Translatome analysis of Tuberous Sclerosis Complex-1 patient-derived neural progenitor cells reveal rapamycin-dependent and independent alterations

Tuberous sclerosis complex (TSC) is an inherited neurocutaneous disorder caused by mutations in TSC1 or TSC2 genes, with patients often exhibiting neurodevelopmental (ND) manifestations termed TSC-associated neuropsychiatric disorders (TAND) including autism spectrum disorder (ASD). The hamartin-tuberin (TSC1-TSC2) protein complex inactivates mechanistic target of rapamycin complex 1 (mTORC1) signaling, leading to increased protein synthesis via inactivation of translational repressor eIF4E-binding proteins (4E-BPs). In TSC1-null neural progenitor cells (NPCs), we previously reported early ND phenotypic changes, including increased proliferation/altered neurite outgrowth, which were unaffected by mTORC1-inhibitor rapamycin. Here, using polysome-profiling to quantify translational efficiencies at a transcriptome-wide level, we observed numerous TSC1-dependent alterations in NPCs, largely recapitulated in post-mortem brains from ASD donors. Although rapamycin partially reversed TSC1-associated alterations, most neural activity/synaptic- or ASD-related genes remained insensitive but were inhibited by third-generation bi-steric, mTORC1-selective inhibitor RMC-6272, which also reversed altered ND phenotypes. Together these data reveal potential implications for treatment of TAND.


INTRODUCTION
Tuberous sclerosis complex (TSC) is an inherited multisystem disorder involving a range of symptoms including epilepsy, autism spectrum disorder (ASD), intellectual disability (ID), and slow growing hamartomas in many organs. TSC is caused by mutations in the TSC1 or TSC2 genes, encoding the tumor suppressor proteins hamartin (TSC1) and tuberin (TSC2) 1,2 . The TSC proteins act as a central hub relaying signals from diverse cellular pathways to control mammalian/mechanistic target of rapamycin complex 1 (mTORC1) activity, which regulates cell growth and proliferation 3,4 . Aberrant activation of mTORC1 in TSC has led to rapamycin (Rap) analogs ("rapalogs") emerging as a lifelong therapy for TSC hamartomas, as their discontinuation leads to resumed growth of TSC-associated lesions [5][6][7][8] . Recent clinical trials revealed that rapalogs reduce epilepsy in 40% of TSC patients 9 . In contrast, rapalogs are ineffective in treating 3 TSC-associated neuropsychiatric defects (TAND) and autism 10,11 . Therefore, new treatments for TSC that are superior to rapalogs with respect to anti-proliferative effects in tumors, and efficacy toward the non-tumor CNS manifestations of the disorder are needed.
Among its many activities 24 , mTORC1 plays a major role in regulating gene expression by modulating how efficiently mRNAs are translated into proteins. Consistent with a key role of mRNA translation in determining proteome composition, translatomes (commonly defined as the pool of mRNA associated with ribosomes) 25 resemble proteomes more closely than corresponding transcriptomes [26][27][28] . mTORC1 regulates cap-dependent translation by modulating the assembly of eukaryotic initiation factor (eIF) 4F, a complex consisting of a cap binding protein (eIF4E), a DEAD-box RNA helicase (eIF4A), and a large scaffolding protein (eIF4G). mTORC1 activation leads to direct phosphorylation of two key substrates involved in regulating translation initiation: eIF4E binding proteins (4E-BPs) and ribosomal protein p70S6 kinases (S6Ks). 4E-BPs are a family of translation inhibitors consisting of three members, the best studied being 4E-BP1, which when de-phosphorylated competes with eIF4G for binding to eIF4E and prevents eIF4F complex formation. Once phosphorylated, 4E-BPs dissociate from eIF4E, facilitating eIF4F complex formation 29 . Further, activation of S6K by mTORC1 also affects translation initiation by: (i) increasing eIF4A availability through phosphorylation and degradation of a negative repressor, PDCD4, and (ii) phosphorylating eIF4B which stimulates eIF4A helicase activity and promotes initiation complex formation 30,31 . Interestingly, a recent study reported that mTORC1-driven translation is high in human pluripotent stem cells and is suppressed during neural differentiation 32 . Moreover, numerous changes in mRNA translation without corresponding changes in mRNA levels have been observed across human neuronal development, highlighting the importance of translational control for developing neurons 32 .
Our comparisons of transcriptomes between isogenic NPCs revealed a quantitative genotypedependent response whereby genes upregulated/downregulated in TSC1-heterozygous NPCs were further increased/decreased in TSC1-null cells when compared to genetically matched CRISPR-corrected WT cells. Interestingly, this included genes linked to ASD, epilepsy and ID 17 .
However, despite alterations in mRNA translation being a major mechanism modulating gene expression downstream of mTOR in cancer cells 33 , translatome studies are lacking in TSC stem cell models. Recent studies have documented that early neurodevelopmental events, such as NPC proliferation, neurite outgrowth and migration, that precede synaptogenesis also play a role in disease pathogenesis of ASD and other neuropsychiatric disorders [34][35][36][37][38][39][40] . The enhanced proliferation and neurite outgrowth consistently observed in TSC1-null NPCs when compared with isogenic WT controls suggest that this may underlie early neurodevelopmental defects in TSC 17 .
Using our isogenic NPC model generated from TSC1 patient-derived iPSCs, we identified TSC1sensitive mRNA expression and translation. Strikingly, TSC1-sensitive mRNA translation observed in NPCs was recapitulated in human ASD brain samples from the Brodmann area 19 when contrasted to controls. Furthermore, although polysome-profiling revealed a partial reversal of TSC1-sensitive translation upon rapamycin treatment, most genes related to neural activity/synaptic regulation or ASD showed rapamycin-insensitive translation. However, these genes could be reversed by the third-generation bi-steric mTORC1-selective inhibitor RMC-6272, which more efficiently suppresses 4E-BP1-phosphorylation in NPCs when compared to rapamycin 41 . This was accompanied by reversal of rapamycin-insensitive phenotypes in TSC1null NPCs, suggesting that more efficient targeting of mTORC1 may be an attractive treatment strategy in ASD.

TSC1 loss leads to widespread alterations of mRNA translation in patient-derived isogenic neural progenitor cells (NPCs)
To determine the impact of TSC1 loss-of-function mutations on mRNA levels and translation in NPCs, we used skin fibroblasts from a patient with a heterozygous mutation in TSC1 exon 15 (1746C>T, Arg509X) to derive isogenic TSC1-null NPCs (-/-) and corrected TSC1-WT NPCs (+/+) as described previously 17 (Fig. 1a, Fig. S3a). This isogenic cell pair was then used to assess TSC1-associated changes in gene expression at multiple levels using polysome-profiling 42 .
During polysome-profiling, total cytosolic mRNA is fractionated depending on ribosome association and a pool of polysome-associated mRNA is isolated in parallel with total cytosolic mRNA (Fig. 1a). The resulting RNA pools from TSC1 -/and TSC1 +/+ NPCs were quantified using RNA sequencing followed by analysis using anota2seq 43,44 to identify three modes of TSC1associated gene expression alterations: (i) changes in polysome-associated mRNA not paralleled by corresponding alterations in total mRNA levels (denoted "translation" and, under conditions when translation elongation is unaffected, interpreted as changes in translational efficiency leading to modulation of protein levels); (ii) congruent changes of polysome-associated and total mRNA (denoted "abundance" representing alterations in mRNA levels impacting protein levels downstream of e.g. modulation of transcription and/or mRNA-stability); and (iii) alterations in total mRNA not paralleled by corresponding changes in polysome-associated mRNA (denoted "offsetting" and interpreted as instances where mRNA translation opposes alterations in protein levels imposed by modulation of mRNA levels [discussed in detail elsewhere] 45,46 ). As visualized 6 by a scatterplot comparing TSC1-associated changes in total cytosolic and polysome-associated mRNA (Fig. 1b), and densities of p-values or FDRs (Fig. 1c) from anota2seq analysis, this revealed numerous TSC1-associated alterations in translation and abundance together with instances of translational offsetting. To validate the observed changes in gene expression, we selected 200 genes whose translation or abundance was increased or decreased in TSC1 -/as compared to TSC1 +/+ NPCs and quantified their expression pattern using the NanoString nCounter Gene Expression Analysis technology. This largely confirmed the gene expression modes (Fig. 1d). To further validate changes in mRNA translation, we focused on a subset of genes that did or did not show accompanying alterations in mRNA levels (Fig. 1e, left), and performed RT-qPCR using total or polysome-associated mRNA as input. This identified larger TSC1-associated changes in polysome-associated mRNA as compared to total mRNA for all genes in the validation subset (Fig. 1e, right). Next, we assessed the potential functional impact of these changes in translational efficiencies using ClueGO-based gene ontology analysis 47 . This revealed that translation of mRNAs encoding proteins annotated to e.g. ion transport, immune system function, RNA polymerase II and oxidative phosphorylation were selectively altered ( Fig.   1f; Table S1a-b). Overall, these data show that loss of TSC1 function in NPCs leads to reprogrammed gene expression via alterations in both mRNA abundance and translation.

Alterations of mRNA translation in ASD patient compared with control brains
We next sought to assess whether similar patterns of mRNA translation to those identified as TSC1-associated in NPCs are observed in post-mortem brain tissue from ASD patients. To this end, we obtained 10 brain samples collected from neurotypical and ASD-affected donors matched for age (4-9 years old) and gender (male). Furthermore, all samples originated from the Brodmann Area 19 (BA19), a part of the occipital lobe cortex involved in responses to visual stimuli ( Fig. 2a-b). As brain samples were small, we used a recently developed optimized polysome-profiling technique amenable to small tissue samples 48 to obtain, to our knowledge, a first dataset of transcriptome-wide alterations of mRNA translation in ASD-affected post-mortem brain tissue. Similarly to above (Fig. 1b-c), we analyzed the resulting dataset using anota2seq 7 43,44 (Fig. 2c-d). Despite the limited statistical power to detect gene expression changes due to the scarce availability of matched ASD and control samples, we observed an enrichment of low p-values for ASD-associated alterations in translation (Fig. 2d). This was further supported by a left-shifted distribution of FDRs (Fig. 2d) together with a larger range of fold changes (ASD vs control) for polysome-associated mRNA as compared to total cytosolic mRNA (Fig. 2e).
Strikingly, ClueGO gene ontology analysis revealed that genes regulated via translation are involved in previously ASD-implicated functions including oxidative phosphorylation, MAPK pathway and alternative polyadenylation ( Fig. 2f; Table S1c-d). Therefore, albeit limited by the low availability of tissues for studies, these data support reprogrammed translation in postmortem brain tissue from donors affected by ASD.

TSC1 -/-NPCs recapitulate mRNA translation in brains of ASD patients
We compared the data sets obtained from NPCs and BA19 samples to determine if there are overlaps in gene expression programs. First, we assessed whether transcripts showing TSC1sensitive translation in NPCs showed altered gene expression when comparing samples originating from ASD vs control brains. Strikingly, transcripts whose translation was increased when comparing TSC1 -/to TSC1 +/+ NPCs (i.e. those identified in Fig. 1b) showed increased translation also when comparing ASD to controls (as levels of polysome-associated mRNA were increased while total mRNA levels where unchanged; Fig. 3a). Similarly, transcripts that were translationally suppressed in TSC1 -/relative to TSC1 +/+ NPCs were also translationally suppressed in BA19 from ASD relative to control subjects (Fig. 3a). We next performed the reciprocal analysis by assessing whether transcripts with altered translation in ASD vs control samples (i.e. those identified in Fig. 3a) also showed TSC1-sensitive expression in NPCs.
Indeed, transcripts showing increased or decreased translation when comparing BA19 samples from ASD to controls revealed similar translation patterns when comparing TSC1 -/to TSC1 +/+ NPCs (Fig. 3b). Accordingly, the NPC model captures alterations in mRNA translation occurring in brains of ASD patients.

8
To further explore similarities between the two data sets, we assessed whether synaptic genes (Table S2a) 49 , and transcripts whose translation was previously identified as induced upon overexpression of eIF4E ("eIF4E-sensitive") 50 , are regulated in the NPC and ASD datasets. In agreement with hyperactivation of the mTORC1/eIF4E axis, previously identified eIF4E-sensitive transcripts were translationally activated in TSC1 -/relative to TSC1 +/+ NPCs as well as in ASD vs control brains (as judged by their increased polysome-association in the absence of changes in total mRNA; Fig. 3c-d; Table S2b). In contrast, synaptic genes showed increased levels of both polysome-associated and cytosolic mRNA in both datasets ( Fig. 3c-d). These findings further underline that a more complete understanding of ASD-associated gene expression changes and their mechanistic underpinnings requires studies of mRNA translation.

Rapamycin only partially reverses TSC1-associated translation
As discussed above, TSC1 loss leads to hyperactivated mTORC1 signaling, which affects mRNA translation both globally and selectively 33,51 . Accordingly, mTORC1 inhibitors have been considered as a strategy to treat phenotypes resulting from loss of TSC1 5,6,8 . To assess whether these agents reverse TSC1-associated mRNA translation in NPCs, we used polysome-profiling in cells treated with the mTORC1 inhibitor rapamycin (Fig. 1a). Anota2seq analysis comparing TSC1 -/-NPCs in the presence or absence of rapamycin revealed that short-term treatment almost exclusively modulated mRNA translation ( Fig. 4a-b). Consistent with rapamycin inhibiting translation via the mTORC1/eIF4E axis, translation of mRNAs previously identified as eIF4Esensitive (same subset as in Fig. 3c-d) was suppressed in rapamycin-treated TSC1 -/-NPCs ( Fig.   4c, Table S2b). In addition, rapamycin reduced polysome-association of mRNAs transcribed from some genes with synaptic activity (same subset as in Fig. 3c-d) (Fig. 4d, Table S2a). Next, using the same strategy as above ( Fig. 3a-b), we assessed whether identified rapamycin-sensitive translation is modulated when comparing TSC1 -/vs TSC1 +/+ NPCs or ASD vs control brains. Although rapamycin reverses TSC1-associated changes of mRNA translation in NPCs (Fig. 4e), previous studies have indicated that the effect on the translatome is only partial -likely due to rapamycin incompletely reducing phosphorylation of 4E-BPs 52 . Consistent with incomplete reversal of mTORC1-sensitive translation by rapamycin, neurite outgrowth in TSC1 -/cells was not rescued by rapamycin 17 . To assess the extent to which TSC1-associated translation is sensitive to rapamycin, we separated mRNAs whose translation was altered in TSC1 -/vs TSC1 +/+ NPCs into subsets with rapamycin-reversed (Fig. 5a) or -insensitive translation (Fig. 5b). Indeed, this revealed that only a subset of TSC1-associated translation changes was reversed by rapamycin, while most transcripts were insensitive. A new generation of bi-steric mTORC1selective inhibitors, which suppresses phosphorylation of 4E-BP1 to a greater extent, has recently been developed 41,53 . To evaluate whether these inhibitors may reverse TSC1associated mRNA translation more efficiently than rapamycin, we used our NanoString nCounter Gene Expression Analysis code set to analyze effects on translation in TSC1 -/-NPCs treated with RMC-6272, an mTORC1-selective bi-steric third-generation inhibitor 41,54 compared to rapamycin.
We focused the analysis on NanoString targets with TSC1-associated changes in mRNA translation, separated these into those whose translation was reversed or insensitive to rapamycin, and compared the effects of mTOR allosteric inhibition with mTORC1 bi-steric inhibition. As expected, transcripts showing rapamycin-sensitive translation were also sensitive to RMC-6272 (Fig. 5c). Conversely, transcripts whose translation was insensitive to rapamycin ( Fig. 5d, left) were largely sensitive to RMC-6272 (Fig. 5d, right). Accordingly, these studies suggest that more efficient inhibition of mTORC1 with a bi-steric inhibitor reverses TSC1associated alterations in mRNA translation to a greater extent than rapamycin and may therefore have distinct effects on ASD-associated phenotypes.

TSC1 NPCs show genotype-dependent phenotypes that are reversed by RMC-6272
We previously reported increased cell size as well as early neurodevelopmental phenotypes, including proliferation rate and neurite outgrowth, in TSC1 -/as compared with TSC1 +/+ NPCs.
Rapamycin treatment, while decreasing cell size, did not rescue the increased proliferation or neurite outgrowth 17 . Based on our results indicating that a subset of rapamycin-insensitive genes with genotype-dependent altered translation were reversed by RMC-6272 (Fig. 5d), we tested the ability of RMC-6272 to rescue early neurodevelopmental phenotypes in TSC1 -/-NPCs.
Immunoblotting of TSC1 +/+ and TSC1 -/-NPCs treated with rapamycin or RMC-6272 for 2 or 24 h showed attenuation of phosphorylated ribosomal S6 (p-S6 S240/244) but only RMC-6272 reversed phosphorylation of 4E-BP1. In addition, consistent with our previous report 17 , p-eIF4E was increased upon rapamycin treatment, but remained unchanged in RMC-6272 treated cells (Fig. 6a, Fig. S3b). Furthermore, treatment of NPCs with 50 nM rapamycin or 10 nM RMC-6272 for 24 h led to a similar reduction in cell size when compared with DMSO-treated control cells ( Fig. 6b). As reported previously 17 , proliferation and neurite outgrowth were unaffected by rapamycin treatment. In contrast, RMC-6272 rescued both these phenotypes. A 4-day treatment with 10 nM RMC-6272 inhibited proliferation of TSC1 +/+ and TSC1 -/-NPCs, as determined by viable cell counts using trypan blue exclusion, with cell numbers at days 1-4 (D1-4) remaining close to seeding (D0; Fig. 6c). Furthermore, as we have previously shown, TSC1 -/-NPCs displayed a significant increase in both neurite number and length compared to TSC1 +/+ NPCs ( Fig. 6d-e, left panel). Strikingly, treatment with RMC-6272 led to a significant reduction in neurite number and length, compared with DMSO-treated cells, while rapamycin did not affect the number or length of neurites ( Fig. 6d-e, right panel). Taken together, these data support that treatment with the third-generation mTORC1 inhibitor leads to more complete inhibition of mTORC1 downstream targets including 4E-BP1 and is more potent than rapamycin in rescuing the altered early neurodevelopmental phenotypes such as proliferation and neurite outgrowth in TSC1-mutant NPCs.

DISCUSSION
It is well established that loss of TSC1 or TSC2 results in activation of mTORC1 signaling, which has led to FDA approval for treatment of TSC-associated tumors with first-generation mTORC1 inhibitors such as rapalogs everolimus/RAD-001. However, rapalogs have not been very effective 11 for treating TSC-associated neuropsychiatric defects and autism 10,11 . The mTORC1 signaling pathway plays a critical role in protein synthesis in normal cells including stem cells, and in human disease through regulation of translation initiation (reviewed in 55,56 ).The mTORC1/eIF4F axis is therefore critical in shaping the proteome. Although transcriptome-wide studies of TSCassociated mRNA translation have been performed in mouse embryonic fibroblasts 57 , the effects of TSC-loss in patient-derived NPCs have not been assessed. Here we sought to bridge this gap in knowledge.
Here we, for the first time, reveal the complex pattern of gene expression alterations downstream of TSC1 loss in patient-derived NPCs encompassing both changes in mRNA abundance as well as numerous alterations in translational efficiencies. Interestingly, TSC1-dependent alterations in mRNA translation observed in NPCs were largely recapitulated in human ASD brains. In addition, our study of TSC1-associated gene expression also indicated ample translational offsetting, which denotes a poorly characterized gene expression mode possibly representing adaptation 45,46 . Although this may be of interest to fully understand how TSC1 loss reprograms gene expression, as this mode of regulation was not observed in human ASD brains, we did not study it further herein. Furthermore, although polysome-profiling revealed a partial reversal of TSC1associated gene expression alterations following rapamycin treatment, most genes related to neural activity/synaptic regulation or ASD that showed TSC1-dependent translation were rapamycin-insensitive. Among mTOR inhibitors, first-generation allosteric rapalogs effectively suppress phosphorylation of mTORC1 target S6K1, but not 4E-BP1 in many cell types.
Furthermore, allosteric rapalogs activate AKT, a downstream target of mTORC2, by negative feedback loops 24 , which prompted development of second-generation, orthosteric mTOR kinase inhibitors (active site mTOR inhibitors) including Torin 1, AZD8055 and TAK-228/MLN0128, which potently inhibit both mTORC1 and mTORC2. As mTORC2 promotes lipogenesis, glucose uptake and cell survival through downstream targets AKT and SGK, active site mTOR inhibitors appears to be more toxic than rapalogs 58 . Orthosteric mTOR inhibitors efficiently reduce phosphorylation of 4E-BP1 when compared with rapalogs, but exhibit short residence time compared to rapamycin, resulting in poor in vivo efficacy 33 . The limited clinical benefits of firstand second-generation mTOR inhibitors led to the recent development of a third-generation mTORC1-directed inhibitor RapaLink-1. As a prototype of the bi-steric class of mTOR inhibitors, RapaLink-1 links the high affinity of rapamycin for mTORC1 with the effective active site mTOR inhibition of TAK-228 33 . RapaLink-1 was shown to be highly potent in reducing phosphorylation of both S6K1 and 4E-BP1 while retaining approximately 4-fold selectivity for mTORC1 as compared to mTORC2. RapaLink-1 also showed antitumor efficacy in glioma models in vivo with no significant toxicities 33,58 . More recent bi-steric compounds show higher selectivity for mTORC1 over mTORC2 (more than 30-fold selectivity), along with potent suppression of 4E-BP1 phosphorylation 41,53 . These bi-steric mTORC1-selective inhibitors, including RMC-6272 and its clinical counterpart RMC-5552, show strong antitumor activity either alone or when combined with other treatments in several preclinical cancer models. RMC-5552 also demonstrates preliminary evidence of anti-tumor activity at tolerated doses 41,54 . Here we reveal that RMC-6272 is not only more potent than rapamycin in inhibiting mTORC1, but also reverses some of the translational changes not reversed by rapamycin (Fig. 5d). These findings are consistent with previous comparisons between the effects of rapamycin and the active site mTOR inhibitor PP242 on transcriptome-wide translation in cancer cells 52 . More importantly, unlike rapamycin, RMC-6272 can rescue early neurodevelopmental phenotypes such as proliferation and neurite outgrowth in TSC1 -/-NPCs ( Fig. 6 and 17 ), raising the question whether 4E-BP1-dependent translation could be essential for some of the neurodevelopmental phenotypes in TSC and other mTORC1-activated neurodevelopmental disorders.
In addition to TSC, dysregulated mTORC1 signaling is also observed in other syndromic ASDs   Table S3.

Polysome fractionation and RT-qPCR for NPCs
Lysate preparation for polysome-profiling from 4 biological replicates was carried out as previously described 64 . Briefly, TSC1 NPC lines were seeded at 40,000 cells/cm 2 , with a total of 5x10 7 cells seeded per drug treatment condition. The next day after seeding, cells were treated for 2h with 50 nM rapamycin, 10 nM RMC-6272, or DMSO as a vehicle control. Treated cells were then rinsed with 1X PBS containing 100µg/ml cycloheximide (CHX) (Sigma, St. Louis, MO), harvested by scraping on ice in PBS/CHX, and pelleted by centrifugation at 300g for 10min at +4C. Cell pellets were lysed in 10 mM Tris-HCl (pH 8), 140 mM NaCl, 1.5 mM MgCl2, 0.1% sodium deoxycholate, 0.1% Triton X-100, 1 mM DTT and 100 μg/ml CHX. Lysates were cleared by spinning for 2 min at 13000g and quickly frozen on dry ice. When ready for processing, lysates were thawed and loaded onto a 10-50% sucrose gradient, centrifuged for 2h15 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 OD254. 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 14 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 (3 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 post-mortem brain samples
Polysome fractionation of post-mortem Brodmann area 19 samples 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 42,48 . RNA sequencing libraries were generated using the Smart-seq2 protocol as described previously 48 . 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.

Analysis using anota2seq
Genes with 0 mapped RNA sequencing read in one or more samples were discarded resulting in analysis of 12950 and 11998 genes in post-mortem brain samples and NPCs, respectively. The data was TMM-log2 normalized and analyzed using anota2seq 44

RNA-Seq data preprocessing and quality control
Quality of sequencing reads (paired-end for NPC data set; single-end for postmortem brain data using, in addition to default parameters, "-no-mixed" and "-no-discordant" parameters for NPC dataset) 66 . The aligned reads were summarized using the "featureCounts" function of the

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

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 independently for genes that were found to be translationally regulated in either data set. The difference between each tested gene set and the background was quantified at the 50th quantile and the Wilcoxon ranksum 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 over-expression 50 or genes associated with synaptic function 49 .

Target gene selection and generation of custom nanostring panel
For NanoString nCounter analysis 68 , a custom panel of 200 target genes identified by Anota2seq analysis as regulated by "translation" or "mRNA abundance" were selected: (i) genes with log2FC > 2 and FDR < 0.15 in any of the contrasts applied when analyzing the NPC data set, (ii) targets annotated to ASD/NDD pathology with log2FC > 1 and FDR < 0.15 in the NPC data set and (iii) negative controls were identified based on standard deviation (<0.3) between samples, mean log2 TMM signal in the top 50th quartile, deltaT <0.1 (from anota2seq analysis) and deltaP <0.1 (from anota2seq analysis).

Sample preparation and data analysis
RNA integrity for polysome-associated and cytosolic mRNA (4 replicates of each condition) was Therefore, global normalization was performed using 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 (Figs. 1d and 5c-d).

Validation of differential translation using RT-qPCR
To validate using RT-qPCR, polysomes from TSC1 -/or TSC1 +/+ NPCs (3 biological replicates) were fractionated and pooled as described for NPCs above. RNA was extracted using TRIzol and cDNA prepared using M-MuLV Reverse Transcriptase (New England Biolabs, Ipswich, MA) and oligo(dT)20 primers. 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 Table S4. The level of each mRNA was normalized to b-actin (ACTB) using the comparative CT method and compared across conditions as indicated in figure legends.

Cell size, proliferation and neurite outgrowth assays
For cell size, proliferation, and neurite outgrowth, assays were performed as previously described Processes that were at least two times the length of the cell body were considered as neurites.
The average neurite number per cell and the average neurite length per cell were analyzed using HCA-Vision software V.2.2.0 (CSIRO, Canberra, Australia).

Immunocytochemistry
Cells were fixed with 4% PFA for 20 min at room temperature and washed 3 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 Table S2 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 17 . 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 Table S2 for primary antibodies). All immunoblotting data shown is a representative of at least 3 biological replicates.

Statistical information
For RNA sequencing, statistical analyses from 3 biological replicates were performed using   Fig. 3a) assessing regulation of transcripts whose translation were altered in ASD vs control BA19 in the comparison of TSC1 -/vs TSC1 +/+ NPCs. c-d. ECDF plots assessing genes related to synaptic activation and transcripts whose translation increased upon eIF4E overexpression. Signatures were evaluated in TSC1 -/vs TSC1 +/+ NPCs (left two panels) and ASD vs control BA19 (right two panels). Fold-changes were calculated using polysome-associated mRNA (c) or cytosolic mRNA (d). Wilcoxon rank-sum test p-vales are indicated for the comparison of each gene set relative to the background (i.e. genes not in gene sets). Comparison of TSC1 -/cells in presence vs absence of rapamycin (similar to Fig. 1B-C). c-d.