Cell lines and reagents

NPCs derived from one female TSC1 patient donor, including TSC1-het (+/-) parental cell line, as well as CRISPR-deleted TSC1-null (−/−) and CRISPR-corrected TSC1-WT (+/+) lines (one clone/genotype), along with culture conditions, have been previously described [17]. Rapamycin was from EMD Millipore (Burlington, MA), and RMC-6272 (previously known as RM-006) was generously provided by Revolution Medicines, Inc. (Redwood City, CA). All antibodies are listed in Additional file 2: Table S1.

Polysome fractionation and RNA sequencing

Lysate preparation for polysome profiling from three biological replicates was carried out as previously described [48]. Briefly, TSC1 NPC lines were seeded at 40,000 cells/cm2, with a total of 5 × 107 cells seeded per drug treatment condition. The next day after seeding, cells were treated for 2 h with 50 nM rapamycin, 10 nM RMC-6272 or DMSO as a vehicle control. Treated cells were then rinsed with 1X PBS containing 100 µg/ml cycloheximide (CHX) (Sigma, St. Louis, MO), harvested by scraping on ice in PBS/CHX and pelleted by centrifugation at 300g for 10 min at + 4C. Cell pellets were lysed in 5 mM Tris–HCl (pH 7.5), 1.5 mM NaCl, 2.5 mM MgCl2, 0.5% sodium deoxycholate, 0.5% Triton-X-100, 2 mM DTT and 100 μg/ml CHX. Lysates were cleared by spinning for 2 min at 13,000 rpm and quickly frozen on dry ice. When ready for processing, lysates were thawed and loaded onto a 10–50% sucrose gradient, centrifuged for 2 h and 15 min at 35,000 rpm in a SW41 rotor using a Sorvall Discovery 90SE centrifuge. The gradients were fractionated on a Teledyne ISCO Foxy R1 apparatus while monitoring the OD254nm. Fractions corresponding to mRNA associated with more than two ribosomes were pooled and the RNA extracted using TRIzol (Thermo Fisher, Waltham, MA) according to the manufacturer’s protocol. Prior to loading samples on the sucrose gradient, RNA was extracted from 10% of the lysate using TRIzol, and the resulting RNA was denoted as total RNA. RNA sequencing libraries were prepared from the resulting samples using Illumina v2.5 Kits and sequenced (three biological replicates) on an Illumina NextSeq 500 at the Canada's Michael Smith Genome Sciences Centre (BC Cancer Research Institute, Vancouver, Canada).

Polysome fractionation of postmortem brain samples

Polysome fractionation of postmortem samples from the Brodmann area 19 (BA19) brain region from ASD-affected donors (n = 6) and matched controls (n = 4) provided by NIH NeuroBiobank was performed using an optimized sucrose gradient, as previously described [49, 50]. For included donors, there is only incomplete information of comorbidities, medications and severity. Moreover, RNA sequencing data did not support TSC1-mutations in any of the ASD donors. RNA sequencing libraries were generated using the Smart-seq2 protocol as described previously [50]. Single-end 51 base sequencing was performed using the HiSeq2500 platform and the HiSeq Rapid SBS kit v2 chemistry at the National Genomics Infrastructure, Science for Life Laboratory, Stockholm, Sweden. Bcl to fastq conversion was performed using bclfastq_v2.19.1.403 from the CASAVA suite.

RNA-Seq data preprocessing and quality control

Quality of sequencing reads (paired-end for NPC dataset; single-end for postmortem brain dataset) was confirmed using FastQC (v0.11.4; http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). BBmap (v. 36.59: https://www.osti.gov/servlets/purl/1241166; parameters: k = 13, ktrim = n, useshortkmers = t, mink = 5, qtrim = t, trimq = 10, minlength = 25) was used to trim reads for Illumina Truseq and Nextera adapter sequences and low-quality base calls. For the NPC dataset, BBmap (v. 36.59) was used to remove sequencing reads mapping to rRNA sequences obtained from the SILVA ribosomal RNA gene database [51]. The resulting reads were aligned to the human reference genome (build hg38) using HISAT2 (v.2.1.0 and v.2.0.4 for NPC and postmortem brain datasets, respectively, using, in addition to default parameters, “–no-mixed” and “–no-discordant” parameters for the NPC dataset) [52]. The aligned reads were summarized using the “featureCounts” function of the RSubread (v.2.6.4) R/Bioconductor package [52] and the reads were assigned using the hg38 GTF annotation from the UCSC database [53] (parameters: isPairedEnd = isPairedEnd = autosort = T, allowMultiOverlap = F, strandSpecific = 2 for NPC dataset and ignoreDup = FALSE, useMetaFeatures = TRUE, countMultiMappingReads = FALSE for postmortem brain dataset). A summary of the read counts at each preprocessing step was plotted using ggplot2 (v.3.3.6) (Additional file 1: Fig. S1A and Fig. S2A). Expression of TSC1 was further assessed in the NPC dataset using TMM-log2 normalized counts obtained by running calcNormFactors function from edgeR (v.3.34.1) [50] and voom function from limma (v.3.48.3) [52] (Additional file 1: Fig. S1B). Principal component analysis (PCA) on normalized counts [50] was performed using the PCAtools R package (v. 2.4.0; https://github.com/kevinblighe/PCAtools; parameters removeVar = 0.75 and scale = T) and visualized using eigencorplot, screeplot and biplot functions from the PCAtools R package (v.2.4.0; https://ggplot2.tidyverse.org; Additional file 1: Fig. S1C-E and Fig. S2B-D).

Analysis of gene expression alterations using anota2seq

Genes with 0 mapped RNA sequencing read in one or more samples were discarded resulting in analysis of 12,950 and 11,998 genes in postmortem brain samples and NPCs, respectively. The data were TMM-log2 normalized and analyzed using anota2seq [54] [v. 1.14.0; parameters: minSlopeTranslation = − 1, minSlopeBuffering = − 2, maxSlopeTranslation = 2, maxSlopeBuffering = 1, deltaPT = deltaTP = deltaP = deltaT = log2(1.2)] [54, 55]. During analysis of the NPC dataset, replicate was included in the model to correct for batch effects and three contrasts were assessed, namely (1) TSC1−/− versus TSC1+/+, (2) TSC1−/− + rapamycin versus TSC1−/− and (3) TSC1+/+  + rapamycin versus TSC1+/+, with the following thresholds to identify differentially expressed genes: minEff = log2(1.5) and maxRvmPadj = 0.15 [i.e., fold change > log2(1.5) and FDR < 0.15]. In the postmortem dataset, 6 ASD-affected brains were compared to 4 neurotypical, and relaxed thresholds were applied: minEff = log2(1.25) and pVal = 0.05.

Gene ontology analysis

Genes identified as regulated via the “translation” mode in anota2seq (i.e., transcripts with an increase or decrease in polysome-associated mRNA levels without corresponding changes in total cytosolic mRNA levels) were used as input for gene set enrichment analysis using Cytoscape (v.3.8.2.) plug-in ClueGO (v.2.5.8) [56] with FDR cutoff = 0.001 or 0.01 for NPCs and postmortem brain samples, respectively, together with p value cutoff = True, Correction Method Used = Benjamini–Hochberg, Statistical Test Used = Enrichment (Right-sided hypergeometric test), Kappa = 0.4, Min. Percentage = 10, Min GO Level = 7, Max GO Level = 15, Number of Genes = 3, GO Fusion = false, GO Group = true, Over View Term = SmallestPValue, Group By Kapp Statistics = true, Initial Group Size = 1, Sharing Group Percentage = 50.0, Ontology Used = GO_BiologicalProcess-EBI-UniProt-GOA-ACAP ARAP_13.05.2021_00h00, KEGG_13.05.2021, REACTOME_Reactions_13.05.2021, Evidence codes used = All, Identifiers used = SymbolID. The resulting networks were set to show the most significant term identified for each group.

Analysis of gene signatures using empirical cumulative distribution functions

To cross-compare RNA-seq datasets, empirical cumulative distribution functions (ECDFs) of log2 fold changes for polysome-associated and cytosolic mRNA were plotted for genes that were found to be translationally regulated in either dataset. The difference between each tested gene set and the background was quantified at the 50th quantile and the Wilcoxon rank-sum test was used to determine whether there was a significant shift between the background and each signature. The same approach was used to assess signatures of transcripts whose translation was previously identified as increased upon eIF4E overexpression [57] or genes associated with synaptic function [58].

NanoString nCounter Gene Expression AnalysisTarget gene selection and generation of custom NanoString panel

For NanoString nCounter analysis [59], a custom panel of 200 target genes identified by anota2seq analysis as regulated by “translation” or “mRNA abundance” were selected: (1) genes with log2FC > 2 and FDR < 0.15 in any of the contrasts applied when analyzing the NPC dataset, (2) targets annotated to ASD/NDD pathology with log2FC > 1 and FDR < 0.15 in the NPC dataset and (3) negative controls were identified based on standard deviation (< 0.3) between samples, mean TMM-log2 signal in the top 50th quartile, deltaT < 0.1 (from anota2seq analysis) and deltaP < 0.1 (from anota2seq analysis).

NanoString analysis

Samples from each condition were randomized on cartridges and processed by the KIGene Core Facility (Karolinska Institute, Sweden) using 100 ng input for cytosolic mRNA and 300 ng input for polysome-associated mRNA. The “newRccSet” function from the NanoStringQCPro (v.1.24.0) R/Bioconductor package was used to preprocess raw data. Genes with expression less than 6.27 (log2 scale; TSC1−/− vs TSC1+/+ comparison) and 5.63 (log2 scale; TSC1−/− + Rap vs TSC1−/− and TSC1−/− + RMC-6272 vs TSC1−/− comparisons) in 3 or more samples were excluded [thresholds were determined by calculating the mean log2 expression level of negative control genes + 2 standard deviation (SD)]. This resulted in analysis of 175 genes for the TSC1−/− versus TSC1+/+ comparison while 164 transcripts were included in TSC1−/− + Rap versus TSC1−/− and TSC1−/− + RMC-6272 versus TSC1−/− comparisons. For the TSC1−/− versus TSC1+/+ comparison, the geNorm function of the CtrlGene (v.1.0.1) R/Bioconductor package identified BUD31, BASP1, and GAB2 as housekeeping genes for normalization and data were normalized using the contentNorm function (from the NanoStringQCPro package) with the following parameters: method = “housekeeping,” summaryFunction = “mean” and hk = (BUD31, BASP1, GAB2). An additional step of variance stabilizing normalization (vsn) was then performed using the justvsn function from the vsn(v.3.60.0) [60] R/Bioconductor package. For the TSC1−/− + Rap versus TSC1−/− and TSC1−/− + RMC-6272 versus TSC1−/− comparisons, no housekeeping genes could be identified (possibly as the RMC-6272 treatment was not included during selection of housekeeping genes) Therefore, global normalization was performed using the contentNorm function with the following parameters: method = “housekeeping,” summaryFunction = “mean.” Similar to above, vsn normalization was then performed. Log2 fold changes were then calculated and plotted.

Validation of differential translation using RT-qPCR

To validate using RT-qPCR, polysomes from TSC1−/− or TSC1+/+ NPCs (three biological replicates) were fractionated and pooled as described for NPCs above. RNA was isolated using TRIzol (ThermoFisher, Waltham, MA) and cDNA was prepared using M-MuLV Reverse Transcriptase (New England Biolabs, Ipswich, MA) and oligo(dT)20 primers using manufacturer’s recommendations. RT-qPCRs were performed with SsoFast Evagreen Supermix (Bio-Rad, Hercules, CA) using the CFX96 PCR system (Bio-Rad Hercules, CA). Primers for RT-qPCR are detailed in Additional file 3: Table S2. The level of each mRNA was normalized to the geometric mean of β-actin (ACTB), Phosphoglycerate Kinase 1 (PGK1) and Hypoxanthine Phosphoribosyl transferase 1 (HPRT1) using the comparative CT method and compared across conditions as indicated in figure legends.

Cell size and proliferation analysis

For cell size and proliferation assays using trypan blue exclusion methods [61], three biological replicates per treatment (each including three technical replicates) were performed. Briefly, cells were rinsed in PBS and detached using Accutase enzyme detachment medium (ThermoFisher, Waltham, MA). Next, for each sample, 10 µl of suspended cells was combined with 10 µl of trypan blue, and 10 µl of the resulting mixture was added to two replicate chambers of a disposable Countess chamber slide and inserted into the Countess II automated cell counter (ThermoFisher, Waltham, MA), according to the manufacturer’s instructions. Quantitation of viable and dead cell counts as well as size of viable cells in microns, after exclusion of objects consistent with cellular debris, was performed. For each technical replicate, two repeat measurements, including counts of viable versus dead cells as well as size of viable cells, from the two chambers were averaged and recorded. Proliferation was also assessed using flow cytometry based on expression of the proliferation marker Ki-67 as well as real-time, cellular image-based analysis. For Ki-67-based methods, TSC1+/+ and TSC1−/− NPCs were seeded onto Geltrex-coated 150 mm plates at 0.75 × 106 cells/plate (two biological replicates per cell line, per treatment condition). Following overnight attachment, NPCs were treated with 50 nM rapamycin, 10 nM RMC-6272 or DMSO as vehicle control for 72 h. To assess the Ki-67 proliferation status, monoclonal AlexaFluor488-conjugated Ki-67 antibody (Cell Signaling Technologies, Danvers, MA) was employed and immunostaining was performed according to the manufacturer’s instructions. Briefly, NPCs were harvested using Accutase, pelleted, washed with PBS and fixed for 15 min at room temperature with 4% paraformaldehyde. Following fixation, cells were washed twice with PBS and permeabilized with ice cold 100% methanol to a final concentration of 90% with gentle vortexing and incubated for 30 min on ice. Thereafter, cells were stained using Ki-67 at 1:50 for 1 h, and data were acquired using a BD LSR II Flow Cytometer (BD Biosciences). For all treatment conditions, 1 × 104 cells were acquired and recorded. Data analyses were carried out using FlowJo 10.8.1 (FlowJo LLC, Ashland, OR, USA). Percentage of Ki-67 positive cells (acquired through FITC channel) were determined by gating with respect to unstained control for both the TSC1+/+ and TSC1−/− NPCs, respectively. For image-based methods, TSC1+/+ and TSC1−/− NPCs were seeded at 0.3 × 105 cells/well of a 24-well Geltrex-coated plate (three biological replicates per cell line, per treatment). The next day after seeding, medium was exchanged for fresh medium containing the fluorescent nuclear marker NucSpot650 (Biotium, Fremont, CA) at 1:500 dilution, according to the manufacturer’s instructions, along with 50 nM rapamycin, 10 nM RMC-6272 or DMSO as a vehicle control. Medium containing DMSO without NucSpot650 was used as an unstained control. Briefly, live, time-lapse images were acquired using an Incucyte SX5 (Sartorious, Göttingen, Germany) with the 10 × objective. Immediately after seeding, an initial image was acquired to measure the baseline confluence value. NucSpot650 and compounds were added 28 h following seeding, with images acquired every 2 h. The Phase channel was used to image cells, and the NIR channel was used to image nuclei stained with NucSpot650 with a 400 ms exposure. Image acquisition and analysis were performed with the Incucyte 2021C version of the software. We used the Basic Analyzer module to segment cells for the confluence metric and NucSpot650 for the nuclear count metric. Briefly, the parameters for the cell segmentation included a 1.1 Segmentation Adjustment and a minimum area filter set at 300 µm2, and NucSpot650 stained nuclei were segmented using a Top-Hat Segmentation with a radius of 20 µm and NIRCU Threshold value of 2. To filter out dead nuclei, a threshold for the mean max intensity was set at 50 NIRCU. For each well in a 24-well plate, 36 non-overlapping locations were imaged.

Neurite outgrowth assays

For neurite outgrowth, NPCs (6250/cm2) were seeded on Poly-D-lysine coated wells (0.1 mg/ml; Sigma, St. Louis, MO) and Fibronectin (5 µg/ml, Corning, Corning, NY) in growth factor-depleted neural expansion medium (30% NEM) containing 1:1 of neurobasal media and advanced DMEM/F12 (ThermoFisher, Waltham, MA), 1 × penicillin/streptomycin and 0.3 × neural induction supplement (ThermoFisher, Waltham, MA). Cells were grown in the presence of DMSO, 50 nM rapamycin or 10 nM of RMC-6272 for 48 h and fixed with 4% paraformaldehyde (PFA; Microscopy Sciences, Hatfield, PA) for 20 min prior to immunostaining. Cover slips from three biological replicates were analyzed. For each cell line, images from eight independent non-overlapping fields/treatment condition were analyzed using HCA-Vision software V.2.2.0 (CSIRO, Canberra, Australia), which was developed to trace and quantify neurite structure in several parameters including: (1) number of cells; (2) average number of neurites/cell, defined as number of root points where neurites emerge from the cell body; (3) average neurite outgrowth/cell, defined as total length of all neurite structures including primary and branched outgrowths emerging from a body; and (4) average number of extremities/cell, defined as the number of termination points for all segment structures/cell [62]. Images were acquired on a Nikon Eclipse TE2000-U microscope using a Nikon DS-QiMc camera and NIS-Element BR 3.2 imaging software.

Immunocytochemistry

Cells were fixed with 4% PFA for 20 min at room temperature and washed three times with PBS. Non-specific labeling was blocked, and cell membranes permeabilized in a single step, using 4% normal goat serum (NGS) in PBS containing 0.1% Triton-X-100 and 0.05% Tween-20 for 45 min at room temperature. Primary antibodies were diluted in 2% NGS/0.1%Triton-X-100/PBS and incubated for 2 h in the dark at room temperature (see Additional file 2: Table S1 for primary antibodies). Coverslips were mounted in ProLong Gold antifade reagent with DAPI (Invitrogen, Carlsbad, CA) and immunofluorescence was visualized on a Nikon Eclipse TE2000-U microscope. Images were acquired using a Nikon DS-QiMc camera and NIS-Element BR 3.2 imaging software.

Immunoblot analyses

Immunoblotting was performed as previously described [24]. Briefly, cells were lysed in RIPA buffer, and protein lysates were resolved on Novex 4–12% or 10–20% Tris–Glycine gels (Invitrogen, Carlsbad, CA), transferred to nitrocellulose (Bio-Rad, Hercules, CA) and then incubated with primary antibodies (see Additional file 2: Table S1 for primary antibodies). All immunoblotting data shown are a representative of at least three biological replicates.

Statistical analysis

For RNA sequencing, statistical analyses from three biological replicates were performed using RStudio (R v.4.1.1). Changes in translational efficiency were assessed using batch-adjusted analysis of partial variance (APV) in combination with a random variance model implemented in the anota2seq bioconductor package. The p values obtained from the analysis were adjusted using the Benjamini–Hochberg (BH) method. ECDFs were used to cross-compare RNA-seq datasets and assess selective regulation of signatures, and significance was assessed using the Wilcoxon rank-sum test relative to the background. Right-sided hypergeometric tests were used to identify GO terms enriched for genes identified by anota2seq with an FDR cutoff 0.001 or 0.01 for NPCs or postmortem brain samples, respectively (BH method). All tests were two-tailed unless otherwise indicated. For cell size and proliferation, p values were determined from three biological replicates by one-tailed Student’s t test. For neurite outgrowth assays, quantitation from three biological replicates was performed from using HCA-Vision software, and p values were calculated by one-tailed Student’s t test. Plots for neurite data were generated using GraphPad Prism9. Details related to each figure regarding iPSC-derived NPCs and number of independent experiments and replicates are shown in Additional file 8: Table S7.