Identification of increased S6 phosphorylation in lymphoblastoid cell lines from individuals with autism
Dysregulated PI3K/mTOR signaling in the brain has been detected and successfully targeted to correct phenotypes in several mouse models of autism, including FXS [1, 27–31]. Defects in this signaling pathway might thus be a shared, targetable pathological mechanism in autism disorders of diverse etiologies. We and others have previously shown that altered PI3K/mTOR signaling, which contributes to neuronal dysfunction and autistic-like phenotypes, can be detected in peripheral cells from individuals with FXS, such as lymphoblastoid cell lines and fibroblasts [24, 32, 33]. To assess if abnormal PI3K/mTOR-mediated signaling as a shared molecular defect in autism is detectable in peripheral cell lines from humans with idiopathic autism, we analyzed lymphoblastoid cell lines from the Autism Genetic Research Exchange (AGRE) collection and the Simons Simplex Collection (SSCs). AGRE collects lymphocytes from autistic individuals from simplex and multiplex families with none, one, or several additional siblings diagnosed with autism, whereas the SSC contains lymphoblastoid cell lines from families with one autistic child that has at least one healthy sibling. S6 phosphorylation and S6 expression from 21 individuals with autism from AGRE were compared to an unaffected control using ELISA analyses (Fig. 1). Three out of 21 individuals showed significantly increased S6 phosphorylation compared to an unaffected control (Fig. 1a, one-way ANOVA F(21,85) = 3.5; p < 0.0001, Dunnett’s post hoc comparisons of autism cell lines to the unaffected control *p < 0.05). Two of these three also had significantly increased S6 expression, suggesting that increased phosphorylation was due to an overall increase in S6 (Fig. 1b, one-way ANOVA F(21,85) = 5.47; p < 0.0001, Dunnett’s post hoc comparisons of all autism cell lines to the unaffected control *p < 0.05). No differences in total S6 levels between any other cell line and the control cell line were detected. Consequently, one cell line (A4) had significantly increased pS6/S6 ratios suggesting a defect in the upstream signaling pathway (Fig. 1c, one-way ANOVA F(21,85) = 2.34; p = 0.003, Dunnett’s post hoc comparisons of all autism cell lines to the unaffected control *p = 0.013). A list of p values for all pairwise comparisons can be found in Additional file 2: Table S1a. Independent repeats of assays in the unaffected control used in Fig. 1 as well as experiments using an additional unaffected control line (CTR-2) supported the reproducibility of the assay results and showed that there was no significant difference in S6 phosphorylation in the two control lines (Additional file 1: Figure S1a–c). Nonetheless, differences in S6 phosphorylation could be due to autism-unrelated inherited genetic predispositions, and a lymphoblastoid cell line generated from a random, unrelated healthy individual might thus not be the ideal control. To address this issue, we followed a second approach, in which we took advantage of the SSC, a collection of lymphoblastoid cell lines from autism simplex families, in which for every individual with autism, at least one healthy sibling is available. Here, we directly compared every cell line from affected individuals with the cell line from the respective unaffected sibling. In 2 out of 37 proband-sibling pairs (S1 and S11), there was a consistent increase in the ratio of phospho-S6 to S6 in the proband compared to his healthy sibling in four repeats, with p < 0.05 for one pair (S1) after false discovery rate correction (Fig. 2, two-way-ANOVA, family and disease status as fixed factors; list of p values for all pairwise comparisons in Additional file 2: Table S1b). To further investigate whether differences in S6 phosphorylation could be due to autism-independent genetic predispositions in unrelated individuals, we compared phospho-S6/S6 ratios in the unaffected siblings. A one-way ANOVA showed a significant difference (F(36,109) = 1.7, p = 0.018), which was mainly driven by sibling S37 (see Additional file 1: Figure S1d, and associated figure legend). Of note, this was an effect of the family S37, as the phospho-S6/S6 ratio of the proband S37 was very low, too, and there was no difference between sibling S37 and the proband S37 (FDR-adjusted p = 0.919; Fig. 2 and Additional file 2: Table S1b). These results show that most unaffected controls have similar phospho-S6/S6 ratios (CTR and CTR-2, healthy siblings S1-S36). This data further supports the validity of our approach to directly compare autistic individuals to their healthy family members to account for effects of autism-unrelated genetic predispositions or environmental influences.
Elevated expression and activity of the PI3K catalytic subunit p110δ in a cell line with increased S6 phosphorylation
We next assessed whether lymphoblastoid cell lines provide a suitable tool to identify molecular mechanisms leading to increased cell signaling. First, we performed Western blot analyses and confirmed that, similar as observed in the ELISA assays, S6 phosphorylation of cell line A4 (AGRE) was increased compared to cell line A21 and the unaffected control (Fig. 3a, b, one-way ANOVA F(2,9) = 35.36, p < 0.0001; Tukey’s post hoc analyses *p = 0.0004; #
p < 0.0001). We have previously shown that increased expression of the PI3K catalytic subunit p110β contributes to increased S6 phosphorylation in lymphoblastoid cell lines from patients with FXS [24]. We thus hypothesized that increased phospho-S6/S6 ratios in patient cell line A4 may be caused by increased expression and activity of a PI3K catalytic subunit. In contrast to FXS, the PI3K catalytic subunit p110β was not changed (Fig. 3a, c, one-way ANOVA F(2,9) = 0.66, p = 0.54). However, Western blot analyses revealed that cell line A4 had increased expression of the class I PI3K subunit p110δ compared to cell line A21 and the unaffected control (Fig. 3a, d, one-way ANOVA F(2,9) = 10.42, p = 0.0045; Tukey’s post hoc analyses *p = 0.0144; #
p = 0.0056). In line with the increased expression of p110δ, p110δ-associated PI3K activity was also elevated in cell line A4 compared to cell line A21 and the unaffected control (Fig. 3e, one-way ANOVA F(2,21) = 14.73, p = 0.0001; Tukey’s post hoc analyses *p = 0.0002, #
p = 0.0006). We further tested if similar defects were detectable in cell lines S1 and S11 (SSC; see Fig. 2). Western blot analyses showed increased S6 phosphorylation in S1 and S11 but not in S3 (Fig. 4, paired t tests, (a) n = 3, t(2) = 10.7, p = 0.009; (b) n = 4, t(3) = 10.39, p = 0.002; (c) n = 4, t(3) = 0.72, p = 0.525), thereby confirming our ELISA results (Fig. 2). In contrast to cell line A4, neither S1 nor S11 showed consistent and significant differences in the expression of the class IA PI3K subunits p110α, p110β, or p110δ (Additional file 3: Figure S2), suggesting that other upstream or downstream defects in the PI3K/mTOR pathway may cause altered S6 phosphorylation. Alternatively, mutations within PI3K catalytic subunits may change their activity but not expression and lead to increased S6 phosphorylation.
Increased p110δ expression may be associated with autism diagnosis in a multiplex family
We next took advantage of the fact that the AGRE collection stores lymphoblastoid cell lines and data from entire multiplex families. Cell line A4 was derived from an individual that has two unaffected parents (A4-F, A4-M), one unaffected sister (A4-S), and three brothers with autism (A4-B1, A4-B2, A4-B3), one of which is his twin brother (A4-B1) (Fig. 5a). Western blot analyses of the parents, the sister, and two brothers showed significant differences in p110δ expression among family members (Fig. 5b, one-way ANOVA, F(5,30) = 3.14, p = 0.021). Post hoc analyses comparing all family members to the healthy sister suggested that p110δ expression in the three affected sons was increased compared to the unaffected parents and sister, although statistical significance was not reached (p values for post hoc comparison to the unaffected sister shown in Additional file 4: Table S2a). Similarly, phospho-S6- and S6-specific ELISA analyses showed that phospho-S6/S6 ratios are significantly different among family members, and post hoc analyses suggested that phospho-S6/S6 ratios may be increased in the three family members with autism (Fig. 5c, one-way ANOVA, F(5,24) = 3.77, p = 0.012; p values for post hoc comparisons to the unaffected sister shown in Additional file 4: Table S2b). To further assess if increased p110δ expression and increased phosphorylation of S6 are associated with autism in this specific multiplex family, we compared pooled data from the autistic family members with those from unaffected family members. Analysis of pooled expression data from the three unaffected and the three affected lymphoblastoid cell lines showed that overall p110δ expression and S6 phosphorylation were significantly increased in lymphoblastoid cells from individuals with autism compared to unaffected individuals in this specific family (Fig. 5d, e, paired t tests, (d) t(5) = 4.99, p = 0.004; (e) t(4) = 4.81, p = 0.009). However, we cannot exclude that the difference observed in the pooled data, in particular in pS6/S6 ratios, was mainly driven by cell line A-4, which was originally identified in our screen. The cell line of one of the affected brothers (A4-B3) did not grow well and therefore was excluded from the analysis.
Elevated S6 phosphorylation and protein synthesis rates in cell line A4 are reduced with a p110δ-selective inhibitor
Increased PI3K activity and S6 phosphorylation contribute to increased protein synthesis in FXS mouse models and cell lines from individuals with FXS [1, 24, 26–28, 33]. To assess whether elevated p110δ expression and activity in cell line A4 likewise leads to increased basal protein synthesis, we quantified protein synthesis rates in cell line A4 and the unaffected sister cell line A4-S using puromycin-labeling and anti-puromycin-specific Western blotting [25]. We detected increased protein synthesis rates in cell line A4, which were reduced to control levels by the p110δ-selective inhibitor IC87114 [34] (Fig. 6a, two-way ANOVA F
interaction(1,6) = 10.73, p
interaction = 0.017; F
treatment(1,6) = 9.34, p
treatment = 0.022; F
autism(1,6) = 1.20, p
autism = 0.316; Sidak’s post hoc analyses *p = 0.010, #
p = 0.011). Protein synthesis rates in the cell line from the unaffected sibling were not affected by the p110δ inhibitor (p = 0.907). The specificity of the effect of the p110δ-inhibitor on protein synthesis rates was further confirmed by experiments showing that a selective inhibitor of the PI3K catalytic subunit p110β, TGX-221, did not significantly affect protein synthesis in cell line A4 (Fig. 6b, n = 4, paired t test, t(3) = 0.34, p = 0.76). In contrast, the p110δ-selective inhibitor reduced S6 phosphorylation in both cell lines, suggesting that additional p110δ-dependent mechanisms, apart from increased S6 phosphorylation, lead to increased protein synthesis rates in the cell line A4 (Fig. 6c, two-way ANOVA with Sidak’s post hoc tests F
autism(1,3) = 35.01, p
interaction = 0.001; F
treatment(1,3) = 163.2, p
treatment = 0.001; F
autism(1,3) = 4.15, p
autism = 0.135; *p = 0.003, #
p = 0.0005; $
p = 0.0019). Of note, the p110β-selective inhibitor also reduced S6 phosphorylation in cell line A4 (Fig. 6d, n = 5, paired t test, t(4) = 5.75, p = 0.005), similarly as we have observed previously in lymphoblastoid cell lines from a patient with FXS and a healthy control [24].