TDP-43 Overexpression in Human Cells Reproduces Key ALS Pathology and Gene Changes

Scientists have built a human cell line that overproduces TDP-43, the misbehaving protein at the center of most cases of amyotrophic lateral sclerosis, and used it to map in detail what happens inside cells when the protein accumulates. The new model, described in the journal Discover Neuroscience, is a stable, tetracycline-inducible system derived from HEK293 cells. When the researchers switched on full-length human TDP-43, the cells developed three hallmark features of TDP-43 pathology: the protein piled up over time, an increasing share of it shifted from the nucleus into the cytoplasm, and the cells became measurably less viable. RNA sequencing then revealed sweeping changes in gene expression and RNA splicing, and a comparison with a large dataset from ALS patients showed that 64 percent of the altered genes in the cell model were also dysregulated in patient tissue.

TDP-43, encoded by the TARDBP gene, is a widely expressed, highly conserved RNA-binding protein that moderates RNA splicing, transport, stability, and transcriptional regulation. In disease, it mislocalizes from the nucleus to the cytoplasm and forms toxic, ubiquitin-positive aggregates in more than 90 percent of ALS cases and roughly half of frontotemporal lobar degeneration cases. The same pathology has also been documented in Huntington’s disease, Alzheimer’s disease, and limbic-predominant age-related TDP-43 encephalopathy. Despite this central role, the field has struggled with a basic question: whether disease is driven by a loss of normal TDP-43 function in the nucleus, a toxic gain of function in the cytoplasm, or both. The new overexpression model is designed to probe the gain-of-function side of that question, complementing the many depletion-based systems that dominate the literature.

The team, led by Samantha A. DeSando and colleagues, created the model using the Flp-In T-REx 293 system, which inserts a codon-optimized TARDBP gene into the chromosome under a hybrid CMV/TetO2 promoter. Because the construct is integrated rather than transiently transfected, and because the expressed protein carries no fluorescent tag such as GFP, the researchers avoided artifacts that can complicate interpretation, including tag-driven changes in protein localization and transfection-related toxicity. Dose-ranging studies identified 25 nanograms per milliliter of tetracycline as the lowest concentration producing sufficient overexpression, giving the system tunable control over how much TDP-43 the cells make.

Induction with that tetracycline dose for 24 to 48 hours produced a robust and time-dependent rise in TDP-43 protein in the engineered cells, while an empty-vector control line showed no increase. Normalized protein levels rose 2.1-fold at 24 hours and 3.1-fold at 48 hours, both statistically significant. Western blot analysis of separated cellular compartments showed that TDP-43 accumulated in both the nucleus and the cytoplasm, but the cytoplasm gained more, pushing the cytoplasmic-to-nuclear ratio up 1.5-fold at 24 hours and 2.4-fold at 48 hours, the latter reaching statistical significance. That shift matters because an overabundance of TDP-43 together with an increased cytoplasmic-to-nuclear ratio is a primary pathological marker in ALS and frontotemporal dementia patients.

The accumulating protein also hurt the cells. In an MTS cytotoxicity assay, 48 hours of TDP-43 induction increased cell death by 26.9 percent relative to controls, a highly significant effect that coincided with the peak in protein levels. The empty-vector line showed no viability change after tetracycline treatment, and neither line was affected at 24 hours, tying the toxicity specifically to TDP-43 accumulation rather than the inducing drug. The authors note that the cell death observed with the modest induction dose was modest in absolute terms, but that cytotoxicity continues to climb beyond 48 hours, which they say allows the assay to be refined for specific screening needs. They position this phenotypic readout as a practical test bed for evaluating small molecules or other interventions aimed at reducing TDP-43-driven cytotoxicity or restoring normal protein distribution.

Transcriptomic profiling exposed the downstream consequences. Comparing the TDP-43 line to the empty-vector control at zero, 24, and 48 hours of induction, the researchers identified 433 upregulated and 516 downregulated genes at 24 hours, expanding to 1,233 upregulated and 1,053 downregulated genes at 48 hours. Of the total significant differentially expressed genes, 754, or 79 percent of the 24-hour set, appeared at both timepoints. Because cytotoxicity emerged at 48 hours, the team prioritized the 24-hour data as a window into early, direct effects of increased TDP-43 rather than secondary consequences of dying cells. Gene Ontology analysis of that early timepoint flagged 97 significantly enriched terms, with the top pathways including nervous system development, chemical synaptic transmission, organism development, and cell adhesion, all consistent with TDP-43’s known functions. The number of enriched terms grew to 129 by 48 hours, with 56 shared between the two timepoints.

Splicing, one of TDP-43’s core regulatory duties, was also disrupted. SpliceSeq analysis counted 664 significant alternative splicing events across 608 unique genes at 24 hours, rising to 1,537 events across 1,329 genes at 48 hours. Applying a stricter filter for changes of at least 25 percent in the spliced fraction left 278 events at 24 hours, spanning exon skipping and intron retention. Three genes drew particular attention: PTS, SAFB, and TBC1D1, each showing exon skipping. The spliced PTS variant would produce a severely truncated protein predicted to be inactive, since the active site is lost. The SAFB variant involves an in-frame deletion of 37 amino acids, and the TBC1D1 variant would truncate a protein critical to glucose metabolism. The authors caution that exon numbering can differ across genome annotations and that nucleotide sequence is the most reliable way to pinpoint splicing events.

To test whether the model reflects human disease, the researchers compared its 24-hour gene expression changes against RNA-seq data from 1,579 samples in the New York Genome Center ALS Consortium dataset, analyzing ten tissue types separately and pooling the significant genes. The combined patient list contained 22,028 genes altered in at least one tissue. Strikingly, 610 of the 949 differentially expressed genes in the induced cell line, or 64 percent, were also altered in ALS patient samples. Among a focused subset of 13 genes highlighted from the cell data, seven were significantly dysregulated in at least one patient tissue as well. NUP85, a nuclear pore protein, was downregulated in both the cell line and patient cerebellum. SREBF2, a master regulator of cholesterol metabolism, was downregulated in the cell line and in five patient tissues, including cerebellum, medial motor cortex, and cervical, lumbar, and thoracic spinal cord. VAMP5 was upregulated in the cell line and in patient cervical and lumbar spinal cord, though it moved in the opposite direction in cortical tissues. WDR41, involved in autophagy regulation, was dysregulated in the cell line and in seven patient tissue types.

Not every comparison aligned neatly. CDC23 and DKC1 were upregulated in patient hippocampus but downregulated in the overexpressing cells, and PTS, downregulated and mis-spliced in the cell line, was upregulated in six patient tissue types. The authors suggest that such directional differences may still reflect overall dysfunction of the same genes, but they flag the discrepancy as a question for future work, noting that a knockdown model might better recapitulate the patient profile for some targets. ELP1, RBM12, and NUP88 showed no significant changes in the patient cohort despite clear dysregulation in the cell line, lowering their priority. The team validated the RNA-seq findings by qPCR, confirming the expression changes for the gene subset and the splicing changes for PTS, SAFB, and TBC1D1. The full dataset has been deposited in the NCBI Gene Expression Omnibus under accession GSE329332.

The authors are careful about the model’s limits. HEK293 cells are not neurons and do not express known ALS-associated splicing markers such as STMN2 and UNC13A, so the researchers recommend evaluating the transcriptomic changes in a neuronal model of TDP-43 overexpression as a next step. They also suggest examining phosphorylated TDP-43 species, detergent-insoluble protein, and C-terminal fragments in future characterizations. Still, they argue the platform fills a real gap: existing overexpression models often lack comprehensive RNA-seq data benchmarked against patient datasets, and overexpression systems are better suited than depletion models for testing drugs designed to target TDP-43 directly. With its tunable induction, tag-free protein, scalable assays, and substantial transcriptomic overlap with patient tissue, the cell line offers ALS researchers a practical tool for hunting disease-relevant genes and screening therapeutic candidates aimed at taming one of neurodegeneration’s most consequential proteins.

Subject of Research: Development and transcriptomic characterization of an inducible TDP-43 overexpression cell model for ALS research

Article Title: TDP-43 overexpression induces cellular dysfunction and ALS-associated transcriptional changes

Article References: DeSando, S. A., McClay, K., Banks, A. N., Cheung, M., Reitz, A. B., & Freeman, K. B. (2026). TDP-43 overexpression induces cellular dysfunction and ALS-associated transcriptional changes. Discover Neuroscience, 21(1), Article 95. https://doi.org/10.1186/s13064-026-00324-y

Image Credits: AI Generated

DOI: 10.1186/s13064-026-00324-y

Keywords: TDP-43, ALS, FTLD, RNA sequencing, HEK293, transcriptomics, alternative splicing, cytotoxicity, nucleocytoplasmic transport, gene expression, proteinopathy, drug discovery