iPSC Models Are Rewriting Neurodegeneration Research, but Aging Still Sets the Limits
Patient-derived induced pluripotent stem cells have become one of the most important attempts to make human neurodegeneration experimentally tractable. In a review published in the Journal of Clinical Investigation, Wang, Elia, Liu, and Finkbeiner argue that the appeal of these models is straightforward: they preserve the donor’s genetic background while providing human neurons and glia that can be studied at scale. That combination matters because most major neurodegenerative diseases do not arise from a single mutation with a clean mechanistic chain. They emerge from polygenic risk, environmental exposures, and the slow biology of aging.
The central problem is not lack of effort. It is biology.
Phase III trial failure rates in neurodegenerative disease remain above 95%, and only a small fraction of preclinical findings ever reach regulatory approval. The review ties that failure rate to a familiar translational gap: animal models and simplified systems often miss the human-specific molecular features that drive disease, especially when the causal architecture is diffuse rather than monogenic. Human neurons are not interchangeable with rodent neurons, and several disease-relevant pathways simply do not map across species with enough fidelity to predict therapeutic benefit.
Human Genetics Explains Why Some Models Fail Before They Start
The review treats neurodegeneration as a set of biologically distinct problems rather than a single disease category. In Alzheimer disease, Parkinson disease, Lewy body dementia, amyotrophic lateral sclerosis, and frontotemporal dementia, most cases are sporadic. They arise from the accumulation of many variants of modest effect, plus stochastic and environmental inputs. A model that starts from one engineered mutation may faithfully reproduce a narrow molecular defect, yet still miss the broader network failure that actually unfolds in patients.
Familial mutations remain crucial, especially when they point to convergent mechanisms. Variants in APP, PSEN1, and PSEN2 drive inherited Alzheimer disease; SNCA, LRRK2, and PRKN shape inherited Parkinson disease; and TARDBP, SOD1, and C9orf72 are central in ALS-FTD biology. Yet those high-penetrance mutations account for only a minority of cases, leaving the larger clinical burden to polygenic risk and disease biology that is much harder to engineer from scratch. In that setting, patient-derived iPSCs offer a different kind of value: they bring the entire genetic architecture into the dish.
The review is blunt about the advantage. A patient line can retain both known and unknown contributors to risk, including rare variants, common GWAS hits, and interactions that would never be captured by editing a single locus onto a generic background.
Disease Signal Emerges at Different Depths in Different Disorders
Not every assay window is equally informative. For monogenic disease, the strongest signal may appear close to the initiating molecular event, where the effect of the causal variant is still large and not yet buffered by homeostatic compensation. For polygenic disease, the opposite is often true. Small upstream effects accumulate downstream, so convergent cellular endpoints can be more sensitive than assays aimed at the first molecular perturbation.
This logic has immediate implications for experimental design. A model of familial ALS built around a pathogenic mutation may be most revealing when investigators look at early splicing or proteostasis defects. A sporadic Alzheimer model, by contrast, may require readouts such as synaptic dysfunction, lysosomal stress, or transcriptional convergence before the disease biology becomes visible. The review frames this as a matching problem: the closer the model reflects the architecture of the disease, the more likely it is to produce a meaningful phenotype.
Human-specific splicing biology illustrates the point. TDP-43 targets such as UNC13A and STMN2 are poorly conserved in rodents, even though the genes themselves exist. A rodent may carry the ortholog, but not the same regulatory sequence logic. For diseases in which aberrant RNA processing is central, that difference can erase the very phenotype researchers need to measure.
Reprogramming Solves Access, Then Creates a New Problem
Reprogramming somatic cells into pluripotency solved a major access problem. It made patient-specific neurons renewable and experimentally flexible. It also created a new obstacle: the process resets epigenetic age. Cells that began as aged somatic tissue are pushed back toward a fetal-like state, and with that reset comes the loss of the long biological history in which late-onset neurodegeneration develops.
For disorders that usually emerge after decades of accumulated damage, this is a serious constraint. Age is not just a risk factor; it is part of the disease context. Epigenetic drift, oxidative stress, altered proteostasis, mitochondrial decline, and changes in chromatin state all shape whether a neuron remains stable or tips into pathology. A young iPSC-derived neuron may have the genotype of an affected patient without the cellular memory that made the patient vulnerable in the first place.
The review treats this as the field’s defining challenge. Recreating disease without recreating age may produce elegant biology that still fails to answer the clinically important question: why does the disorder begin now, and not twenty years earlier?
Approximating Age Without Losing Human Relevance
Several strategies try to restore aging features in iPSC-derived neurons. Others aim to bypass the reset entirely by using direct conversion approaches that preserve the donor’s native epigenetic age. Each path has strengths and trade-offs. Induced aging can help expose age-dependent phenotypes, but it may also introduce artificial stressors that are not identical to natural aging. Direct conversion can preserve age-related signatures, but often at the cost of flexibility and expansion capacity.
The review favors a pragmatic approach rather than a single preferred platform. Model choice should follow the biological question. If the goal is to define a convergent pathway downstream of many risk variants, an iPSC system may be ideal. If the goal is to capture the first aging-dependent step that unmasks a sporadic phenotype, a different cellular strategy may be more informative.
One sentence captures the field’s pressure point: aging can be removed in one step, but disease onset cannot.
What These Models Can Deliver, and What They Still Cannot
The review does not present iPSCs as a universal answer. It argues for precision in model selection. Two-dimensional monocultures, mixed cultures, and organoids each expand the range of biological questions that can be asked, but each also introduces trade-offs in maturity, cellular composition, throughput, and interpretability. The more complex the system becomes, the closer it may come to human tissue architecture, yet the harder it becomes to isolate mechanism.
For neurodegeneration research, that trade-off is unavoidable. Human iPSC systems can capture patient genetics, human-specific molecular regulation, and, in some settings, disease-relevant phenotypes that would be invisible elsewhere. They can also mislead if investigators ask them to model a decades-long aging process after resetting cells to an embryonic state. The practical lesson is not that one platform wins. It is that the biological question must dictate the model, not the other way around.
The review closes by proposing exactly that framework: match the model system to the disease architecture, the target cell type, and the role of aging in phenotype emergence. For a field where translational attrition remains severe, that discipline may be as important as any new assay.
What iPSC models cannot yet reproduce with full fidelity is the long human aging trajectory that turns genetic risk into late-onset neurodegeneration.