Transcriptome of iPSC-derived neuronal cells reveals a module of co-expressed genes consistently associated with autism spectrum disorder

The study addresses how heterogeneous genetic risks in autism spectrum disorder (ASD) may converge on shared molecular pathways detectable at the transcriptomic level. Prior work shows ASD-associated genes converge on pathways related to immune response, neurotransmission, and neurodevelopment, but postmortem adult/child brain studies may miss transient prenatal dysregulation. iPSC-derived neuronal models enable examination of early neurodevelopmental transcriptional changes. The authors aim to characterize transcriptomic profiles of iPSC-derived neural progenitor cells (NPC) and post-mitotic neurons from a cohort enriched for high-functioning, normocephalic ASD males, compare these profiles to fetal and adult brain expression to infer developmental context, identify differentially expressed genes and co-expression modules associated with ASD, and evaluate concordance with prior iPSC and postmortem brain studies to pinpoint robust ASD-associated gene networks that could serve as biomarkers or therapeutic targets.

Whole-transcriptome analyses of ASD postmortem brain have repeatedly reported alterations in immune response, neurotransmission, and neurodevelopment pathways. However, postnatal brain tissue may not capture prenatal, developmentally restricted dysregulation relevant to ASD pathogenesis. iPSC-derived neuronal studies have examined relatively small idiopathic ASD cohorts, including studies focused on macrocephalic patients (a minority of ASD cases) and an analysis enriched for low-functioning normocephalic patients. Given ASD’s genetic and phenotypic heterogeneity, larger and more diverse cohorts and endophenotypes need investigation. It also remains unclear how findings from in vitro iPSC-derived neurons relate to and are consistent with postmortem brain transcriptomes across studies and developmental stages.

- Cohort: Six male ASD individuals (five high-functioning; one low-functioning, IQ < 70) negative for Fragile X; six male controls without neurodevelopmental diagnoses. Ethical approval obtained (protocol 1.133.486). - Genomics: Peripheral blood DNA assayed by array-CGH (Agilent 4x180K) for CNVs and whole-exome sequencing (Nextera Rapid Capture Exome; Illumina HiSeq 2500). Variant prioritization per Supplementary pipeline. - iPSC generation: Reprogrammed SHED (stem cells from human exfoliated deciduous teeth) using retroviral SOX2, c-MYC, OCT4, KLF4. Two to three clones per donor. Pluripotency and tri-lineage differentiation verified by immunocytochemistry. - Neural differentiation: NPC generated from iPSC and maintained in DMEM/F12 with N2, B27, FGF (20 ng/ml), EGF (20 ng/ml). Some NPC transduced with SYN::EGFP lentivector (GFP under Synapsin promoter). Neuronal differentiation for 4 weeks in Neurobasal medium with 1 µM retinoic acid. GFP+ neurons enriched by FACS (SYN::EGFP). - Immunocytochemistry: Assayed cell identity markers—NPC: SOX2, Nestin; neurons: CTIP2, MAP2; GFP detection—using standard fixation, primary/secondary antibodies, and confocal imaging. - RNA-seq: Libraries from NPC (29 lines), heterogeneous neuronal cultures (16 lines), and sorted GFP+ neurons (7 lines) prepared with TruSeq RNA kit, sequenced (100 bp paired-end, HiSeq 2500). Alignment with TopHat2 (hg19), gene counts with HTSeq. Public data: GEO GSE142670. Cell-type deconvolution used to estimate neuronal proportions; four neuronal samples with <51% neurons excluded. Expressed genes defined as FPKM ≥ 1 in >50% samples (NPC: 13,818 genes; neurons: 15,026 genes). Batch/cell-type proportion normalization with RUVSeq; parameters tuned to remove neuron proportion correlations while improving replicate clustering. - Developmental and regional identity inference: Compared sample transcriptomes to BrainSpan Atlas across 4 post-conception weeks to 60 years using machine learning (Stein et al.) to predict developmental period and regional identity. - Differential expression and network analysis: Differential expression tested with dream model (variancePartition) at FDR ≤ 0.05. WGCNA performed separately for NPC and neurons to derive co-expression modules; module eigengenes correlated with ASD status, batch, and neuronal proportion. Functional enrichment via DAVID, Ingenuity Pathway Analysis, and cameraPR; protein-protein interaction enrichment via STRING. Module preservation assessed against BrainSpan fetal brain. Code and data available at https://github.com/griesik/ASDiPSCTranscriptome. - Neuronal morphology: FACS-purified GFP+ neurons co-plated with non-transduced heterogeneous neurons under non-mixed or mixed (opposite group) conditions. After 72 h, cells fixed and imaged (InCell Analyzer 2200). Morphometrics quantified with ImageJ Neurphology plugin; statistics via geepack (R). - Proteomics: Mass spectrometry-based proteomics on NPC from 6 clones (3 ASD, 3 controls); 2D reverse-phase UPLC (Waters) coupled to Synapt G2-Si. Data processed in Progenesis QI 2.1; identification/quantification against Uniprot human proteome, algorithms per Cassoli et al. - Genetic enrichment: Enrichment of ASD de novo variants (from compiled sources) and SFARI/ID gene sets tested in module gene sets using two-sided Fisher’s exact test with FDR correction; only protein-coding genes considered. - Module overlap: WGCNA userListEnrichment used to compare modules to those reported in prior iPSC and postmortem ASD transcriptome studies, defining a robust synapse-associated module overlapping with MNeul-turquoise across fetal/neonatal brain and iPSC neuron datasets.

- ASD vs control transcriptomes in iPSC-derived cells revealed: • In NPC: dysregulation of a co-expressed gene module enriched for protein synthesis/translation-related functions. • In neurons: dysregulation of a synapse/neurotransmission-related module and a separate translation-related module. - Proteomics in NPC identified protein-level changes providing potential molecular links between NPC and neuronal modules dysregulated in ASD. - Cross-study comparison showed the synapse-related module is consistently upregulated in iPSC-derived neurons (which resemble fetal brain expression) but downregulated in ASD postmortem brain tissue, indicating robust disease association with developmental directionality (opposite dysregulation across development). - iPSC-derived neuronal cells’ transcriptomes resembled fetal brain stages critical to ASD pathophysiology, supporting model relevance. - Genetic characterization identified: • Patient F2613: de novo 17p13.3 duplication (ABR, BHLHA9, TUSC5, YWHAE, CRK, MYO1C) plus a de novo predicted-damaging missense in DOCK1 (ASD-candidate). • Patient F2688: compound heterozygous missense variants in RELN (high-confidence ASD gene) causing reduced secretion/signaling; additional de novo loss-of-function splice-site mutation in CACNA1H. • Remaining patients lacked de novo LoF in LoF-intolerant brain-expressed genes but carried inherited rare potentially damaging missense variants in SFARI genes. - RNA-seq metrics and filtering: NPC n=29 lines; neurons n=16 (4 excluded for <51% neuronal content); expressed gene sets: NPC 13,818; neurons 15,026; analyses controlled for neuronal proportion (RUVSeq).

Findings support the hypothesis that diverse ASD genetic risks converge on specific molecular programs detectable in early neurodevelopment. The consistent identification of a synapse/neurotransmission co-expression module across independent iPSC-derived neuron datasets, and its inverse regulation relative to postmortem adult/child ASD brains, suggests a developmentally dynamic dysregulation: early upregulation (fetal-like iPSC neurons) transitioning to downregulation in later life. This pattern strengthens the module’s relevance to ASD pathophysiology and highlights timing as a critical factor for interpreting transcriptomic signatures. The detection of translation/protein synthesis modules in both NPC and neurons implies broad impacts on proteostasis that may intersect with synaptic function. Proteomic data corroborate transcriptomic module relationships at the protein level, reinforcing biological plausibility. Together, these results clarify how in vitro models relate to in vivo brain development and help reconcile discrepancies between iPSC-derived and postmortem studies, pointing to a robust synaptic gene network as a candidate biomarker and therapeutic target in ASD.

The study demonstrates that iPSC-derived NPCs and neurons from predominantly high-functioning, normocephalic ASD individuals exhibit consistent dysregulation of co-expression modules related to protein synthesis and synapse/neurotransmission. Cross-cohort and cross-tissue comparisons identify a synaptic module robustly associated with ASD that is upregulated in fetal-like iPSC neurons but downregulated in postmortem brains, indicating developmental stage-specific directionality. This network emerges as a promising biomarker and potential therapeutic target for ASD. Future research should validate this synaptic signature longitudinally across developmental stages, expand cohorts to encompass diverse ASD endophenotypes, integrate multi-omics (transcriptome/proteome) in matched samples, and experimentally test interventions targeting components of the synaptic network.

- Cohort size is modest (n=6 ASD males), limiting statistical power and generalizability. - ASD group is enriched for high-functioning, normocephalic individuals; findings may not extend to other ASD endophenotypes (e.g., macrocephaly, low-functioning). - iPSC-derived neuronal cultures are heterogeneous; although neuronal proportion was estimated and adjusted, variability remains and four samples with <51% neurons were excluded. - In vitro models approximate fetal-like states and may not recapitulate all in vivo developmental processes or cell-type interactions present in the brain. - Cross-study comparisons rely on heterogeneous datasets, platforms, and analysis pipelines, which may introduce confounds despite observed consistencies.