https://www.cell.com/iscience/fulltext/S2589-0042(26)02293-5
chatGPT(5.5paid):
The uploaded paper is a 2026 iScience review, “Cryptic transcription in aging”, by Wu et al. Its central argument is that loss of transcriptional fidelity during ageing is partly caused by normally silent intragenic transcription start sites (cTSSs) becoming active as chromatin regulation deteriorates.
Summary
1. What is cryptic transcription?
Normally, RNA polymerase II begins transcription at the canonical promoter near the 5′ end of a gene. Cryptic transcription instead starts from promoter-like sequences buried within the gene body.
The resulting RNAs therefore generally lack part of the normal 5′ transcript and may produce:
- truncated proteins,
- proteins lacking regulatory domains,
- frameshifted “neoproteins,”
- unstable RNAs degraded by the RNA exosome,
- or transcripts that interfere with transcription of the normal gene.
The authors argue that increased cryptic transcription is an evolutionarily conserved feature of ageing, first established in yeast and C. elegans and now increasingly demonstrated in mammalian stem cells, fibroblasts and tissues.
2. The proposed epigenetic mechanism
The most important mechanistic model in the paper is shown in Figure 2 on page 3.
In young cells:
SETD2 → H3K36me3 → KDM5B + DNMT3B → suppression of intragenic promoters
More specifically:
- SETD2 deposits H3K36me3 over actively transcribed gene bodies.
- H3K36me3 recruits KDM5B, which removes promoter-associated H3K4me3 from gene bodies.
- H3K36me3 also recruits DNMT3B, maintaining intragenic CpG methylation.
- H3K36me3 promotes FACT recruitment and proper nucleosome replacement behind RNA polymerase II.
Together these make the gene body relatively resistant to inappropriate transcription initiation.
During ageing/senescence:
↓H3K36me3 → ↓KDM5B/DNMT3B → ↑H3K4me3 + ↓DNA methylation
At the same time:
AP-1 → p300/CBP → ↑H3K27ac
The resulting chromatin increasingly resembles a genuine promoter. RNA polymerase II can then initiate transcription from inside the gene.
So the paper’s model can be condensed to:
Ageing converts parts of gene bodies from elongation-only chromatin into promoter-like chromatin.
3. Enhancer-to-promoter conversion
One particularly interesting aspect is that cTSSs may not simply appear at random exposed DNA.
Sen et al.'s data, heavily discussed in the review, suggest that some cTSSs correspond to pre-existing intragenic enhancers.
In proliferating cells they exhibit enhancer-like bidirectional transcription. During senescence they acquire:
- ↑ H3K4me3,
- ↑ H3K27ac,
- p300/CBP,
- H2A.Z,
- ↓ H3K36me3,
- ↓ DNA methylation,
and start behaving more like promoters.
This is illustrated particularly clearly in Figure 1 on page 2 and Figure 2 on page 3.
4. Evidence that cryptic transcription rises with age
The review brings together several lines of evidence.
McCauley et al. found increased cryptic transcription in:
- aged mouse haematopoietic stem cells,
- expanded human mesenchymal stem cells,
- activated neural stem cells.
Sen et al., using PRO-seq/PRO-cap, found senescence-specific intragenic TSSs in human IMR90 fibroblasts.
ChRO-cap in mouse liver found more unique cTSSs in old animals.
And reanalysis of 25 ageing/senescence RNA-seq datasets found increased cryptic transcription in 21 of 25 datasets.
Interestingly, it was not clearly increased in some premature-ageing disorders such as Werner and Rett syndromes, suggesting it is not simply a universal consequence of anything resembling ageing.
5. Long genes seem particularly vulnerable
An interesting recurring observation is that genes showing age-related cryptic transcription tend to be relatively long.
The paper notes this in:
- yeast,
- worms,
- mouse HSCs,
- human MSCs.
Genes containing cTSSs tend to have:
- fewer but longer introns,
- shorter exons,
- shorter UTRs.
One proposed reason is that long intronic regions increase the amount of time RNA polymerase spends traversing a gene, giving more opportunity for chromatin maintenance to fail.
However, the authors appropriately acknowledge that some apparent enrichment could arise from detection bias: longer genes simply contain more sequence in which a cryptic TSS can be detected.
Biological consequences
The review proposes several ways cryptic transcription could damage ageing cells.
Abnormal proteins
As shown in Figure 3 on page 5, cryptic RNAs may be translated.
Depending on where transcription begins, this could produce proteins that:
- lack N-terminal domains,
- lack regulatory sequences,
- are frameshifted,
- misfold.
These products might burden:
- molecular chaperones,
- the proteasome,
- autophagy.
Thus cryptic transcription potentially links epigenetic ageing to loss of proteostasis.
Transcriptional interference
A cryptic transcript can potentially interfere with transcription of the legitimate gene.
This mechanism is very well established in yeast but remains much less firmly demonstrated in mammals.
Transcriptional noise
The authors therefore place cryptic transcription within a broader phenomenon:
ageing → deterioration of chromatin control → deterioration of transcriptional fidelity → noisier transcriptome.
The affected genes are enriched in pathways involving:
- signalling,
- cell adhesion,
- cell communication,
- differentiation,
which potentially links cryptic transcription to senescence and the SASP.
What is novel about this paper?
Importantly, this is a review rather than a new experimental study, so its novelty is principally conceptual synthesis rather than new data.
I think there are four particularly useful aspects.
1. It turns several epigenetic observations into one coherent causal chain
The strongest contribution is the integrated model:
**SETD2↓
→ H3K36me3↓
→ KDM5B recruitment↓ + DNMT3B recruitment↓
→ H3K4me3↑ + DNA methylation↓
- AP-1/p300/CBP → H3K27ac↑
→ enhancer-to-promoter conversion
→ cryptic transcription.**
Most individual components were already known. The review presents them convincingly as a single ageing mechanism.
2. It frames cryptic transcription as an ageing phenotype in mammals
The older literature largely treated transcriptional fidelity as a yeast mechanism.
The review makes the case that the phenomenon should now be considered relevant to:
- mammalian stem-cell ageing,
- cellular senescence,
- tissue ageing,
- possibly cancer.
That is an important shift in emphasis.
3. The enhancer-to-promoter model is particularly interesting
Rather than thinking of cryptic transcription simply as random transcription from exposed DNA, the paper emphasizes something more mechanistically specific:
ageing may convert existing intragenic regulatory elements into promoters.
That makes cryptic transcription potentially much more structured and biologically reproducible than simple transcriptional “noise.”
4. It highlights locus-specific epigenetic therapy
The review makes an interesting therapeutic distinction.
Broad epigenetic drugs may actually worsen cryptic transcription. For example:
- DNMT inhibitors,
- HDAC inhibitors,
can activate cryptic antisense transcripts, including potentially oncogenic CAPTs.
The authors therefore propose locus-specific approaches such as:
- SETD2-dCas9 to restore H3K36me3,
- dCas9-HDAC3 to remove H3K27ac.
That is a logical extension of the mechanism, although still highly speculative therapeutically.
Critique
The mechanistic framework is compelling, but I think the paper sometimes moves too quickly from association → mechanism → ageing causality.
1. The biggest weakness is lack of evidence that cryptic transcription causes mammalian ageing
There is good evidence for:
ageing/senescence → increased cryptic transcription.
There is substantially weaker evidence for:
increased cryptic transcription → ageing.
For example, reducing SETD2 causes:
- increased cryptic transcription,
- impaired stem-cell function,
- reduced proliferation.
But SETD2 and H3K36me3 regulate many processes besides cryptic transcription.
Therefore:
SETD2 loss → dysfunction
does not prove
SETD2 loss → cryptic transcription → dysfunction.
The paper itself concedes this at several points, but its concluding language occasionally becomes stronger than the evidence warrants.
This is the most important unresolved causal question.
2. H3K36me3 loss has many effects besides cryptic initiation
The paper focuses heavily on:
H3K36me3↓ → cryptic transcription.
But H3K36me3 is deeply involved in:
- transcription elongation,
- co-transcriptional splicing,
- DNA methylation,
- DNA repair,
- nucleosome dynamics,
- FACT recruitment.
Therefore cryptic transcription could partly be an indicator of failing gene-body chromatin rather than the principal pathological output.
A better causal model may be:
age-associated gene-body chromatin dysfunction
→ cryptic initiation
→ altered elongation
→ altered splicing
→ DNA methylation changes
→ transcriptional noise
rather than cryptic transcription being the dominant downstream cause.
3. The protein-level consequences remain surprisingly speculative
Figure 3 gives considerable emphasis to:
cryptic RNA → aberrant protein → proteostasis stress.
There is some supporting ribosome-profiling evidence, but the paper does not establish that a large fraction of age-associated cryptic RNAs actually produce stable proteins in aged tissues.
Many could instead be:
- rapidly degraded,
- poorly exported,
- poorly translated,
- destroyed by RNA surveillance.
So the pathway
cryptic transcription → aberrant proteome
is plausible but much less firmly demonstrated than
cryptic transcription → aberrant RNA production.
4. Detection is technically difficult
This is a major issue that the review acknowledges well.
Ordinary RNA-seq cannot reliably distinguish:
- genuine cryptic initiation,
- alternative promoter usage,
- degradation products,
- RNA processing intermediates,
- alternative splicing.
PRO-cap and related methods identify transcription initiation much more convincingly, but measure nascent RNA rather than mature stable transcripts.
The authors’ suggestion of combining:
PRO-cap + long-read RNA sequencing
is therefore particularly sensible.
Until this becomes common, the quantitative extent of biologically meaningful cryptic transcription during ageing remains uncertain.
5. “Cryptic” may sometimes be too pejorative
Some apparently cryptic promoters may actually represent physiologically regulated alternative transcription start sites.
A previously weak intragenic promoter becoming more active with age does not automatically mean that it is pathological.
The key experiment is therefore not simply:
Does this transcript increase with age?
but:
Does selectively preventing this particular transcript improve cellular function without changing the canonical gene?
That sort of experiment is still largely missing.
6. The sex differences are intriguing but presently inconsistent
The review reports:
- female bias in aged mouse liver,
- female-associated effects in mouse neural stem cells,
- male-specific association in one human fibroblast dataset.
Those observations are interesting, but presently they are too inconsistent across species and tissues to support a general mechanistic model.
The discussion of sex-specific cryptic transcription is therefore more hypothesis-generating than established biology.
A particularly important implication for ageing biology
I think the most important conceptual point in this review is actually somewhat broader than the authors’ stated model.
It suggests that age-associated transcriptional dysfunction is not necessarily just a failure to make enough RNA from legitimate promoters.
Instead ageing may progressively reduce the signal-to-noise ratio of transcription:
[
\text{Transcriptional quality}
\approx
\frac{\text{correct full-length transcripts}}
{\text{correct transcripts + erroneous transcripts}}
]
That means an ageing cell could have apparently normal or even elevated total RNA synthesis while nevertheless suffering a substantial decline in informational fidelity.
This could connect several ageing observations:
loss of H3K36me3 / chromatin organisation
↓
cryptic TSS activation + altered elongation + abnormal splicing
↓
increased isoform diversity / NMD / aberrant RNAs
↓
loss of coherent gene-expression programmes
↓
declining cellular function.
That broader model is, in my view, more compelling than treating cryptic transcription alone as an independent hallmark of ageing.
Bottom line
This is a useful and timely review rather than a breakthrough experimental paper. Its strongest contribution is assembling a credible mechanistic chain connecting SETD2/H3K36me3 loss, DNA hypomethylation, promoter-like histone marks and enhancer-to-promoter conversion to increased intragenic transcription during mammalian ageing. Its major weakness is that the evidence currently establishes cryptic transcription much more strongly as a marker/consequence of ageing chromatin dysfunction than as a demonstrated driver of organismal ageing.
I am with chatGPT on this. I think it is a result of chromatin dysfunction rather than a driver of ageing. (back to my acetylation hobbyhorse)