Autism-specific PTEN p.Ile135Leu variant and an autism genetic background combine to dysregulate cortical neurogenesis

Summary

Alterations in cortical neurogenesis are implicated in neurodevelopmental disorders including autism spectrum disorders (ASDs). The contribution of genetic backgrounds, in addition to ASD risk genes, on cortical neurogenesis remains understudied. Here, using isogenic induced pluripotent stem cell (iPSC)-derived neural progenitor cells (NPCs) and cortical organoid models, we report that a heterozygous PTEN c.403A>C (p.Ile135Leu) variant found in an ASD-affected individual with macrocephaly dysregulates cortical neurogenesis in an ASD-genetic-background-dependent fashion. Transcriptome analysis at both bulk and single-cell level revealed that the PTEN c.403A>C variant and ASD genetic background affected genes involved in neurogenesis, neural development, and synapse signaling. We also found that this PTEN p.Ile135Leu variant led to overproduction of NPC subtypes as well as neuronal subtypes including both deep and upper layer neurons in its ASD background, but not when introduced into a control genetic background. These findings provide experimental evidence that both the PTEN p.Ile135Leu variant and ASD genetic background contribute to cellular features consistent with ASD associated with macrocephaly.

Fu et al. performed bidirectional genome editing on iPSCs derived from individuals with autism and controls. They uncovered that the PTEN p.Ile135Leu variant, found in an ASD-affected individual with macrocephaly, dysregulated cortical neurogenesis in an ASD-genetic-background-dependent fashion by using 2D neural progenitor cells and 3D cortical organoids.

Introduction

Autism spectrum disorders (ASDs) are a group of phenotypically complex and genetically heterogeneous neurodevelopmental disorders.¹ Variants in nearly 1,000 genes have been associated with ASDs.^2,3 Many of these gene variants are de novo, but a role for inherited common variation is likely to be a major contributor to ASD genetic risk.⁴ Approximately 20% of ASD-affected individuals display early brain overgrowth^5,6 and 15% of ASD-affected individuals with early brain overgrowth possess variants in PTEN.⁷ PTEN is a well-known tumor suppressor gene that also acts as a lipid phosphatase that dephosphorylates PIP3 to PIP2, reducing PI3K/AKT activation and leading to reduced cell proliferation or increased apoptosis.^8,9 Compared to tumor-related PTEN mutations that result in increased risk of hereditary cancers, a majority of ASD-associated PTEN variants do not substantially disrupt the lipid phosphatase function of PTEN.¹⁰ In addition, individuals with PTEN variants do not invariably display ASD, and many individuals with ASD do not possess variants in known ASD risk genes. This provides further support for the hypothesis that, in addition to key ASD risk genes such as PTEN, ASD genetic background—the spectrum of variants and mutations found in throughout the genome and/or epigenome—may also contribute to ASD pathology.^11,12 However, experimental support for this hypothesis is lacking.

Here, we have utilized induced pluripotent stem cell (iPSC) models and genome editing to address this question. iPSCs have been extensively used for neurodevelopmental disease modeling.^{13,14,15,16,17} iPSCs can be used to produce NPCs and neurons in 2D culture, as well as 3D self-organizing brain organoids,¹⁸ which mimic the neurogenesis trajectory during human fetal brain development.^19,20 Bidirectional genome editing using CRISPR-Cas9 on iPSCs derived from both healthy control and individuals with ASD enables the effects of key ASD risk genes as well as the contribution of the genetic background to ASD etiology to be determined.

We previously reported increased NPC proliferation in iPSC-derived NPCs from eight individuals with ASD and early brain overgrowth, compared with five age- and gender-matched control lines.¹⁶ We identified a missense variant in PTEN (c.403A>C [p.Ile135Leu] [GenBank: NM_000314.8]) in one of the eight individuals with ASD through whole-exome sequencing (WES).¹⁶ Therefore, we selected this individual-derived iPSC line named “Apex” and one control iPSC named “Chap” and produced isogenic lines with wild-type PTEN (PTEN WT/WT), the ASD PTEN variant (PTEN WT/Ile135Leu), and complete disruption of PTEN (PTEN KO/KO) in both ASD and control backgrounds. We followed the convention of cell line naming used in our previous publication.¹⁶ For the ASD lines, cell line name starts with A (“Arch,” “Apex” to distinguish different ASD lines). For the control lines, their names start with C (“Chap” and “Clay”). Using NPC cultures and self-organizing organoids, we found that the PTEN p.Ile135Leu variant led to increased NPC proliferation, enlarged organoid size, dysregulated genes related to neurogenesis, and, specifically on the ASD genetic background, the overproduction of deep and upper layer neurons. These effects were modified by the genetic background, since the ASD genetic background itself also led to increased NPC proliferation and dysregulation of genes related to neurogenesis depending on the PTEN genotype. Importantly, the organoid size enlargement effect of the PTEN p.Ile135Leu variant was replicated in another control genetic background, Clay, supporting that PTEN p.Ile135Leu is a pathogenic variant. In addition, the ASD background effects on proliferation and on genes involved in neurogenesis were also found in another ASD genetic background, Arch, with the key ASD risk gene CTNNB1 stop-gain variant¹⁶ (c.226C>T [p.Gln76^∗] [GenBank: NM_001098209.2]) corrected by CRISPR-Cas9 genome editing. Furthermore, we observed PTEN WT/Ile135Leu organoids displayed accelerated upper layer neuronal maturation in the ASD background. Lastly, we identified that both the PTEN p.Ile135Leu variant and ASD genetic background led to dysregulated synaptic signaling in multiple neuronal subtypes. These studies provide strong experimental evidence that both a specific ASD variant and ASD genetic background are important for cellular surrogates of brain overgrowth in ASD.

Material and methods

Human iPSCs

All studies that involved the use of the cell lines and/or data mentioned in this paper were approved by the Institutional Review Board of Case Western Reserve University. Written informed consent was provided by the parents for the development of the iPSC lines.¹⁶ One control iPSC Chap and one ASD iPSC Apex described in our previous paper¹⁶ were used as the parental cell lines for generation of the isogenic PTEN iPSC panels. We introduced the ASD PTEN c.403A>C variant into clones of Chap iPSC by using CRISPR-Cas9 genome editing.²¹ This editing process also generated PTEN KO/KO clones in Chap. Additionally, we corrected the ASD PTEN c.403A>C variant in Apex to WT/WT and similarly generated Apex PTEN KO/KO. Similarly, we included another control iPS cell line, Clay, and we installed the same PTEN variant into Clay by using CRISPR-Cas9 genome editing. Finally, we corrected the CTNNB1 p.Gln76^∗ variant observed in another ASD iPS cell line, Arch, by using genome editing. Detailed phenotypic characterization of Clay, Chap, Arch, and Apex iPSCs were reported in the previous publication.¹⁶ Summary information for all nine cell lines used in this paper is provided in Table S1. All cell lines have been confirmed to be karyotypically normal at NanoString via Human Karyotype Panel or at Thermo Fisher Scientific via Karyostat⁺ or at Cell Line Genetics via G-banding. All iPSCs were maintained in mTeSR plus medium (StemCell Technologies) with a feeder-free culture protocol in six-well plates coated with Vitronectin (Gibco, A31804).²² iPSCs were cultured at 37°C and 5% CO₂ with feeding of 2 mL mTeSR plus per well every other day. Passaging of iPSC colonies was carried out with 0.5 mM EDTA (Thermo Fisher Scientific).

Generation of isogenic iPSCs with CRISPR-mediated genome editing

We amplified a 1.6 kb PCR product from Apex genomic DNA spanning the heterozygous PTEN c.403A>C mutation site. We fused this product, containing both wild-type and mutant PTEN fragments with and without the point mutation, with the HindIII digested pUC19 vector via Gibson Assembly²³ and transformed this into competent cells to produce the donor plasmids. We confirmed donor plasmid identity via Sanger sequencing for the plasmids derived from single colonies to determine whether the clone contained the wild-type PTEN donor plasmid, which was used for CRISPR correcting mutation in ASD iPSC, or the mutant PTEN donor plasmid, which was used for CRISPR installing PTEN p.Ile135Leu variant. We co-transfected plasmids (100 ng) containing double nickase cas9 (Cas9n) and two guide RNAs (guide RNA #1 5'-TGTCATCTTCACTTAGCCAT-3', guide RNA #2 5'-AAAGCTGGAAAGGGACGAAC-3') to target Cas9n to the PTEN locus, along with the appropriate donor plasmid (400 ng) with Lipofectamine Stem Transfection Reagent (Invitrogen, STEM00001) in E8 medium in one well of a 24-well plate. After 48 h of transfection, iPSCs were dissociated into single cells with TrypLE Express Enzyme (Gibco, 12605093), which were then seeded into six-well plates at 250 cells per well. We then picked single colonies into individual wells of 48-well plates for further expansion and Sanger sequencing. We then subjected Sanger positive clones to targeted amplicon sequencing with 2 × 250 bp Miseq to ensure purity of clones and that no unintentional indels introduced in the edited clones had occurred. For the generation of the Arch CTNNB1 correction line, we used a similar strategy. Two guide RNAs (guide RNA #1 5'-GTTCCCACTCATACAGGACT-3', guide RNA #2 5'-TTCACTTAAGAACAAGTAGC-3') were used for targeting CTNNB1 c.226C>T locus. Fastq files from Miseq were processed with CRISPResso (version: 2.0.29) for CRISPR-edit analysis.²⁴

Generation of dorsal cortical NPCs

We differentiated the isogenic iPSCs from the control and ASD PTEN panels into dorsal NPCs by using the PSC Neural Induction Medium protocol (A1647801, Gibco). Briefly, iPSCs were seeded in six-well plates at approximately 20% confluence. On the second day, the media was replaced with neuronal induction medium (NIM), which was used for 7 days with 3 mL media changes every other day. At day 7, cells were dissociated into single cells with Accutase (Stemcell Technologies, 07922) and were seeded at 10⁶ cells per well in neural expansion medium (NEM) in six-well plates coated with Geltrex (A1413201, Gibco). This was considered as passage 0 (P0). NPCs were passaged when reaching 85% confluence.

Immunofluorescence staining for 2D NPCs

NPCs were washed in PBS, fixed with 4% paraformaldehyde in PBS for 15 min at room temperature, washed three times with PBS, each for 5 min, and permeabilized with 0.5% Triton X-100 in PBS for 10 min. Blocking was performed with 0.1% Triton X- and 1% BSA in PBS for at least 1 h, followed by overnight primary antibody incubation at 4°C. The following primary antibodies were diluted in the blocking solution: FOXG1 (Abcam, 1:500) and PAX6 (BD Biosciences; 1:500). Following three 5 min washes in PBS, the cells were incubated with the appropriate secondary antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 555 (Thermo Fisher Scientific; 1:500) for 30 min at room temperature, followed by three 5 min washes in PBS. Cells were counterstained with DAPI in PBS (Sigma Aldrich, 1:2,000) for 5 min, then washed three times. Fluorescence was visualized with the Leica DM6000 inverted microscope in the CWRU Light Microscopy Imaging Core. Images were acquired with the Q-Imaging Retiga Xi Firewire High-Speed, 12-bit cooled CCD camera and Volocity software.

Immunoblotting

When NPCs reached 85% confluences in six-well plates, medium was aspirated, cold PBS was added, cells were scraped off the plate, and resuspended in Mammalian Protein Expression Reagent (M-PER, 78501, Thermo Fisher Scientific) supplemented with Halt Protease and Phosphatase Inhibitor Single-Use Cocktail (78442, Thermo Fisher Scientific). Protein lysate quantification was done with the Bradford assay (Bio-Rad). 10 μg protein lysates were loaded into a NuPAGE 4%–12% 17-well gel (NP0329BOX, Invitrogen) and transferred onto nitrocellulose membranes. Blots were blocked in Intercept (TBS) Blocking Buffer for 1 h, primary antibodies (PTEN, 9188S; p-AKT-Ser473, 4060S; p-AKT-Thr308, 2965S; total AKT, 2920S from Cell Signaling Technology, Beta Actin sc-47778, Santa Cruz Biotechnology) were incubated in TBS blocking buffer overnight (1:1,000 dilution), and then blots were washed and incubated with fluorescent secondary antibody (1:20,000) for 1 h. Blot images were captured and quantified with Odyssey XF Imaging System.

Population doubling time

NPCs at passage 3 were seeded at a density of 7 × 10⁴/cm² in four wells of the six-well plate pre-coated with Geltrex. When NPC wells reached 85% confluence, cells were dissociated with Accutase and cells from three wells were counted. Doubling time was calculated on the basis of culture time and the ratio of cell number before seeding and after harvesting. Cells in the fourth well of the six-well plate were collected for bulk RNA sequencing (RNA-seq). Similar procedures were done continuously until P10. Proliferation assays were done simultaneously for all isogenic clones in the control and ASD backgrounds. Three separate proliferation assays were done for the isogenic clone sets.

Bulk RNA-seq

RNA extraction

NPCs were collected by adding 1 mL Trizol into one well of a six-well plate washed with DPBS and stored at −80°C from P3, P4, and P5 across three batches. RNA was then extracted in a single batch with the Direct-zol RNA Miniprep Plus Kit (Zymo Research). The extracted RNA was subjected to 150 bp unstranded paired-end poly-A RNA-seq on the Illumina platform Novaseq (Novogene).

Data alignment and analysis

We aligned paired-end RNA-seq reads to the human reference transcriptome (Ensembl version 103) by using Kallisto version 0.46.2²⁵ for transcript abundance quantification. All downstream analysis was performed with RStudio (2022.07.1, build 554) with R version 4.2.1. We used Tximport R package to summarize transcript quantification data to genes²⁶ followed by the R package edgeR TMM method for normalization.²⁷ Genes with <1 count per million (CPM) among three sample replicates were filtered out. Variance-stabilization for the normalized filtered data was carried out with limma voom,²⁸ and differentially expressed genes were identified with linear modeling with limma with Benjamini-Hochberg multiple testing correction.

To perform gene set enrichment analysis (GSEA), we used the Molecular Signatures Database (MSigDB v.7.5.1) and we used C5 subcluster “biological process” for Gene Ontology (GO) analysis. All gene lists that were used for the input were ranked on the basis of the t statistic of the limma voom output, and we performed the analysis in the R package ClusterProfiler 4.5.1 by using default parameters (minGSsSize = 10, maxGSSize = 500). We used the “comparecluster” function to visualize the GO term enrichment from multiple passages at the same time.²⁹

Genetic analysis for the ASD line

WES for the ASD line Apex fibroblast was performed in our previous paper.¹⁶ Stop-gain and stop-loss variants as well as single-nucleotide variants (SNVs) were pulled out from the original publication;¹⁶ in total 68 genes with either stop-gain or stop-loss variants were found, and 380 genes were found to have SNVs in Apex.

Differential expression genes in the ASD genetic background for 2D NPCs

We performed differential expression analysis by comparing nine samples from Apex and nine samples from Chap NPCs, comparing Apex PTEN correction vs. Chap (nine samples each), and comparing Arch CTNNB1 correction vs. Chap (nine samples each). We used the decideTests function in the limma R package to identify the differentially expressed genes with cutoff adjusted p value = 0.05 and log2 fold change = 0.6 with Benjamini-Hochberg multiple testing correction.

Cortical organoid production

We used published organoid culture protocols with modifications.^14,30,31 When iPSC cultures reached ∼80% confluency, the medium was aspirated and wells rinsed twice with PBS. 1 mL of Accutase was added per well of the six-well plate and incubated for 2 min at 37°C, 5% CO₂, then 2 mL mTESR plus medium was added to the six-well plate. Cells were then scraped off the plate. We performed gentle trituration by using a 5 mL pipette to achieve a single cell suspension, which was transferred to a 15 mL conical tube, collected by centrifugation at 1,200 rpm for 4 min. Cells were resuspended in Cortex Differentiation Medium (Glasgow-MEM, 20% KSR, 1× NEAA, 1× Sodium Pyruvate, 1× Glutamax, 0.7% Beta-Mercaptoethanol [BME], 1% Pen/Strep) plus ALI (1 μM A83-01 [Stem cell Technologies], 100 nM LDN-193189 [Sigma-Aldrich], 3 μM IWR-1 [Sigma-Aldrich]) plus 10 μM ROCK inhibitor Y-27632 (Tocris). Cells were counted with a hemocytometer and seeded at 9,000 cells per well of a 96-well V-bottom ultra-low attachment plate (S-Bio, MS-9096VZ) in 100 μL volume for 48 h. On day 2, 100 μL cortex differentiation medium with ALI and ROCK inhibitor were added. From day 4 to day 10, 100 μL of the old medium was aspirated and 100 μL of fresh cortex differentiation medium with ALI was added every other day. From day 10 to day 18, IWR-1 was excluded, and only AL (A83-01 + LDN-193189) was added to the cortex differentiation medium. At day 18, organoids in 96-well plates were transferred to ultra-low attachment six-well plates (Costar, #3471) in early organoid medium (DMEM/F12, 1× N2, 1× Glutamax, 1× chemically defined lipid concentrate, 1% Pen/Strep, and 0.1% Fungizone), with approximately 16 organoids per well and three wells for each genotype and cultured on an orbital shaker (Thermo Fisher Scientific) at 85 rpm at 37°C, 5% CO₂. Medium changes were performed every other day. At day 35, the culture medium was switched to late organoid medium (DMEM/F12, 10% FBS, 1× N2, 1× Glutamax, 1% Pen/Strep, 0.25% Fungizone, 0.1% Heparin, 1× chemically defined lipid concentrate) supplemented with fresh 1% Matrigel. From day 70 onwards, culture medium was switched to final organoid medium (DMEM/F12, 10% FBS, 1× N2, 1× B27 without vitamin A, 1% Pen/Strep, 0.25% Fungizone, 0.1% Heparin, 1× chemically defined lipid concentrate) with 2% Matrigel added fresh, and medium was changed every other day.

For cortical organoids for Clay and Clay PTEN WT/Ile135Leu, a previously published cortical organoid protocol¹⁴ was used with modification. The same day mTeSR plus media change was performed before dissociating iPSCs into single cell with Accutase. 9,000 cells were seeded in the V-bottom ultra-low attachment 96-well plate in cortex differentiation medium supplemented with 5 μM SB431542 and 3 μM IWR-1 as well as 20 μM ROCK inhibitor Y-27632 at day 0. On day 2, 100 μL cortex differentiation medium with 5 μM SB431542, 3 μM IWR-1, and 20 μM ROCK inhibitor were added. From day 4 to day 18, 100 μL of the old medium was aspirated and 100 μL of fresh cortex differentiation medium supplemented with 5 μM SB431542 and 3 μM IWR-1 was added; medium change was performed in this fashion every other day.

Organoid size analysis

Bright field images of organoids at day 14, day 18, week 4, and week 8 were obtained via the Leica DMi1 Inverted Microscopes, and the organoid size was determined with a pixel-based method with customized script (https://github.com/shuaifu93/codes-for-PTEN-manuscript) with ImageJ version 1.53k.

Fixation and frozen sectioning of organoids

Organoids were fixed on ice in 4% paraformaldehyde for 30 min, followed by three rinses in PBS, and allowed to sink in 30% sucrose in PBS at 4°C overnight. Organoids were then placed in cryomolds (Tissue Tek, 4565) with OCT compound (Tissue Tek, 4583) and 30% sucrose (1:1), snap frozen on dry ice, and stored at −80°C. Organoids were sectioned via cryostat (Leica) sequentially at 20 μm thickness. Sections were placed on microscope glass slides, dried overnight, and stored at −20°C for subsequent immunohistochemistry.

Immunohistochemistry of organoid frozen sections

Sections were thawed and air-dried at room temperature and then washed two times in TBS-T (Tris-Buffered Saline, 0.1% Triton X-100) for 10 min per wash. Slides were blocked with 10% donkey serum in TBS-T for 30–45 min at room temperature in a humidifying chamber. Slides were then incubated with primary antibodies diluted in blocking solution at 4°C overnight. The next day, slides underwent four washes in TBS-T (10 min per wash) and then were incubated with secondary antibodies at room temperature for 1 h. Slides underwent another three washes in TBS-T (10 min per wash). Slides were counterstained with DAPI in TBS for 5 min then washed in TBS for 5 min before coverslips were mounted with fluoromount (Sigma, F4680). Tissue sections were imaged at a 20× objective on a Hamamatsu Nanozoomer S60 slide scanner in the CWRU Light Microscopy Imaging Core. Image analysis was performed on NPD.view2 (Windows [ver. 2.9.29]) and ImageJ (version 1.53k).

Dissociation of brain organoids and scRNA-seq

Organoids were dissociated into single-cell suspension with the Papain dissociation system (Worthington Biochemical, LK003150). We then loaded cells onto the Chromium Next GEM Chip G Single Cell Kit, (10x Genomics, 1000120) and processed them with Chromium Controller to generate single-cell gel beads in emulsion. We used the Chromium Next GEM Single Cell 3′ Kit v.3.1 (10x Genomics, 1000268) for all scRNA-seq library preparations. Libraries were pooled from different samples on the basis of molar concentrations and sequenced on NovaSeq (Illumina) with 28 bases for read 1 and 91 bases for read 2 (Novogene).

scRNA-seq data analysis

Read alignment and processing

Raw demultiplexed fastq files were uploaded into the 10x Genomics cloud, and reads were aligned to Human (GRCh38) 2020-A with Cell Ranger v6.1.2 with default settings. The generated “filtered cell matrix HDF5 file” for each sample was loaded into Seurat 4.1.1.³² We removed all cells expressing less than 200 genes as well as cells containing more than 30% mitochondria content. We used SCTransform v2 to process each Seurat object to account for sequencing depth and batch effect³³ and then selected the top 3,000 variable features by using the SelectIntegrationFeatures function. We integrated the organoid datasets collected from the same timepoint as well as fetal brain dataset³⁴ by using canonical correlation analysis (CCA). Principal-component analysis (PCA) was performed on the scaled data for the variable genes. We then used PCA to cluster cells by using the Seurat FindNeighbors function with reduced dimension 1:30 followed by FindClusters with resolution = 1.5. UMAP was used for dimension reduction and visualization. We derived the top expressed genes in each cluster by first normalizing SCT reads with PrepSCTFindMarkers and then using the FindMarkers function in the “Seurat” R package. Cell types for each cluster were manually annotated on the basis of the fetal brain dataset,³⁴ marker gene expression (Figures 5H, S6A and S6B), and literature searches. All UMAPs were visualized with SCP R package version 0.4.0 (https://github.com/zhanghao-njmu/SCP).

Figure 5.

Reproducible effects of ASD genetic background on proliferation and dysregulation of neurogenesis in additional control and ASD lines

(A) Day 14 organoid size quantification for organoids derived from Clay and Clay PTEN WT/Ile135Leu iPSC lines.

(B) Day 18 organoid size quantification for organoids derived from Clay and Clay PTEN WT/Ile135Leu iPSC line.

(C) Week 4 organoid size quantification for organoids derived from Chap and Apex PTEN correction iPSC lines.

(D) Week 8 organoid size quantification for organoids derived from Chap and Apex PTEN correction iPSC lines.

(E) Day 18 organoid size quantification for organoids derived from Chap and Arch CTNNB1 correction iPSC lines.

(F) Day 18 organoid size quantification for organoids derived from Clay and Arch CTNNB1 correction iPSC lines.

(G) Week 4 organoid size quantification for organoids derived from Clay and Arch CTNNB1 correction iPSC lines.

(H) Week 8 organoid size quantification for organoids derived from Clay and Arch CTNNB1 correction iPSC lines. Unpaired t test was used for (A)–(H). ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001; ns, not significant.

(I) UMAP plots for visualizing clusters of NPC subtypes and neuron subtypes for week 10 cortical organoids, UMAP plot in the right panel demonstrates that scRNA-seq cell populations for each of the isogenic lines displayed similar spreads.

(J) Dot plot for the visualization of the markers genes for each annotated cell type of the integrated dataset.

(K) GSEA for the effect of ASD genetic background in NPC subtypes comparing Apex PTEN correction vs. Chap.

(L) GSEA for the effect of ASD genetic background in NPC subtypes comparing Apex PTEN correction vs. Clay.

(M) GSEA for the effect of ASD genetic background in NPC subtypes comparing Arch CTNNB1 correction vs. Chap.

(N) GSEA for the effect of ASD genetic background in NPC subtypes comparing Arch CTNNB1 correction vs. Clay.

(O) GSEA comparing Apex PTEN correction vs. Arch CTNNB1 correction across NPC subtypes.

(P) GSEA comparing Clay vs. Chap across NPC subtypes. GO terms were preselected for visualization and p values were adjusted with Benjamini-Hochberg correction.

Differential expression analysis for scRNA-seq

We used the Seurat’s FindMarkers to perform differential expression between genotypes of interest by using their default, the Wilcoxon test. In addition, we removed genes that are expressed in less than 10% of the cells by using min.pct = 0.1. We filtered all the genes by using logfc.threshold = −Inf and then ranked the genes for GSEA on the basis of avg_log2FC. We performed GSEA between genotypes by using clusterprofiler 4.5.1, and we used the comparecluster function to visualize the Gene Ontology (GO) term enrichment across different NPC subtypes or neuronal subtypes with default settings.²⁹

Pseudotime analysis

We performed Monocle3 (version 1.2.9) pseudotime analysis³⁵ on filtered datasets including cycling progenitor cells, ventricular radial glia cells (vRGs), truncated radial glia cells (tRGs), outer radial glia cells (oRGs), IPCs, Cajal-Retzius cells, deep layer excitatory neurons, and upper layer excitatory neurons as well as newborn excitatory neurons. SOX2⁺ radial glia cells were set as the root cells. We down-sampled the cell number to ensure equal numbers between genotype comparisons, and we plotted the distribution of cells from deep and upper layer neuron clusters along the pseudotime trajectory by using ridgeplots.

Results

PTEN p.Ile135Leu variant leads to increased NPC proliferation in both control and ASD genetic background

To test the effect of the ASD PTEN p.Ile135Leu variant and genetic background on the cellular phenotypes associated with ASD, we generated isogenic PTEN iPSCs in both ASD and control backgrounds. We corrected the PTEN c.403A>C variant in the ASD iPSC line (referred to as “Apex”) containing this variant by using CRISPR-Cas9^D10A double nickase genome editing.²¹ We also introduced the same heterozygous PTEN c.403A>C variant into the one control iPSC line (referred to as “Chap,” Figures S1A and S1B). In addition, we abolished PTEN expression by creating PTEN KO/KO lines in both control (Chap) and ASD (Apex) genetic backgrounds (Figure 1B). We validated our CRISPR-edited clones by using Sanger (Figure S1C) and Miseq (Figure S1D) amplicon sequencing. We confirmed that our edited iPSCs were isogenic with the parental lines on the basis of short tandem repeat (STR) profiles (data not shown) and were karyotypically normal by using either NanoString human karyotype panel or traditional G-banding (Figures S1E and S1F). Thus, we produced two sets of isogenic PTEN iPSC lines: one set on the control background (Chap PTEN WT/WT, Chap PTEN WT/Ile135Leu, and Chap PTEN KO/KO) and one set on the ASD genetic background (Apex PTEN WT/WT, Apex PTEN WT/Ile135Leu, and Apex PTEN KO/KO).

Figure 1.

PTEN p.Ile135Leu variant dysregulates genes important for neurogenesis and impairs canonical PTEN activity in the ASD but not in the control genetic background

(A) Isogenic PTEN panel iPSCs were differentiated to neural progenitor cells (NPCs). Chap WT/WT refers to control line Chap with wild-type PTEN, Chap WT/I135L refers to control line Chap installed with the ASD heterozygous PTEN c.403A>C variant, and Chap KO/KO denotes control line Chap with complete disruption of PTEN. Apex WT/WT refers to ASD line Apex with heterozygous PTEN c.403A>C variant corrected to wild-type with CRISPR-Cas9. Apex WT/I135L refers to the original ASD line Apex, which contains heterozygous PTEN c.403A>C variant. Apex KO/KO refers to ASD line Apex with complete disruption of PTEN. From passages 2 to 6, cells were plated at the same density and population doubling time at each passage was calculated. Results of all lines (three independent cell culture replicates per line) are presented as mean ± SD. p values were calculated with one-way ANOVA with Sidak’s correction for multiple comparisons. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001; ns, not significant.

(B) Representative immunoblot results, 10 μg lysates from isogenic PTEN NPCs at passage 5 were used for immunoblot for PTEN, beta-actin, p-AKT-Ser473, p-AKT-Thr308 as well as total-AKT.

(C) Quantification of the ratio of p-AKT-S473 to total AKT, n = 8, four independent experiments were performed and each with duplicate sample loading.

(D) Quantification of the ratio of p-AKT-T308 to total AKT, n = 8, four independent experiments, duplicate samples were loaded for each experiment. For both (C) and (D), results are presented as mean ± SD, p values were calculated with repeated measure one-way ANOVA with Sidak’s correction for multiple comparisons. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001; ns, not significant.

(E) GSEA for the effect of PTEN WT/Ile135Leu in the control background. We used the comparecluster function to run GSEA for multiple passages at the same time, with showCategory = 5. p values were adjusted with the Benjamini-Hochberg correction.

(F) GSEA for the effect of PTEN WT/Ile135Leu in the ASD genetic background. We used the comparecluster function to run GSEA for multiple passages at the same time, with showCategory = 6. p values were adjusted with the Benjamini-Hochberg correction.

To first test the effect of the ASD PTEN p.Ile135Leu variant and genetic background on the NPC proliferation phenotype, we differentiated the three isogenic lines in the control genetic background (Chap PTEN WT/WT, Chap PTEN WT/Ile135Leu, and Chap PTEN KO/KO) and three isogenic lines in the ASD genetic background (Apex PTEN WT/WT, Apex PTEN WT/Ile135Leu, and Apex PTEN KO/KO) into dorsal forebrain NPCs expressing the markers PAX6 and FOXG1 (Figure S2). We measured proliferation by using the population doubling time assay. The PTEN WT/Ile135Leu NPCs on both control and ASD genetic backgrounds displayed increased NPC proliferation as indicated by the significantly decreased doubling time across multiple passages (Figure 1A). Similar effects on NPC proliferation were observed when we abolished the expression of PTEN in both control and ASD genetic background, suggesting that the point mutation provides sufficient reduction of PTEN activity compared to the complete PTEN knockout for accelerated NPC proliferation. Additionally, the Apex PTEN WT/WT NPCs proliferated faster than Chap PTEN WT/WT NPC (Figure 1A), indicating that ASD genetic background was also important for increased NPC proliferation (Figure 1A).

PTEN p.Ile135Leu variant disrupts PTEN activity in part by activating PI3K/AKT

To test whether the p.Ile135Leu variant impairs PTEN function, we performed immunoblot by using lysates from NPC P5, P6, and P7 of all six isogenic PTEN NPCs and looked for readouts of canonical PTEN activity by using anti-p-AKT-Ser473 and -Thr308 antibodies (Figures 1B–1D, S3A, and S3B). As expected, PTEN KO/KO led to the complete loss of PTEN and dramatically decreased PTEN activity, indicated by increased p-AKT-Thr308 and p-AKT-Ser473 levels in both control and ASD genetic backgrounds. The Apex PTEN WT/Ile135Leu NPCs in the ASD background displayed a slight decrease of PTEN canonical activity indicated by very mild increases in p-AKT-Thr308 and p-AKT-Ser473 levels compared to PTEN KO/KO in the Apex ASD genetic background (Figures 1B–1D). To our surprise, we found that Chap PTEN WT/Ile135Leu NPCs in the control background displayed no effect on the levels of both p-AKT-Ser473 and p-AKT-Thr308 (Figures 1B–1D), suggesting that the PTEN WT/Ile135Leu variant affects PTEN activity differently in control versus ASD genetic backgrounds.

PTEN p.Ile135Leu variant dysregulates neurogenesis in NPCs as assessed by RNA-seq

To examine the effects of PTEN variant at the transcriptomic level, we then performed bulk RNA-seq on NPCs derived from each of the isogenic PTEN iPSC lines. In total, 54 RNA samples were sequenced: three experimental replicates for each isogenic iPSC-derived NPC group across three different passages averaging 30M paired-end reads per sample. We employed GSEA to provide sensitive and unbiased gene expression profiles for each PTEN genotype and genetic background.^36,37 We then focused on the top enriched GO terms due to PTEN variants in each genetic background across all passages for analysis. The PTEN WT/Ile135Leu variant in the control genetic background led to downregulated genes enriched for GO terms related to neurogenesis (regulation of neurogenesis) and neuron development (regulation of neuron projection development, axon development, forebrain development, regulation of axonogenesis, telencephalon development, sensory system development, camera type eye development, and sensory organ development). We identified upregulated genes enriched for GO terms related to mitochondria-related processes (ATP metabolic process, oxidative phosphorylation, ATP synthesis coupled electron transport, mitochondrial translation, and mitochondrial respiratory chain complex assembly) and upregulated genes enriched for GO terms related to biosynthetic process (e.g., nucleoside phosphate biosynthetic process, sterol biosynthetic process, etc.) as well as GO terms related to ribosome biogenesis and protein folding (Figure 1E).

We performed similar GSEA for the PTEN WT/Ile135Leu NPCs in the ASD genetic background. Downregulated genes enriched for GO terms related to embryonic development (embryonic organoid morphogenesis, and embryonic skeletal system development) were found. In addition, we identified downregulated genes enriched for GO terms related to cytoplasmic translation and upregulated genes enriched for GO terms related to microtubules. Interestingly, upregulated genes in the control genetic background were enriched for GO terms related to ribosome biogenesis, while in the ASD genetic background, genes enriched for the same GO terms were downregulated in the ASD genetic background because of PTEN WT/Ile135Leu variant. Strikingly, GO terms related to neuron development (cerebral cortex development) that were enriched by downregulated genes in the control background were enriched by upregulated genes in the ASD genetic background (Figures 1E and 1F). These results suggest that GO enrichment due to the PTEN WT/Ile135Leu variant is dependent on the genetic background, and the ASD genetic background can even reverse the direction of GO terms enriched due to this PTEN WT/Ile135Leu variant in the control background.

ASD genetic background dysregulates neurogenesis in NPCs as assessed by RNA-seq

Our experimental design also allowed us to examine the effects of ASD vs. control genetic background with PTEN WT/WT and PTEN WT/Ile135Leu as well as PTEN KO/KO groups, focusing on the GSEA-significant terms identified in an unbiased fashion (shown in Figure 1). For the PTEN WT/WT group, the ASD genetic background resulted in an enrichment of downregulated genes associated with GO terms related to neurogenesis (regulation of neurogenesis) and neuron development (cerebral cortex development, forebrain development, and axon development) as well as microtubule-related terms (Figure 2A). We also observed upregulated genes enriched for GO terms related to biosynthetic process, mitochondria translation, ribosome biogenesis, embryonic development, cytoplasmic translation, and protein folding (Figure 2A).

Figure 2.

ASD genetic background dysregulates genes important for neurogenesis in 2D NPC culture

(A–C) GSEA for the effect of Apex ASD genetic background with PTEN WT/WT group (A), with PTEN WT/Ile135Leu group (B), and with PTEN KO/KO group (C). GO terms were preselected for visualization and p values were adjusted with the Benjamini-Hochberg correction.

For the PTEN WT/Ile135Leu group, the ASD genetic background affected similar GO terms, but the direction of the gene expression profile was often inverted. For example, genes enriched for GO terms related to “forebrain development” were now upregulated, while genes enriched for GO terms related to “ribosome biogenesis,” “mitochondria translation,” “biosynthetic process,” “embryonic organ morphogenesis,” and “cytoplasmic translation” were now downregulated (Figure 2B). Furthermore, for the PTEN KO/KO group, the ASD genetic background affected genes enriched for GO terms related to neuron development in both positive (sensory system development) and negative (axon development, regulation of neuron projection development) directions (Figure 2C). These results suggest that the ASD genetic background affected similar pathways as the PTEN p.Ile135Leu variant. Of note, ASD genetic background itself contributed to dysregulation of genes important for neurogenesis, in addition to the effects of the PTEN p.Ile135Leu variant. Finally, as the severity of PTEN disruption increased from point mutation to complete loss of function, there were fewer GO terms altered in neurogenesis/neural development pathways selected in an unbiased fashion (Figure 1), suggesting that the severity of the PTEN mutation and the ASD background together contributed to the dysregulation of neural development.

PTEN p.Ile135Leu variant in ASD genetic background leads to overproduction of neural progenitors and deep and upper layer neurons in cortical organoids

To further interrogate the effect of PTEN p.Ile135Leu during cortical neurogenesis, we derived cortical organoids from our isogenic PTEN panel iPSCs by using a dual SMAD and WNT inhibition protocol^14,30,31 and cultured the organoids until week 21. We quantified organoid sizes and found that PTEN WT/Ile135Leu organoids displayed enlarged size in both control and ASD genetic background at week 4 (Figure 3A). However, at week 8, PTEN WT/Ile135Leu organoids continued enlarging in the control genetic background, while similar organoid enlargement was not observed in ASD background (Figure 3B). This suggests that the effect of this ASD-specific PTEN variant on organoid growth was affected by ASD genetic background, consistent with findings in NPCs. In addition, organoids produced in the ASD genetic background were always larger than those from the control background regardless of the PTEN genotype. We fixed organoids and performed immunohistochemistry for NPC subtypes as well as deep and upper layer neurons on the isogenic PTEN organoids between weeks 4 and 21(Figures 3C, 3J, S4B, S5B, and S5D) (Figures S4A, S4C, S4D, S5A, and S5C). As deep layer neurons and upper layer neurons can be generated from both intermediate progenitor cells (IPCs) and (oRGs,³⁸ we then stained for TBR2, an IPC marker, and HOPX, an oRG-specific marker. Increased production of IPCs at weeks 4 and 10 as well as oRGs at week 10 (Figures 3D, 3E, and 3K) were found in Apex PTEN WT/Ile135Leu vs. Apex PTEN WT/WT ASD background organoids. Surprisingly, no significant differences were observed for the PTEN WT/Ile135Leu variant on either IPCs or oRGs production in the Chap control genetic background. As expected from the IPC and oRG findings, we identified a significantly increased proportion of cells expressing deep layer neuron markers such as CTIP2 and TBR1 (Figures 3F, 3G, 3H, and 3I), as well as the overproduction of upper layer neurons as indicated by SATB2 staining (Figures 3L and 3M). This was found in Apex PTEN WT/Ile135Leu vs. Apex PTEN WT/WT ASD background organoids but not in the Chap PTEN WT/Ile135Leu vs. Chap PTEN WT/WT control background organoids. These results suggest that PTEN WT/Ile135Leu variant leads to increased production of NPC subtypes including IPCs and oRGs only in the ASD genetic background, leading to the overproduction of the deep layer and superficial layer neurons.

Figure 3.

PTEN p.Ile135Leu variant leads to NPC subtypes and neuron overproduction in the ASD genetic background but not in the control genetic background

(A–M) Organoid size quantification for week 4 (A) and week 8 (B). One-way ANOVA was used, and p values were adjusted with Sidak’s correction for multiple comparisons. ^∗p < 0.05, ^∗∗∗∗p < 0.0001; ns, not significant. Representative IHC images for the isogenic PTEN organoids for week 4, week 8, week 10, week 15, and week 21 for the NPC marker PAX6, the IPC marker TBR2 (C), and the upper layer neuron marker SATB2 (J), scale bar, 250 μm. Quantification of TBR2/DAPI ratio at week 4 (D) and week 10 (E). Quantification of HOPX/DAPI ratio at week 10 (K). Quantification of week 4 and 10 deep layer neuron proportion (F, G, H, and I) and upper layer neuron proportion (L and M). We used one-way ANOVA with Sidak’s correction for multiple comparisons to calculate statistics for (D)–(I) and (K). For (L) and (M), unpaired t test was used. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001; ns, not significant. We used standard deviation (SD) to plot all error bars.

PTEN p.Ile135Leu variant and ASD genetic background dysregulate neurogenesis in cortical organoids

Since the cortical organoids produce heterogeneous cell types, we performed single-cell RNA sequencing (scRNA-seq) by using the 10× platform on week 10 and 21 cortical organoids made from each of the isogenic ASD and control background lines. We mixed three individual organoids from each genotype at each timepoint to account for organoid-to-organoid variation. After quality control, we profiled 95,227 cells for 12 samples, averaging 55k reads per cell. To assist with cell type annotation, we integrated our brain organoid datasets at each timepoint with a published, comprehensively annotated human fetal brain scRNA-seq dataset (https://cells.ucsc.edu/?ds=cortex-dev).³⁴ Our organoid scRNA-seq data integrated well with the fetal brain dataset. Diverse NPC and neuronal subtypes were produced at both 10 (Figure 4A) and 21 (Figure 4B) weeks, including IPC, oRGs, vRGs, tRGs, and cycling progenitors, as well as deep layer excitatory neurons, upper layer excitatory neurons, and interneurons (Figures 4A, 4B, S6A, and S6B). Several unknown clusters do not overlap with the brain dataset, which are likely non-brain cells due to mis-differentiation, which has also been found in other organoid datasets.^19,39,40,41 The integrated dataset enabled us to select specific cell types to perform differential expression analysis with GSEA between genotypes to study the effects of the PTEN p.Ile135Leu variant and ASD genetic background.

Figure 4.

PTEN p.Ile135Leu variant and ASD genetic background lead to dysregulated neurogenesis in NPC subtypes in cortical organoids

(A–G) UMAP plots for visualizing clusters of NPC subtypes and neuron subtypes for week 10 (A) and week 21 cortical organoids (B). Right panel UMAP plots of cells for both week 10 (A) and week 21 organoids (B) demonstrate that scRNA-seq cell populations for each of the isogenic lines displayed similar spreads. (C) GSEA on the effect of PTEN WT/Ile135Leu in the control genetic background in NPC subtypes from both week 10 and week 21 organoids. S indicates cycling progenitors at S phase; M indicates cycling progenitors at M phase; vRG indicates ventricular radia glia cells; tRG indicates truncated radial glia cells; IPC indicates intermediate progenitor cells; oRG indicates outer radial glia cells. w10 indicates week 10, w21 indicates week 21. (D) GSEA on the effect of PTEN WT/Ile135Leu in the ASD genetic background in NPC subtypes from both week 10 and week 21 dataset. GSEA for the effect of ASD genetic background in NPC subtypes with PTEN WT/WT group (E), with PTEN WT/Ile135Leu group (F) and with PTEN KO/KO group (G). GO terms were preselected for visualization and p values were adjusted using Benjamini-Hochberg correction.

First, we performed GSEA to study the effect of the PTEN WT/Ile135Leu variant in the control genetic background in NPC subtypes including IPCs, oRGs, vRGs, and tRGs as well as cycling progenitor cells (S phase, M phase) at week 10 and week 21. The PTEN WT/Ile135Leu organoids displayed downregulated genes enriched for GO terms related to neuron development and neurogenesis at weeks 10 and 21 (Figure 4C), consistent with findings from NPC cultures (Figure 1E). Interestingly, the PTEN p.Ile135Leu variant affected genes related to GO terms such as neuron development and neurogenesis in all NPC subtypes (Figures 4C and S6C). We identified additional downregulated genes enriched for GO terms such as “glia cell differentiation” and “regulation of gliogenesis,” suggesting that in addition to dysregulating neurogenesis, this PTEN variant affected gliogenesis. Consistent with NPC cultures, we observed upregulated genes enriched for GO terms related to mitochondria including “ATP metabolic process” in M phase cycling progenitor and “oxidative phosphorylation” in oRG at week 21. Genes enriched for GO terms related to “ribosome biogenesis” were also upregulated as in 2D NPC cultures. Unexpectedly, we found that the PTEN p.Ile135Leu variant downregulated genes enriched for GO terms related to biosynthetic processes in NPC subtypes including vRGs, IPCs, and oRGs in 3D organoid culture, whereas these GO-term-enriched genes were upregulated in 2D NPC culture (Figure 1E, 4C, and S6C).

Second, we studied PTEN WT/Ile135Leu organoids on the ASD genetic background for the effects of the PTEN variant on the NPC subtypes. The PTEN WT/Ile135Leu ASD organoids displayed upregulated genes enriched for GO terms related to neuron development and neurogenesis across NPC subtypes in both week 10 and week 21. We additionally identified dysregulated genes related to GO terms such as “glia cell differentiation,” “regulation of gliogenesis,” and “regulation of glia cell differentiation,” which were upregulated in cycling progenitor cells, tRGs, and oRGs at week 21 but were downregulated in tRGs at week 10 (Figures 4D and S6C). These results suggest that PTEN p.Ile135Leu variant also dysregulates gliogenesis in the ASD genetic background. Consistent with 2D NPC findings, we observed microtubule-related GO terms were enriched by genes upregulated in IPCs and oRGs as well as cycling progenitor cells. However, GO terms related to ribosome biogenesis and cytoplasmic translation were consistently enriched by genes that were downregulated by this PTEN variant in the ASD genetic background, concordant with 2D NPC bulk RNA-seq findings (Figures 1F, 4D, and S6C).

Third, we asked whether ASD genetic background exerted similar effects on NPC subtypes in 3D culture as in 2D NPC culture. In the PTEN WT/WT group, the ASD genetic background resulted in downregulated genes enriched for GO terms related to neurogenesis and gliogenesis irrespective of NPC subtype (Figure 4E), which is consistent with 2D NPC RNA-seq findings (Figure 2A). Of note, genes that are associated with GO terms related to neuron development were both up- and downregulated from the ASD genetic background at week 21 in oRGs, whereas uniform downregulation effect was observed at week 10 for this cell type. ASD genetic background led to upregulated genes enriched for GO terms including “cytoplasmic translation,” “ribosome biogenesis,” “embryonic organ morphogenesis,” and “mitochondria translation,” all terms were identified in 2D NPC culture (Figures 2A, 4E, and S6D). In the PTEN WT/Ile135Leu group, the ASD genetic background resulted in upregulated genes enriched for GO terms related to neuron development, neurogenesis, and gliogenesis in NPC subtypes, completely consistent with NPC bulk RNA-seq results (Figures 2B, 4F, and S6D). Interestingly, we observed that the ASD genetic background displayed downregulated genes enriched for GO terms related to ribosome biogenesis, cytoplasmic translation, and mitochondria at week 10. However, these GO terms were activated at week 21 (Figures 4F and S6D). Contrary to the effect of ASD genetic background on the PTEN WT/Ile135Leu group, we observed uniform upregulated genes enriched for GO terms related to ribosome biogenesis and cytoplasmic translation, as well as uniform downregulation of genes associated with mitochondria-related GO terms in NPC populations in organoids with the PTEN KO/KO group. In addition, the effect of ASD genetic background on neuron development and neurogenesis became asynchronized across NPC subtypes (Figures 2C, 4G, and S6D).

These results provide additional evidence that the PTEN p.Ile135Leu variant found in an ASD individual and ASD genetic background affect similar pathways, including dysregulating neurogenesis and gliogenesis. The ASD genetic background can reverse the effect of a mild ASD-specific PTEN p.Ile135Leu variant, while the effect of the ASD genetic background can also be modulated by the more severe complete loss of function of PTEN.

Reproducible effects of ASD genetic background on proliferation and dysregulation of neurogenesis in additional control and ASD lines

We next addressed whether the effect of ASD genetic background was generalizable to other control and ASD cell lines. We first introduced the same heterozygous PTEN c.403A>C variant into another control iPSC, referred to as Clay¹⁶ (Figure S1H), to produce a second set of control background lines for analysis (Clay PTEN WT/WT, Clay PTEN WT/Ile135Leu). Next, we chose another iPSC line, Arch, derived from an individual with ASD and macrocephaly, which contains a CTNNB1 p.Gln76^∗ variant.¹⁶ We corrected this variant by using CRISPR-Cas9 genome editing to create another ASD genetic background (Figure S1G).

As a surrogate for proliferation, we measured early organoid size when organoids consist mainly of neural progenitors. As noted above, PTEN WT/Ile135Leu organoids displayed enlarged size in the control genetic background at week 4 (Figure 3A) and week 8 (Figure 3B), although PTEN WT/Ile135Leu organoids in the ASD background only enlarged at 4 weeks (Figure 3A). This organoid size enlargement phenotype due to the PTEN p.Ile135Leu variant is also generalizable in another control genetic background, as we observed dramatic cortical organoid size increase at both day 14 and day 18 after installing the same heterozygous PTEN variant into the control cell line Clay (Figures S1H, 5A, and 5B). Next, we replicated the organoid size enlargement observed in Apex PTEN correction vs. control Chap organoids (Figures 5C and 5D) in another ASD genetic background, Arch CTNNB1 correction. We subjected both control iPSC lines Chap and Clay as well as the Arch CTNNB1 correction iPSC line to the same organoid culture protocol. Intriguingly, Arch CTNNB1 correction organoids exhibited enlarged organoid size at day 18 compared with Chap (Figure 5E). At both day 18 and week 8, Arch CTNNB1 correction cortical organoids grew bigger than Clay (Figures 5F and 5H), but at week 4, no organoid size differences were observed Arch CTNNB1 correction and Clay (Figure 5G), which is most likely because of the cell line sensitivity differences to timing of organoid development. In summary, we have observed ASD genetic background leads to enlarged organoid size by using two cortical organoid models derived from control lines and two models derived from ASD genetic background lines.

We next differentiated control iPSC cell line Chap and ASD genetic background iPSC line Arch CTNNB1 correction into 2D NPC (Figure S2) and performed bulk RNA-seq at P5, P6, and P7. We performed GSEA by comparing Arch CTNNB1 correction and Chap across three different passages. Indeed, we found that compared to Chap, ASD genetic background Arch CTNNB1 correction NPCs exhibited downregulated genes involved in GO terms related to neurogenesis (forebrain generation of neuron), neuron development (axon development, cerebral cortex development), microtubule-related (microtubule polymerization) and upregulated genes involved in GO terms related to embryonic development (embryonic organ morphogenesis), and GO terms related to protein folding; all of these GO terms were also found in 2D NPC Apex PTEN correction vs. Chap (Figures S1I and 2A).

To test whether this new ASD genetic background dysregulates neurogenesis in organoids at the transcriptome level, as we saw in the Apex genetic background, we further cultured Clay and Arch CTNNB1 correction organoids to 10 weeks and performed scRNA-seq analysis with the same procedures that we processed the original Chap and Apex isogenic PTEN panel cortical organoids with. We integrated week 10 scRNA-seq datasets for two control background organoids, Chap and Clay, and two ASD genetic background organoids, Apex PTEN correction and Arch CTNNB1 correction, as well as the previously published fetal brain datasets.³⁴ Cell types were annotated for each cluster on the basis of known marker gene expression, and organoids derived from all four of these lines produced a variety of NPC and neuronal subtypes found in the human fetal brain (Figures 5I and 5J). Focusing exclusively on NPC subtypes including cycling progenitors, vRGs, tRGs, oRGs, and IPCs, we performed pairwise differential gene expression analysis among the two control background and two ASD background cortical organoids by using GSEA. Preselected GO terms related to neurogenesis, gliogenesis, mitochondria, and ribosome biogenesis as well as cytoplasmic translation, which already appeared in both bulk RNA-seq and scRNA-seq on the isogenic PTEN panel, were used to study the effect of ASD genetic background on the gene expression profiles. Of note, in this analysis, with two control and two ASD genetic backgrounds, we were able to study the effects of ASD genetic background of two different biological ASD background replicates, while at the same time, each ASD genetic background was compared to two independent control biological replicates. For GSEA of Apex PTEN correction vs. Chap, genes involved in neurogenesis were downregulated in vRGs, IPCs, and oRGs, whereas genes related to ribosome biogenesis and cytoplasmic translation were upregulated in these NPC subtypes (Figure 5K). GSEA comparing Apex PTEN correction with Clay, another control genetic background, revealed that again genes involved in neurogenesis were downregulated, but this time only in vRGs and oRGs (Figure 5L), suggesting that vRGs and oRGs were reproducibly affected by this ASD genetic background irrespective of the control genetic background used for gene expression comparison. Next, we studied the effects of another ASD genetic background by comparing Arch CTNNB1 correction and Chap and found again downregulation of genes involved in neurogenesis in vRGs, IPCs, and oRGs (Figure 5M). These effects in this ASD genetic background were further replicated by comparing Arch CTNNB1 correction and Clay, where we found that genes involved in neurogenesis were downregulated in vRGs, IPCs, and oRGs, as well as cycling progenitors (Figure 5N). In addition, genes related to gliogenesis were affected by the ASD genetic background in both IPCs and oRGs (Figure 5N). Our results were further strengthened by the fact that no neurogenesis-related GO terms were enriched when comparing gene expression profiles of the two control organoids, Clay and Chap (Figure 5P), or the two ASD genetic backgrounds, though we observed gliogenesis-related GO terms were seen in vRGs, IPCs, and oRGs comparing ASD backgrounds (Figure 5O). Overall, these results identify vRGs and oRGs as reproducibly affected NPC subtypes by ASD genetic backgrounds, where we observe genes involved in neurogenesis dysregulated by the ASD genetic backgrounds.

Genetic variants and differentially expressed genes in the ASD genetic background

To gain insight into the ASD background-dependent effect of the PTEN p.Ile135Leu variant on the PTEN activity in 2D NPCs as well as the differential gene expression findings (Figure 1), we first examined the WES data for the ASD line Apex.¹⁶ There were 68 genes with either stop-loss or stop-gain variants in Apex, and none of these overlapped with the 1,045 ASD Simons Foundation Autism Research Initiative (SFARI) risk genes (SFARI-Gene_genes_05-05-2022 release, https://gene.sfari.org) (Figure S7A, Table S2A) or the 339 genes in the PI3K-AKT pathway derived from Molecular Signature Database (MsigDB, Table S2B). In addition, 380 genes with SNVs were identified in WES data from Apex. Of these, two overlapped with both the PI3K-AKT pathway and SFARI risk genes: PTEN, as expected, and MET (Figure S7B). An additional three SNVs overlapped with PI3K-AKT pathway genes, while 14 genes overlapped with 1,045 SFARI risk genes but were not in the PI3K-AKT pathway. Next, we determined whether any of the differentially expressed genes (DEGs), identified comparing autism and control, are part of both the PI3K-AKT pathway and the SFARI autism risk genes. Out of 1,632 genes downregulated Apex vs. Chap (with p(adjust) < 0.05, and log2 fold change ≥ 0.6, Table S2C), 33 genes belong to the PI3K-AKT pathway, four of which are also SFARI ASD risk genes: FGF14, MET, NTRK1, and RELN (Figure 6A). All of these genes are outstanding candidates that alone or together modulate the activity of PTEN in the ASD background. FGF14 is a gene that expresses a brain-specific fibroblast growth factor, and variants in this gene are responsible for an autosomal-dominant form of cerebellar ataxia.⁴² MET is a tyrosine kinase important for stem cell growth and epithelial-mesenchymal interactions.⁴³ MET signaling disruption was also reported in ASD postmortem cerebral cortex.⁴⁴ NTRK1 is a neurotrophin (NGF) receptor that is critical for neuronal development. Variants in this gene result in cognitive disability.⁴⁵ RELN is a ligand for a number of receptors during development and adulthood, while mutations result in defects in neuronal migration and neurogenesis.⁴⁶ An additional 116 downregulated genes are also SFARI ASD risk genes. Among 1,729 genes that are upregulated Apex vs. Chap, 59 genes belonged to the PI3K-AKT pathway, five of which were SFARI ASD risk genes: ITGB3, LAMB1, NTRK2, PRKCA, and TEK (Figure 6B, Table S2D). Likewise, these genes are outstanding candidates that alone or together modulate the activity of PTEN in the ASD background. ITGB3 codes for integrin beta-3, a cell adhesion molecule that has important synaptic functions during development.⁴⁷ LAMB1 codes for laminin beta-1, and variants in this gene cause cobblestone lissencephaly.⁴⁸ NTRK2 is a neurotrophin (NT-3) receptor that is critical for neuronal development. Variants in this gene result in obesity⁴⁹ and mood disorders.⁵⁰ PRKCA codes for protein kinase C-alpha⁵¹ that has been associated with ASD,⁵² while TEK codes for TIE2 and is expressed almost exclusively in the vascular system.⁵³

Figure 6.

Differentially expressed genes in the ASD genetic background

(A) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes, and 1,632 downregulated genes in Apex.

(B) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes, and 1,729 upregulated genes in Apex.

(C) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes, and 1,497 downregulated genes in Apex PTEN correction.

(D) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes, and 1,610 upregulated genes in Apex PTEN correction. Four ASD genes that are dysregulated in Apex NPC line (Bolded gene symbols in A and B) were rescued upon correcting PTEN variant in Apex.

(E) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes, and 1,131 downregulated genes in Arch CTNNB1 correction.

(F) Venn diagram for the identification of the overlapping genes among 339 PI3K/AKT pathway gene sets, 1,045 SFARI ASD genes and 1,881 upregulated genes in Arch CTNNB1 correction. Arch CTNNB1 correction NPCs contain same sets of ASD risk genes that are also dysregulated in Apex or in Apex PTEN correction lines (bolded genes in E and F). All DEGs described here refer to dysregulated genes in 2D NPCs. Venn diagrams were plotted with https://bioinformatics.psb.ugent.be/webtools/Venn/.

We then asked whether any of the above mentioned nine ASD risk genes that are also part of the PI3K-AKT pathway and dysregulated in Apex vs. Chap would be rescued after correcting the PTEN p.Ile135Leu variant by performing similar differential expression analysis between Apex PTEN correction and Chap NPCs. Excitingly, downregulated ASD risk genes that are part of the PI3K-AKT pathway including MET and NTRK1 were rescued after correcting the PTEN variant back to wild type (Figures 6A and 6C, Tables S2C and S2E). In addition, ITGB3 and NTRK2, genes that were upregulated ASD risk genes belonging to the PI3K-AKT pathway were also rescued after correcting the PTEN variant (Figures 6B and 6D, Tables S2D and S2F). This provides additional evidence that PTEN p.Ile135Leu interacts with other ASD risk genes that are part of the PI3K-AKT pathway, which may contribute to the ASD genetic background-dependent effect of the PTEN variant on PTEN activity in 2D NPCs.

We further asked whether the PI3K-AKT pathway ASD risk gene dysregulation observed in the Apex ASD genetic background was replicated in the Arch CTNNB1 correction ASD genetic background. We identified 1,131 downregulated genes in NPCs Arch CTNNB1 correction vs. Chap; 25 of these genes belong to PI3K-AKT pathway, and two of these are also SFARI ASD risk genes, including LAMA1 and THBS1 (Figure 6E, Table S2G). In addition, we identified 1,881 upregulated genes; 63 genes were part of the PI3K-AKT pathway, and seven were also ASD risk genes, namely ITGB3, LAMB1, MET, NTRK2, PRKCA, RELN, and TEK (Figure 6F, Table S2H). Remarkably, nearly all of the dysregulated PI3K-AKT pathway ASD risk genes in the Arch ASD genetic background were also dysregulated in the Apex ASD genetic background (Figures 6A, 6B, 6E, and 6F), and LAMA1 dysregulation was unique to the Arch genetic background (Figure 6E). These results suggest that two different ASD genetic backgrounds converge on similar dysregulation of PI3K-AKT pathway ASD risk genes.

In summary, these results suggest that ASD backgrounds contain dysregulated ASD risk genes that are related to the PI3K/AKT pathway, and the PTEN p.Ile135Leu variant interacts with a number of such dysregulated genes, providing an explanation for the ASD-background-dependent effect of the PTEN p.Ile135Leu variant on PTEN activity. Lastly, we found that two different ASD genetic backgrounds converge on dysregulating ASD risk genes that were part of the PI3K-AKT pathway.

PTEN p.Ile135Leu variant accelerates the neuronal maturation of upper layer neurons in ASD genetic background

To determine whether the PTEN p.Ile135Leu variant affects neuronal maturation, we performed pseudotime analysis with Monocle3³⁵ on the week 10 and week 21 scRNA-seq datasets. We used SOX2⁺ radial glia cells as the root cells to plot the pseudotime trajectory (Figures S8A, S8B, S8E, and S8F) and then plotted the distribution of deep layer excitatory neurons and upper layer excitatory neurons on the basis of developmental pseudotime. At both week 10 and week 21, the PTEN WT/Ile135Leu organoids displayed accelerated upper layer neuron maturation only in the ASD genetic background (Figures S8C and S8D). Interestingly, PTEN WT/Ile135Leu cortical organoids displayed decelerated upper layer neuron maturation in the control genetic background at week 21 but not at week 10. Deep layer neuron maturation remained unaffected by the PTEN p.Ile135Leu variant in both control and ASD genetic background (Figures S8G and S8H). These results suggest that the PTEN p.Ile135Leu variant accelerated upper layer neuron maturation in concert with increased NPC subtype proliferation, leading to overproduction of more mature upper layer neurons in the ASD genetic background.

PTEN p.Ile135Leu variant and ASD genetic backgrounds dysregulate synaptic transmission genes

To examine whether PTEN WT/Ile135Leu and ASD genetic background affected genes important for neuronal function in addition to accelerating neuronal maturation, we performed GSEA on neuron subtypes including deep and upper layer excitatory neurons as well as interneurons. Neurons from the PTEN WT/Ile135Leu control background organoids displayed downregulated genes enriched for GO terms related to synaptic function (synaptic signaling, regulation of transsynaptic signaling, and regulation of synaptic plasticity, Figure 7A). We also found that genes important for action potential were downregulated in PTEN WT/Ile135Leu control background organoids in both deep and upper layer excitatory neurons as well as interneurons. Genes enriched for GO terms important for protein synthesis such as “cytoplasmic translation,” “ribosome biogenesis,” and “ATP metabolic process” were upregulated by the PTEN variant in the control genetic background (Figure 7A).

Figure 7.

PTEN WT/Ile135Leu variant and ASD genetic background dysregulate synaptic transmission

(A and B) GSEA for the effect of PTEN WT/Ile135Leu variant in control genetic background and in ASD genetic background in different neuronal subtypes.

(C–E) GSEA for the effect of ASD genetic background in different neuronal subtypes with PTEN WT/WT group, with PTEN WT/Ile135Leu group, and with PTEN KO/KO group. GO terms were preselected for visualization and p values were adjusted with Benjamini-Hochberg correction. EN_deep indicates deep layer excitatory neurons; EN_upper indicates upper layer excitatory neurons; IN indicates interneurons.

By contrast, in the ASD genetic background, genes enriched for GO terms related to synaptic function (synaptic signaling, regulation of transsynaptic signaling, and regulation of synaptic plasticity) were all upregulated in PTEN WT/Ile135Leu organoids in deep layer and upper layer neurons as well as interneurons, which is the opposite direction of the effect of the same PTEN variant on the control genetic background (Figures 7A, 7B, and S6E). GO terms related to action potential were also enriched by genes upregulated by the PTEN WT/Ile135Leu variant in the ASD genetic background but only in upper layer neurons at week 21. Interestingly, we found that GO terms related to “cytoplasmic translation” were enriched by genes upregulated at week 10 in deep layer neurons, but were enriched by downregulated genes at week 21, whereas GO terms related to mitochondria (electron transport chain, ATP synthesis electron transport chain) were enriched by genes downregulated in upper layer excitatory neurons and interneurons but were enriched by upregulated genes in the deep layer excitatory neurons (Figure 7B). All these results highlight the spatial and temporal regulation of gene expression in mature neuronal populations by PTEN WT/Ile135Leu in the ASD genetic background.

The reversed effect of the ASD PTEN variant on synaptic function in the ASD genetic background versus its effect on the control genetic background motivated us to ask whether the ASD genetic background had a strong effect on neuronal GO terms. In the PTEN WT/WT group, neurons from organoids with the ASD genetic background displayed downregulated genes enriched for GO terms related to synaptic function (synaptic signaling, regulation of transsynaptic signaling, regulation of synaptic plasticity, and action potential) irrespective of neuronal subtypes (Figures 7C and S6E), suggesting that the ASD genetic background indeed affects genes important for neuronal function. Neurons from organoids with the ASD genetic background displayed upregulated genes enriched for GO terms related to cytoplasmic translation and ribosome biogenesis (Figures 7C and S6E). Additional analyses of the effect of the ASD genetic background were carried out. This includes similar GSEA on week 10 deep and upper layer neurons as well as interneurons in three two-way comparisons: one ASD genetic background (Apex PTEN correction) to a different control cell line, Clay; another ASD genetic background (Arch CTNNB1 correction) to the control line Chap; and Arch CTNNB1 correction line to the control line Clay. GSEA results for all the additional comparisons point to dysregulated synaptic function due to the ASD genetic backgrounds (Figures S9A, S9B, S9D, and S9E). Of note, the effect on dysregulating genes involved in synaptic function was not seen when performing GSEA for comparison between the two control lines, Clay and Chap (Figure S9F), though we noticed dysregulated synaptic function genes when performing GSEA between the two ASD genetic background lines, Arch CTNNB1 correction and Apex PTEN correction, in the deep layer neuron (Figure S9C). These results further strengthen our findings that ASD genetic backgrounds dysregulate genes involved in synaptic function.

In the PTEN WT/Ile135Leu group, neurons from organoids in the ASD genetic background activated genes associated with GO terms related to synaptic function in the excitatory upper layer neurons for both week 10 and week 21 but not in deep layer excitatory neurons. In addition, we also observed a temporal effect of the ASD background on genes enriched for synaptic signaling in interneurons from organoids, as they were activated at week 10 and downregulated at week 21 (Figure 7D). In the PTEN KO/KO group, ASD genetic background downregulated genes enriched for GO terms related to synaptic function in both upper layer neurons and interneurons but not deep layer neurons (Figure 7E).

Thus, similar to what we found in NPC subclasses in organoid models, genes enriched for neuronal GO terms such as synaptic function are dysregulated in a spatial- and temporal-dependent interplay between the ASD PTEN p.Ile135Leu variant and the ASD genetic background.

Discussion

ASDs are heterogeneous genetic disorders that result from the complex interplay between genetic and environmental factors.⁵⁴ The number of identified ASD risk genes has grown over the past decade, especially that of those with high heritability and large effect size. One such gene, PTEN, is a major contributor to ASD risk, especially in the 20% of ASD-affected individuals with macrocephaly.^5,6 Although PTEN is a well-known tumor suppressor, approximately 15% of ASD-affected individuals with early brain overgrowth possess variants in PTEN.⁷ The role of ASD-affected individuals’ genetic background effect in modifying the risk of ASD has been more challenging to study, and experimental evidence for the role of genetic background in ASD development remains limited. A recent study⁵⁵ identified an asynchronous effect of ASD risk gene variants on neuron production by modeling the same ASD risk gene variant on different control genetic backgrounds in cortical organoids, providing evidence that genetic background is an important factor when studying the role of a specific ASD risk gene disruption. Here, we have directly studied the cellular impact of a PTEN variant found in an ASD-affected individual in both control and ASD backgrounds by performing bidirectional CRISPR-Cas9 genome editing. We used an ASD-affected-individual-derived iPSC line possessing a point mutation of PTEN to produce an isogenic line with the correction of the PTEN variant, both in the ASD background. We examined a second ASD background after correcting a CTNNB1 variant within that line. Similarly, we studied two control iPSCs and the matched line with the induction of the same PTEN variant in the control backgrounds. These isogenic PTEN iPSC panels enabled us to study the effect of the same ASD variant in both control and ASD genetic background and the effect of ASD genetic background with different PTEN genotypes.

Our studies demonstrated that both the heterozygous PTEN c.403A>C variant and ASD genetic backgrounds dysregulated genes important for cellular features of neurogenesis in induced, stable NPCs across three different passages and in self-organizing cortical organoids across NPC subtypes at both week 10 and week 21. The PTEN p.Ile135Leu variant resulted in increased proliferation in stable NPCs in both control and ASD backgrounds. However, the genetic backgrounds buffered the effect of the PTEN p.Ile135Leu variant differentially. For example, comparing the effect of the same heterozygous PTEN c.403A>C variant in both control and ASD genetic backgrounds, we observed reversed effects of PTEN p.Ile135Leu variant on general GO terms from gene expression profiles related to neuron development and neurogenesis. The ASD genetic background also displayed profound effects dependent upon the PTEN genotype. In the PTEN WT/WT group, NPCs in the ASD background displayed more rapid proliferation compared to NPCs from the control genetic background, and the ASD genetic background resulted in the downregulation of gene expression profiles related to neurogenesis. However, in the PTEN WT/Ile135Leu group, there was a reversal of the effect of ASD genetic background on general GO terms from gene expression profiles related to neurogenesis, which was upregulated. In the PTEN KO/KO group, the effect of ASD genetic background on neurogenesis was non-uniform and asynchronous. Notably, most of these effects were entirely reproducible in three independent cell culture replicates over several passages in NPCs or in several organoids. These findings suggest that the PTEN variants such as the heterozygous PTEN c.403A>C variant found in ASD can be modified by the ASD genetic background to influence neurogenesis, while stronger variants, such as PTEN KO/KO, are less sensitive to background modification of neurogenesis phenotypes, consistent with clinical findings in PTEN-related cancer-prone or ASD syndromes.¹⁰ In addition, similar results were obtained with two independent control and two different ASD genetic backgrounds, namely on proliferation and gene dysregulation.

Our study additionally identified, unexpectedly, that the PTEN p.Ile135Leu variant impairs the canonical PTEN activity in the ASD genetic background but not in the control genetic background in 2D NPCs. Consistent with this biochemical readout, we also identified that the same PTEN variant led to overproduction of NPC subtypes as well as neuronal subtypes only in the ASD genetic background but not in the control genetic background in 3D cortical organoid model. Such an ASD-genetic-background-dependent effect of the PTEN variant on neurogenesis may be a direct effect of the genetic background on PTEN activity. We analyzed genetic variants (stop-gain and SNVs) and DEGs identified in the ASD line. Several ASD risk genes including MET, FGF14, NTRK1, NTRK2, ITGB3, LAMB1, RELN, and PRKCA were identified that were all part of the PI3K/AKT pathway. In addition, correcting the PTEN variant was able to rescue the dysregulation of several ASD risk genes, including NTRK1, NTRK2, MET, and ITGB3. This is consistent with the oligogenic inheritance model for the ASD pathology,^56,57,58 where several genetic or epigenetic variants in the ASD genetic background serve as the genetic modifiers for a key ASD risk gene, in our case, PTEN. Thus, our findings suggest one plausible molecular mechanism in which these additional dysregulated ASD risk genes in the ASD genetic background may affect neurogenesis through modifying the effect of PTEN p.Ile135Leu variant on the PTEN canonical activity. Intriguingly, Similar ASD risk genes that belong to the PI3K/AKT pathway were also dysregulated in another ASD genetic background, in which a CTNNB1 stop-gain variant was corrected via CRISPR-Cas9 genome editing. This result suggests potential convergent PI3K/AKT pathway dysregulation in the ASD genetic backgrounds.

The neuron overproduction finding in our PTEN panel cortical organoid model is also consistent with previous clinical findings, which identified 67% more neurons were present in the prefrontal cortex in postmortem brains from ASD-affected individuals with brain overgrowth compared to healthy controls.⁵⁹ Our studies further indicated that the ASD PTEN p.Ile135Leu variant led to this neuron overproduction by both accelerating neuronal maturation and increasing NPC subtype proliferation only on the ASD genetic background. This clearly indicates the advantages of studying ASD candidate gene variants in ASD genetic backgrounds using cortical organoid models.

In conclusion, our study provides direct evidence that ASD genetic background contributes to ASD pathology by working in concert with the ASD risk gene that displays a variant found in ASD-affected individuals. Our study demonstrates that ASD genetic background modifies the effect of the PTEN p.Ile135Leu variant on gene expression profiles related to neurogenesis, resulting in accelerated neural maturation and elevated IPC and oRG production, which ultimately leads to the overproduction of the neuronal subtypes including deep and upper layer neurons in an ASD-background-dependent manner in cortical organoids. By transcriptomic profiling the isogenic PTEN panel cortical organoids via scRNA-seq, our study revealed dysregulated gene expression profiles related to neurogenesis and gliogenesis in NPC subtypes, consistent with findings from bulk RNA-seq in 2D NPCs. We also uncovered dysregulated expression profiles related to synaptic function in neuronal subtypes resulting from both the PTEN p.Ile135Leu variant and the ASD genetic backgrounds. Of note, our study revealed that both PTEN p.Ile135Leu variant and an ASD genetic background reproducibly affected transcriptomic profiles related to both neurogenesis and gliogenesis in NPC subtypes. It is possible that the timing for the neurogenesis to gliogenesis switch is also impacted. A recent study⁶⁰ identified that EGFR⁺ cells dramatically increased after gestational week 20, marking the onset of human gliogenesis, which was within the timing of the organoid models studied here. Future work on teasing out the effect of the ASD risk gene and ASD genetic background on the neuron to glia switch will yield additional insights on their roles in ASD pathology.

Published: April 24, 2023

Data and code availability

Bulk RNA-seq and scRNA-seq raw data have been deposited at GEO with the following accession numbers: GEO: GSE214323, GSE214422, GSE221882, and GSE221923. All original code for scRNA analysis for this paper has been deposited and is publicly available under the following link: https://github.com/shuaifu93/codes-for-PTEN-manuscript.