Dissociation of the mTOR Protein Interaction Network Following Neuronal Activation Is Altered by Shank3 Mutation

Abstract

The mechanistic target of Rapamycin (mTOR) kinase pathway plays critical roles in neuronal function and synaptic plasticity, and its dysfunction is implicated in numerous neurological and psychiatric disorders. Traditional linear models depict mTOR signaling as a sequential phosphorylation cascade, but accumulating evidence supports a model that includes signaling through dynamic protein-protein interaction networks. To examine how neuronal mTOR signaling networks discriminate between distinct stimuli, we quantified phosphorylation events and protein co-association networks in primary mouse cortical neurons. Unexpectedly, neuronal mTOR activation by IGF or glutamate triggered dissociation-rather than the anticipated assembly-of protein complexes involving mTOR complex 1 (TORC1), mTOR complex 2 (TORC2), and translational machinery, distinguishing neurons from proliferative cells. Applying in vitro homeostatic scaling paradigms revealed distinct combinatorial encoding of synaptic scaling direction: both up- and down-scaling induced dissociation of translational complexes, but downscaling uniquely included dissociation of upstream pathway regulators. Cortical neurons from Shank3B knockout mice, modeling autism-associated Phelan-McDermid Syndrome, displayed baseline hyperactivation of the mTOR network, which reduced the dynamic range of protein interaction network responses to homeostatic synaptic scaling and pharmacological mTOR inhibition. These findings reveal that neuronal mTOR signaling employs stimulus-specific combinations of dissociative protein interaction modules to encode opposing forms of synaptic plasticity.

Keywords: Shank3; homeostatic scaling; mTOR; protein–protein interaction; signal transduction.

FIGURE 1

Mtor dynamics following IGF or Glutamate stimulation in cortical neuron culture. (A) Experimental design. NBA, neurobasal media; DIV, days in vitro. (B) Representative western blots showing phospho‐ and total AKT, S6, ERK and puromycin labelig, with Actin for a loading control. Full blot images can be found in Figure S1 (C–F) Quantification of blots shown in B, N = 4 independent treatments and lysate preparations per condition. Error bars represent standard error of the mean (SEM). * indicates p < 0.05 from starved by analysis of variance (ANOVA) followed by Dunnett post hoc testing. (G) Principal component graph of quantitative multiplex immunoprecipitation (QMI) data, N = 4 independent treatments and lysate preparations per condition. (H) Topological overlap matrix of QMI data. Ends of the symmetrical dendrogram indicate interactions (unlabeled). Dark pixels indicate stronger correlation between interactions on the x and y axes. Colored boxes indicate modules. (I) Module‐trait table showing the correlation coefficient (top number) and p‐value (bottom number) between the eigenvector of each color‐coded module (colored rectangles on the left) and binary‐coded variables as shown in the table below. (J) Heatmap of scaled values of all significantly changed interactions. Each box represents a single interaction measurement from a single biological replicate; columns correspond to a biological replicate while rows correspond to an interaction (listed IP_Probe). Statistical significance calculated by adaptive, non‐parametric, paired statistical test corrected for multiple comparisons (ANC) and correlation network analysis (CNA) statistics as detailed in methods. (K–N) Mean scaled value of all interactions in the turquoise (K) or blue (M) module, and a representative interaction from each module (L,N). Error bars represent SEM. * indicates p < 0.05 by ANOVA followed by Dunnett post hoc testing (K, M) or by ANC (L, N).

FIGURE 2

Mtor dynamics following IGF and Glutamate treatment of Shank3 KO neurons. (A) Representative western blots showing phospho‐ and total AKT with Actin for a loading control. Full blot images can be found in Figure S4 (B) Quantification of blots shown in A. ** indicates p < 0.01 by analysis of variance (ANOVA) followed by Dunnett post hoc testing, N = 6 independent treatments and lysate preparations per condition. Error bars represent standard error of the mean (SEM). (C) Heatmap of all significantly altered interactions as in Figure 1. (D) Mean scaled value of all interactions in the green module. (E) Median fluorescence intensity (MFI) of the mTOR_Raptor co‐association as measured by quantitative multiplex immunoprecipitation (QMI). (F) Mean scaled value of all interactions in the red module. For (D–F), error bars represent SEM, * indicates p < 0.05 by 2‐way ANOVA followed by Tukey post hoc testing (D, F) or adaptive, non‐parametric, paired statistical test corrected for multiple comparisons (ANC) (E).

FIGURE 3

mTOR pathway phosphorylation following homeostatic scaling. (A) Representative western blots showing phospho‐ and total AKT, S6, 4EPB1, mTOR and p70S6K following 12 or 48 h of up‐ or down‐scaling with tetrodotoxin (TTX) or bicuculline (BIC). Full blot images can be found in Figure S5 (B–G) Quantification of blots shown in A. (H) Representative western blot showing puromycin labeling following homeostatic scaling, with actin as a loading control. (I) Quantification of puromycin incorporation. Error bars represent standard error of the mean (SEM), * indicates p < 0.05, ** p < 0.01, **** p < 0.001 compared to DMSO by 2‐way analysis of variance (ANOVA) followed by Dunnett post hoc testing. N = 4 independent treatments and lysate preparations per condition.

FIGURE 4

mTOR network dynamics following homeostatic scaling in cultured neurons. (A) Module‐trait table showing the correlation coefficient (top number) and p‐value (bottom number) between the eigenvector of each color‐coded module (colored rectangles on the left) and binary‐coded trait labels shown in the table below. (B, C) Average scaled value of all interactions in the turquoise (B) and blue (C) modules. Error bars represent standard error of the mean (SEM), * indicates p < 0.05 by 2‐way analysis of variance (ANOVA) followed by Dunnett post hoc testing. (D) Heatmap of the scaled values of all significantly altered interactions, N = 8 independent treatments and lysate preparations per condition. Statistical significance calculated by adaptive, non‐parametric, paired statistical test corrected for multiple comparisons (ANC) and correlation network analysis (CNA) statistics. (E, F) Node‐edge diagrams of all interactions in the turquoise (E) and blue (F) modules. Nodes represent proteins measured by the quantitative multiplex immunoprecipitation (QMI) panel, edges represent interactions significant by both ANC and CNA statistics.

FIGURE 5

Homeostatic scaling in Shank3^−/− neurons. (A) Western blot showing phospho‐ and total S6 following 48 h of up‐ or down‐scaling. Full blot images can be found in Figure S8 (B) Module‐trait table showing the correlation coefficient (top number) and p‐value (bottom number) between the eigenvector of each color‐coded module (colored rectangles on the left) and binary‐coded trait labels shown in the table below. (C) Heatmap of the scaled values of all significantly altered interactions, by adaptive, non‐parametric, paired statistical test corrected for multiple comparisons (ANC) and correlation network analysis (CNA), N = 7 independent treatments and lysate preparations per condition. (D–J) Mean scaled value of all interactions in each module (D, G, I), and representative interactions from each module (E, F, H, J). Error bars represent SEM, * indicates p < 0.05 compared to same genotype (treatment effect), # indicates p < 0.05 compared to same treatment condition (genotype effect), by 2‐way analysis of variance (ANOVA) followed by Dunnett post hoc testing (D, G, I) or ANC (F, E, H, J).

FIGURE 6

Inhibition of mTOR signaling with Rapalink in WT and Shank3B^−/− neurons. (A) Representative western blots showing phospho‐ and total Akt, phospho‐ and total S6, and beta‐Actin in wildtype (WT) and Shank3B^−/− neurons with and without Rapalink treatment. (B, C) Quantification of blots shown in A. Error bars represent standard error of the mean (SEM), * indicates p < 0.05 by 2‐way ANOVA followed by Dunnett post hoc testing; N = 4 independent treatments and lysate preparations per condition. (D) Heatmap showing the scaled values of significantly different interactions identified by both ANC and CNA; N = 4 independent treatments and lysate preparations per condition. (E–J) Mean scaled value of all interactions in each module (E, G, I), and representative interactions from each module (F, H, J). Error bars represent SEM, * indicates p < 0.05, *** p < 0.001 compared to DMSO control within genotype (treatment effect), and # indicates p < 0.05 compared to wildtype with same treatment (genotype effect) by 2‐way ANOVA followed by Dunnett post hoc testing (E, G, I) or ANC (F, H, J).