PP2A regulatory subunit B56β. Nonsignificant changes of Mapk1/Mapk3 are shown in the center. (B) Representative tandem mass spectrometry spectrum of B56β phosphopeptide SHS*SSQFR identified with increased abundance (Log2fold change = 2.489) in Shank3 knockdown neurons. Asterisk denotes phosphorylation at Ser46. (C) Western blotting validation of reduced Akt-mTORC1 signaling in Shank3 knockdown neurons. Neurons were treated with BDNF (50 ng/ml), and cell lysates were prepared in radioimmunoprecipitation assay buffer. Plots are means ± SEM. (D) Impaired Akt activity in PMDS iPSC-derived patient neurons. Cell lysates were prepared as in (C) at 8 weeks in vitro. Plots are means ± SEM. (E) Impaired Akt activation after BDNF stimulation in PMDS iPSC-derived patient neurons. Cell lysates were prepared as in (C) at 8 weeks in vitro.

Shank3 knockdown in primary neurons caused a reduction by a factor of two of Akt phosphorylation at T308 in basal conditions or after BDNF stimulation (Fig. 1C) but had less of an effect at S473. Coincident phosphorylation of both Akt sites (T308 and S473) is required for full Akt activation. Phosphorylation of rpS6 was similarly reduced (Fig. 1C). Attenuated Akt T308 phosphorylation was observed with two additional Shank3 short hairpin RNAs (shRNAs) and was rescued by coexpression of nontargeted Shank3 (fig. S2). mTOR steady-state levels were unchanged. This indicated that reduced mTOR recovery in Kinomatrix capture, which enriches activated kinases through affinity for a promiscuous adenosine triphosphate (ATP)–competitive inhibitor resin, was due to impaired kinase activity after Shank3 knockdown. BDNF and its receptor TrkB also activate ERK and phospholipase Cγ (PLCγ) pathways (fig. S2); however, no impairment was observed in either of these pathways (Fig. 1C and fig. S2). We observed no change in Akt phosphorylation after knockdown of another postsynaptic density protein, PSD95, thus indicating that impaired Akt signaling is specific to Shank3 deficiency rather than generic reduction of synaptic proteins (fig. S2). Phosphorylation of the BDNF receptor, TrkB, and expression of phosphatase and tensin homolog (PTEN) were also unchanged (Fig. 1C), suggesting that attenuated Akt-mTORC1 signaling occurs directly through deregulated Akt phosphorylation. These findings were corroborated in human induced pluripotent stem cell (iPSC)–derived neurons from two PMDS patients who harbor deletions within the SHANK3 locus (fig. S3). As with Shank3 knockdown, Akt T308 phosphorylation was reduced in these SHANK3-deficient neurons (Fig. 1, D and E).

Next, we examined the possibility that Akt-signaling is attenuated by elevated PP2A dephosphorylation activity. B56β, a regulatory subunit of PP2A, is expressed in the brain and defines PP2A specificity for Akt T308 dephosphorylation (fig. S1) (18, 23). Phosphorylated B56β recruits PP2A catalytic (PP2Ac) and structural subunits to Akt for heterotrimeric phosphatase holoenzyme assembly and T308 dephosphorylation (19). In Shank3 knockdown neurons, association of Akt with PP2Ac was above normal by a factor of two (Fig. 2A). Thus, excess of B56β phosphorylation enhances recruitment of PP2A to Akt and reduces phosphorylation of Akt within the catalytic domain at T308, thereby attenuating activity. Addition of the PP2A inhibitor Okadaic acid (OA), alone or together with BDNF, increased Akt T308 phosphorylation in Shank3 knockdown neurons by a factor of two, relative to control (Fig. 2B). Expression of a B56β variant lacking the regulatory phosphorylation sites (B56β 6A) (19) restored Akt phosphorylation in Shank3 knockdown neurons to control levels (Fig. 2C; compare lanes 1 and 2 with 11 and 12). Thus, Shank3 deficiency in neurons engenders enhanced activity of B56β/PP2A, leading in turn to excessively dephosphorylated Akt and deficiency in Akt activity.

Fig. 2 Elevated B56β/PP2A activity and CLK2 expression in Shank3-deficient neurons.

(A) Association of PP2A catalytic subunit (PP2Ac) with Akt is enhanced in Shank3 knockdown neurons. Primary neurons cotransduced with shRNA and HA-Akt lentiviruses were lysed and immunoprecipitated with antibody to hemagglutinin(HA) tag. (B) PP2A-inhibition restores Akt activity in Shank3 knockdown neurons. Neurons were treated with 100 nM Okadaic acid 15 min before and during BDNF (50 ng/ml) incubation, followed by lysis and Western blotting. Data are means ± SEM. (C) Overexpression of a phosphorylation-defective variant of B56β restores Akt activity in Shank3 knockdown neurons. Flag-tagged B56β, or a variant lacking phosphoserines on the indicated sites (B56β 6A) (19), were cotransduced with shRNA viruses. Neurons were treated with BDNF, followed by Western blotting. (D) Increased steady-state levels of CLK2 in Shank3 knockdown neurons. Neurons were transduced with lentiviruses followed by Western blotting. Data are means ± SEM. (E) Attenuated CLK2 ubiquitination in Shank3 knockdown neurons. Neurons were cotransduced with shRNA and Myc-CLK2 lentiviruses. Cell lysates were prepared in immunoprecipitation buffer and immunoprecipitated with antibody to Myctag followed by Western blotting for the proteasome-targeting Ubiquitin K48-linkage. (F) CLK2 inhibition restores Akt activity in Shank3-deficient neurons. Neurons were treated with 20 μM TG003 for 60 min before Western blotting.

B56β is phosphorylated by CLK2 (19) on serine residues in the regulatory peptide sequence. Shank3 knockdown neurons exhibited an increase by a factor of two in CLK2 protein levels (Fig. 2D), consistent with excess phosphorylation on B56β (Fig. 1, A and B). In hepatocytes, CLK2 is up-regulated by insulin induction to homeostatically regulate Akt through B56β/PP2A-mediated dephosphorylation (19, 24). BDNF treatment induced the accumulation of CLK2 in control neurons but not in Shank3 knockdown neurons (fig. S4). This suggests that the regulated expression of CLK2 by ubiquitination (24) is lost in the absence of Shank3. Proteasome inhibition caused a rapid increase of CLK2 in control neurons but not in neurons deficient in Shank3 (fig. S4). Immunoprecipitation of Myc-CLK2 showed less CLK2 ubiquitination in Shank3 knockdown neurons (Fig. 2E). Shank3 knockdown neurons exhibited no change in CLK2 mRNA levels (fig. S5). Treatment with an ATP-competitive CLK2 inhibitor, TG003 (fig. S4), restored Akt T308 phosphorylation in Shank3 knockdown neurons to control levels (Fig. 2F). Thus, increased steady-state levels of CLK2 in Shank3-deficient neurons occurs through impaired ubiquitination and causes excess Akt inactivation through elevated B56β/PP2A activity (fig. S4).

We anticipated that Akt activation, either directly or through CLK2 inhibition, would reverse neuronal impairments associated with Shank3 deficiency. Shank3 knockdown in organotypic brain slice cultures decreased dendritic spine density and was rescued by Shank3 re-expression (fig. S6) (4, 6, 11). We then exposed slices to the small-molecule Akt activator SC79 (fig. S7) (25) or to TG003. We observed that both treatments restored spine density in Shank3 knockdown neurons to control levels (Fig. 3, A and B). This was dependent on Akt activity because it was blocked by co-incubation with a validated Akt inhibitor (Fig. 3B and fig. S7). Reduction of CLK2 in Shank3 knockdown neurons had a similar restorative effect on spine density to that of TG003 treatment (fig. S6). We confirmed that Akt inhibition in wild-type neurons is sufficient to reduce spine density and thereby phenocopy the effect of Shank3 deficiency on reducing spine density through downstream Akt attenuation (fig. S6). Shank3 knockdown in organotypic slices reduced miniature excitatory postsynaptic current (mEPSC) frequency that was restored to control levels with SC79 treatment (Fig. 3C). iPSC-derived neurons from two unrelated PMDS patients exhibited reduced frequency of spontaneous EPSCs (sEPSCs) (Fig. 3D and fig. S8). Treatment with SC79 or TG003 again rescued these deficits in an Akt-dependent manner (Fig. 3D, highlighted groups). Thus, Akt activation alone or through CLK2 inhibition is sufficient to restore synaptic impairments occurring from Shank3 deficiency.

Fig. 3 Activation of Akt or inhibition of CLK2 corrects Shank3-associated deficits.

(A) SC79 rescues spine density in Shank3 knockdown neurons. Neurons were treated on day 14 in vitro for 24 hours with 4 μg/ml SC79 before immunostaining for green fluorescent protein. Plots indicate group means ± SEM [one-way analysis of variance (ANOVA), P < 0.0001, Dunn’s multiple comparisons test]. (B) CLK2 inhibition restores normal spine density in Shank3 knockdown neurons. Organotypic slices were treated for 24 hours with 20 μM TG003 and 10 μM Akt inhibitor (Akti), as indicated (one-way ANOVA, P < 0.0001, Dunn’s multiple comparisons test). Scale bar is 2 μm. (C) SC79 rescues impaired synaptic transmission frequency in Shank3 knockdown neurons. (Upper left) Representative traces of mEPSCs recorded from CA1 neurons in organotypic slice cultures prepared and treated as in (A) and fig. S6 (scale bar: 5 pA, 1 s). (Upper right) Average traces of the mEPSCs amplitude for the cells shown (scale bar: 1.6 pA, 5 ms). (Lower) Plots of recorded mEPSCs. Data are means ± SEM (one-way ANOVA, P < 0.005, Tukey’s multiple comparisons test). (D) SC79 and TG003 correct synaptic transmission impairments in PMDS neurons. (Upper) Representative traces of recorded sEPSCs (scale bar: 20 pA, 5 s). (Lower) Plots of 24 hours treatments on sEPSC frequency and amplitude of PMDS patient-derived neurons at eight weeks in vitro. Data are means ± SEM (one-way ANOVA, P < 0.0001, Dunn’s multiple comparisons test).

To determine whether our findings extend to ASD-like behaviors, we generated a Shank3-deficient mouse model by ablation of Shank3 exon 21 (fig. S9) as previously described (8, 10). The major Shank3 isoforms are absent in our homozygous mice (Shank3ΔC/ΔC), whereas faster-migrating, truncated fragments were detected (fig. S9). Shank3ΔC/ΔC neurons displayed excess CLK2 expression (fig. S9). Neither heterozygous (Shank3+/ΔC) nor Shank3ΔC/ΔC mice exhibited anxious behavior or locomotor skill impairments (fig. S10). Both Shank3+/ΔC and Shank3ΔC/ΔC mice displayed avoidance behavior, assessed by marble burying, that was refractory to treatment with TG003 (fig. S10). In contrast, we observed that only Shank3ΔC/ΔC mice exhibited excess self-grooming, a trait reflecting repetitive behaviors seen in ASD (fig. S10). Treatment of these mice with TG003 significantly decreased self-grooming, although not to wild-type frequency (fig. S10). Mice were then tested in a social motivation paradigm (fig. S10). Wild-type and Shank3+/ΔC mice displayed a significant preference for social investigation, whereas Shank3ΔC/ΔC mice did not (Fig. 4, A and B, and fig. S10). In contrast, Shank3ΔC/ΔC mice treated with TG003 recovered normal preference for social interaction (Fig. 4, A and B, and fig. S10). We observed that this effect persisted 3 days after treatment when a new cohort of animals was tested (fig. S10). The recovery of normal social behavior correlated with restored Akt phosphorylation in synaptosomal fractions (fig. S9). No significant preference was observed between groups when a second intruder was introduced (Fig. 4B). Thus, CLK2 inhibition rescues deficits in social behavior caused by Shank3 deficiency.

Fig. 4 CLK2 inhibition corrects impaired social motivation in Shank3ΔC/ΔC mice.

(A) Shank3ΔC/ΔC mice display impaired motivation for social interaction that is corrected by treatment with CLK2-inhibitor TG003. Interaction times with the intruder mouse (S1) or the object (O1) are plotted for phase 2. Data are means ± SEM with paired t tests (wild-type, P < 0.0005; Shank3ΔC/ΔC +TG003, P < 0.0005), comparing S1 to O1 investigation times within each group. Comparison of social interaction times across groups was by one-way ANOVA with Tukey’s multiple comparisons test (P < 0.0005 for differences among group means). (B) Preference index for S1 versus O1 of interaction times (see the methods) was calculated for each test phase. Data are means ± SEM (one-way ANOVA, P < 0.0001, Tukey’s multiple comparisons test).

Mutations in Akt-mTORC1 pathway regulators PTEN and TSC1/2 yield clinical syndromes that include features of ASD by means of exaggerated Akt-mTORC1 activity (21, 22). However, attenuated Akt-mTORC1 activity has also been associated with Angelman syndrome, another monogenic form of ASD (26, 27). This suggests a bidirectional regulation of the Akt-mTORC1 pathway associated with different ASD genes. Here, we show that Shank3 deficiency leads to enhanced CLK2 expression and attenuated Akt-mTORC1 activity. Cellular and behavioral impairments attributed to Shank3 loss of function have been corrected with IGF-1 treatment (12, 16, 17). We found that IGF-1 treatment restored normal dendritic spine density to Shank3 knockdown neurons in an Akt-dependent manner (fig. S11). We suggest that IGF-1 restores balance in signaling pathways by boosting Akt phosphorylation to counteract elevated dephosphorylation by PP2A (fig. S12). Thus, we propose that direct Akt-reactivation or CLK2 inhibition may be therapeutic targets for intervention in patients with PMDS.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

Databases S1 to S3


References and Notes

  1. Acknowledgments: We thank B. Gomez-Mancilla and T. Doll for help in accessing and reprogramming the PMDS cell lines. We thank B. Kinzel, M. Xiaohong, and D. Breustedt for help in the creation of the Shank3ΔC/ΔC mouse model and S. Legare for help with the in vivo experiments. We thank J. Knehr for the technical performance of the metapair sequencing and M. Beibel for the bioinformatics support. Sequencing data related to this study are available in the National Center for Biotechnology Information’s Short Read Archive under accession number SRP067966. The PMDS iPS lines and the Kinomatrix are available from Novartis under materials transfer agreements. Novartis AG may hold patent applications related to aspects of the disclosed subject matter. The Developmental Molecular Pathways Department is part of the Novartis Institutes for Biomedical Research. The supplementary materials contain additional data.
View Abstract

Stay Connected to Science

Navigate This Article