https://www.nature.com/articles/s41586-026-10877-x#
chatGPT(5.6paid):
Overall assessment
This is an unusually ambitious and technically impressive paper. The authors maintain human cortical organoids for more than five years and show that they continue to undergo coordinated transcriptional, epigenetic, cellular and functional changes resembling aspects of human prenatal and postnatal cortical development.
The strongest conclusion is that human neural cells possess a remarkably durable, largely cell-intrinsic developmental timer that continues running outside the body. The evidence for this is persuasive.
The more dramatic language—that organoids “record” and “recall” time—is a useful conceptual framing, but the underlying molecular mechanism of that memory is not identified. Nor does the paper demonstrate that five-year organoids are equivalent to a five-year-old human cortex in any comprehensive biological sense.
Summary
1. Long-term organoids follow human developmental trajectories
The authors analysed 110 individual organoids spanning 15 days to five years, comprising 424,720 single cells. The newly generated dataset includes 34 organoids aged six months to five years, integrated with 76 younger organoids from previous studies.
When the organoid cells were mapped onto reference datasets from human fetal and postnatal cortex:
- Early organoids mapped mainly to first-trimester brain.
- Three-to-six-month organoids mapped predominantly to the second trimester.
- Organoids older than approximately one year increasingly acquired late-prenatal and postnatal-like transcriptional states.
- Cell-type-specific maturation modules derived from the human cortex changed sequentially with time in culture.
- Astrocytes and glial progenitors showed particularly convincing concordance: about 80% of the genes changing significantly with age in both organoids and human tissue changed in the same direction.
The cultures also acquired structural evidence of oligodendrocyte maturation and myelination.
2. The organoid methylome tracks elapsed culture time
Whole-genome bisulfite sequencing from three months to five years showed:
- Increasing global CpG methylation followed by relative stabilization.
- Age-associated differential methylation at cortical developmental genes and regulatory regions.
- Progressive non-CpG, particularly CpA, methylation—a recognised feature of maturing neurons.
- Developmentally appropriate changes in neuronal and non-neuronal methylation regions previously identified in human cortex.
Two established methylation clocks correlated strongly with time in culture:
- Horvath pan-tissue clock: approximately (r=0.9)
- Cortex-specific clock: approximately (r=0.9)
The median absolute errors were roughly 7 and 20 months, respectively. A fetal-brain clock showed a weaker, borderline association.
Methylation at solo-WCGW sites, used as an indicator of cumulative cell division, remained relatively stable. The authors therefore argue that the clock-like changes are not simply a consequence of continuing proliferation.
3. Conventional culture loses neurons over time
Under the original CDM4 culture conditions:
- Astrocytic and glial populations remained abundant and continued maturing.
- Neurons progressively declined as a fraction of recovered single cells.
- Nevertheless, NeuN-positive neurons remained detectable at 2, 3, 4 and 5.8 years.
- SATB2-positive excitatory neurons were detected at 5.8 years.
Thus, conventional organoids can preserve some neurons for nearly six years, but the cultures become increasingly glial and do not maintain a stable cortical cellular composition.
4. Activity-permissive medium improves excitatory-neuron survival and maturation
The authors introduced an “activity-permissive medium” or APM, based on BrainPhys with GlutaMax, from day 70.
Compared with CDM4, APM produced:
- More SATB2-positive callosal projection neurons.
- More FOS-positive, apparently active excitatory neurons.
- Greater neurite length and branching complexity.
- Higher synaptic density at one year.
- An increase in the proportion of synapses located on dendritic spines from 25% to 52%.
- Enrichment of synaptic and neuronal-maturation gene-expression programmes.
- Preferential retention of excitatory projection neurons for at least 18 months.
- Sustained spontaneous extracellular electrical activity for at least two years.
The electrophysiological activity was sensitive to glutamatergic receptor inhibition, indicating that it represented synaptically mediated network activity rather than merely autonomous cellular firing.
APM did not simply preserve every population equally: it changed progenitor proliferation and favoured excitatory-neuron lineages while reducing the relative representation of several glial and inhibitory populations.
5. Old progenitors retain temporal identity
The most conceptually interesting experiments used “chimeroids.”
The authors dissociated nine-month-old human organoids and allowed the cells to reaggregate. When old cells were cultured alone, they predominantly produced cell types characteristic of their late developmental state.
When old cells were mixed with very young neural progenitors:
- The young environment shifted old progenitors towards a somewhat younger molecular state.
- The old cells restarted excitatory neurogenesis.
- But they did not restart the developmental sequence from the beginning.
- Instead, within two weeks they produced late-born callosal projection neurons, glial progenitors and astrocytes.
- Young cells in the same chimeroid continued to produce early progenitors and early-born neuronal types.
Approximately 49% of the derivatives attributed to the old cells in the heterochronic chimeroids were classified as callosal projection neurons, compared with around 0.5% in old-only reaggregates and 1.1% in conventional nine-month organoids.
Parallel experiments with embryonic mouse cortical progenitors supported the general proposition that older progenitors retain temporally advanced fate tendencies when placed alongside younger cells.
The interpretation is that progenitors preserve a memory of the developmental stages they have already passed through: a young environment can restore neurogenic competence but does not reset their temporal identity to zero.
What is genuinely novel?
1. The length and multimodal depth of the experiment
Previous studies cultured cortical organoids for approximately two years, but this paper extends the timeline beyond five years and combines:
- Single-cell transcriptomics
- Whole-genome DNA methylation
- Immunohistochemistry
- Electron microscopy
- Expansion microscopy
- Neuronal reconstruction
- Multielectrode electrophysiology
- Heterochronic cell-mixing experiments
The novelty is therefore not simply “an organoid survived for five years.” It is the demonstration that multiple biological modalities continue changing coherently across that period.
2. A close correspondence between culture time and methylation age
The near-linear relationship between chronological culture time and predicted DNA-methylation age is striking. It suggests that at least some methylation clocks measure autonomous cellular or developmental processes rather than exposure to the systemic environment, behaviour, inflammation or whole-body physiology.
This is important for interpreting epigenetic clocks: part of their signal may reflect an internally generated measure of elapsed developmental time.
3. Cell-type-specific postnatal-like maturation
Earlier long-term organoid studies relied substantially on bulk measurements. This study resolves how individual neural and glial populations progress, showing that apparent organoid ageing cannot be explained entirely by changes in cellular composition.
The concordance in glia, neuronal CpA methylation and maturation-associated multicellular programmes is particularly valuable.
4. The APM culture method
The APM protocol appears to overcome part of the long-standing problem of excitatory-neuron attrition in long-term organoids. The improvement is supported by molecular, morphological, ultrastructural and electrophysiological measures rather than a single neuronal marker.
This is a practical advance that may make experiments on later human cortical development substantially more feasible.
5. Experimental evidence for persistent temporal memory
Temporal competence transitions in neural progenitors were already known, especially in model organisms. The distinctive contribution here is showing in long-lived human organoids that:
- old progenitors retain neurogenic plasticity;
- young environmental signals can reactivate neurogenesis;
- but reactivated old progenitors rapidly generate temporally appropriate late progeny.
That combination distinguishes temporal identity from irreversible loss of competence.
Critique
1. “Ageing” is repeatedly conflated with development and maturation
The dominant process studied is developmental maturation, not ageing in the gerontological sense.
The five-year organoids move from fetal-like towards early-postnatal-like states. The paper does not establish the emergence of familiar ageing phenotypes such as:
- senescence or SASP accumulation;
- proteostatic decline;
- age-associated mitochondrial dysfunction;
- somatic mutation accumulation;
- loss of epigenetic information;
- impaired autophagy;
- chronic inflammatory signalling;
- neurodegenerative vulnerability.
Consequently, phrases such as “tissue ageing,” “epigenomic ageing” and “old progenitors” need careful interpretation. A nine-month organoid progenitor is developmentally older than a 15-day progenitor, but it is not biologically aged in the usual adult-ageing sense.
A more precise formulation would be that organoids preserve developmental time and early postnatal maturation.
2. Clock agreement does not prove equivalence to an age-matched brain
Epigenetic clocks reduce complex methylation states to a numerical estimate. A high correlation with time establishes clock-like progression but does not show that the entire organoid methylome has reached the corresponding in-vivo state.
Important qualifications include:
- The clocks were not necessarily developed or calibrated for organoids.
- A correlation can be high even when the slope or absolute calibration differs.
- The cortex clock had a median absolute error of about 20 months, substantial relative to the ages studied.
- The fetal-brain clock was much less convincing: (r=0.54), (P=0.056).
- Some CpG sites may measure generic time-dependent methylation rather than functional cortical maturity.
The paper does provide regional methylation and transcriptional evidence beyond clock scores, which helps considerably, but “the clock reads five years” should not be translated into “this is equivalent to a normal five-year-old cortex.”
3. The time series is largely cross-sectional
Different organoids are destroyed at different timepoints. The study therefore observes a population time series, not continuous ageing of the same organoid.
This leaves room for:
- between-batch variation;
- selective survival of particular organoids;
- selective loss of vulnerable cell types;
- changing recoverability during dissociation;
- clonal selection within the cultures.
The authors use several lines, batches and mixed-effects analyses, but a cross-sectional design cannot completely distinguish within-organoid maturation from selection among cells and organoids.
4. The oldest cultures are not compositionally normal cortex
Under conventional conditions, neurons become rare and glia dominate. Demonstrating a residual SATB2-positive population at 5.8 years is meaningful, but it is not equivalent to maintaining a structurally normal cortical neuronal network for that period.
The strongest molecular agreement at the oldest ages may therefore be disproportionately driven by robust glial populations. This matters because the paper’s title could be read as applying equally to all components of the organoid.
The APM data improve the neuronal result, but those cultures were functionally studied for about two years, not the full five-to-six-year interval.
5. APM changes developmental selection, not simply neuronal “health”
APM is described as activity permissive, but the experiments do not isolate activity as the causal factor. BrainPhys differs from CDM4 in several chemical and metabolic respects, and GlutaMax changes glutamine stability and ammonia production.
Moreover, APM:
- increases cycling apical radial glia;
- alters interneuron-progenitor proliferation;
- changes cellular proportions;
- preferentially expands or retains excitatory lineages.
Therefore, the increase in excitatory neurons could arise from altered lineage production, selective survival, altered maturation or all three. A decisive experiment would manipulate electrical activity within the same medium—for example with chronic silencing or stimulation—and ask whether the survival advantage disappears or increases.
6. The temporal-memory experiment is powerful but not definitive mechanistically
The chimeroid experiment demonstrates persistent fate bias, but “memory of elapsed time” remains a phenomenological description.
Several alternatives or contributors remain possible:
- stable chromatin state;
- retained transcription-factor expression;
- DNA methylation;
- accumulated protein or RNA;
- metabolic state;
- lineage selection during dissociation and reaggregation;
- selective survival or expansion of a late-competence subpopulation.
The study does not identify which of these stores temporal information. Perturbing candidate chromatin regulators, methylation machinery or metabolic timing pathways would be needed to reveal the clock mechanism.
7. The chimeroid experiment contains donor and reaggregation confounds
Young and old cells were derived from different genetic backgrounds so that their origins could be distinguished. That is technically convenient, but age and genotype are then partially confounded.
The design would be stronger with:
- reciprocal young/old assignments across donor lines;
- isogenic barcoding of cells from the same line;
- multiple independent old and young donor combinations;
- lineage tracing before and after mixing.
Dissociation itself is also a strong injury and selection event. The striking 49% CPN result may partly reflect which old cells survive, reaggregate or are transcriptionally classified, rather than every old progenitor changing fate.
8. Some orthogonal validation is statistically modest
The scRNA-seq assignment of old-cell progeny to callosal projection neurons is striking, but the corresponding SATB2 immunostaining comparison used only three organoids per condition and gave (P=0.0519). That is suggestive rather than conventionally significant.
SATB2 is also not, on its own, a complete demonstration of mature functional callosal projection-neuron identity. Stronger confirmation could include:
- a broader CPN marker panel;
- patch-clamp physiology;
- axonal projection preference;
- longer follow-up;
- lineage-resolved morphology.
9. Missing systemic and anatomical influences limit biological equivalence
The organoids lack or incompletely model:
- vasculature and blood–brain barrier;
- normal microglia and immune interactions;
- meningeal and vascular signals;
- endocrine and circadian inputs;
- thalamic and sensory afferents;
- organised long-range cortical outputs;
- normal cortical lamination and geometry.
The study therefore shows that much developmental timing is autonomous, not that the systemic environment is dispensable for normal maturation. The authors appropriately recognise that sensory activity and non-neural tissues should be incorporated in future models.
10. Long-gene transcription and splicing remain underexplored
Given the prolonged maturation of human neurons, this dataset could be particularly informative about whether ageing in culture changes:
- transcription of long neuronal genes;
- alternative splicing;
- intron retention;
- spliceosomal fidelity;
- isoform switching;
- nonsense-mediated decay.
The study focuses primarily on gene-level expression modules and methylation. It does not deeply analyse isoform-level maturation or the possibility that residual long-term neurons acquire splicing defects. That is a notable missed opportunity, especially because human neuronal maturation is strongly dependent on long genes and carefully timed isoform transitions.
Bottom line
The paper convincingly demonstrates that human cortical organoids contain an autonomous, multi-year developmental programme. Transcription, DNA methylation, cell identity, morphology and electrophysiology continue to mature in a temporally ordered fashion, and older progenitors retain a persistent temporal state even when exposed to a young environment.
Its most important conceptual result is not that organoids reproduce “brain ageing,” but that human developmental time is biologically encoded within neural cells and can proceed for years without a body.
The central unanswered question is what the clock physically is. DNA methylation is an excellent record of elapsed time here, but the paper does not establish whether it is the causal timer, a downstream memory of transcriptional/chromatin state, or merely one readout of a deeper metabolic or regulatory process.