https://www.pnas.org/doi/epdf/10.1073/pnas.2515183123
chatGPT:
Summary
The paper presents a platform called the Transcriptional Rejuvenation Discovery Platform (TRDP) for finding single transcription-factor perturbations that can push aged or damaged cells back toward a younger transcriptional and functional state. The authors use late-passage human neonatal dermal fibroblasts as a model of cellular aging, comparing them with early-passage fibroblasts, then computationally prioritising transcription factors likely to explain the old-versus-young gene-expression differences. They then perturb the top candidates using Perturb-seq with CRISPR activation and CRISPR inhibition.
The central result is that several single transcription-factor perturbations partially reverse aging-associated transcriptional patterns. The strongest validated candidates were:
| Perturbation | Direction | Claimed rejuvenating effects |
|---|---|---|
| E2F3 | overexpression / CRISPRa | Increased proliferation; reversed some aged fibroblast gene-expression changes |
| EZH2 | overexpression / CRISPRa | Improved mitochondrial activity; reversed fibroblast and liver aging markers |
| STAT3 | repression / CRISPRi | Improved proteostasis and lysosome-related phenotypes |
| ZFX | repression / CRISPRi | Reversed aged fibroblast expression and some functional phenotypes |
In fibroblasts, these perturbations increased Ki67 positivity, improved proteasome gene expression/activity, increased mitochondrial gene expression and TMRE signal, reduced lysosome staining, and lowered some senescence-associated markers such as p21/CDKN1A, TIMP1, and TIMP2. The effects were described as moving late-passage cells toward a middle-passage-like state rather than fully restoring them to a young state.
The most striking in vivo experiment is AAV8-mediated EZH2 overexpression in aged mouse liver. After three weeks, aged mice showed partial reversal of liver aging-associated gene-expression patterns, reduced steatosis, reduced fibrosis, and improved glucose tolerance. The authors argue that this shows a single transcription-factor perturbation can affect not only cultured cells but also an aged tissue.
The paper also compares the transcriptional effects of these perturbations with other aging/rejuvenation datasets, including human skin fibroblast aging and mouse heterochronic parabiosis datasets. The authors suggest that the perturbations converge on shared downstream transcriptional modules, especially involving cell cycle, proteostasis, mitochondrial function, stress response, and tissue identity.
What is novel?
The main novelty is not simply that EZH2, E2F3, STAT3, or ZFX affect aging-associated biology, because several of these factors already have known roles in proliferation, chromatin regulation, inflammation, stemness, fibrosis, or cancer. The novelty is the systematic discovery framework: combining old-versus-young transcriptomics, predicted transcription-factor regulatory modules, Perturb-seq, CRISPRa/CRISPRi, and downstream phenotyping to search for single-factor rejuvenating perturbations.
A second novelty is the emphasis on single transcription-factor perturbations rather than multi-factor partial reprogramming with Yamanaka factors. The study argues that single-factor perturbations can produce some effects comparable to Yamanaka factor overexpression in fibroblast assays, but without obvious dedifferentiation in the tested conditions.
A third important novel claim is that EZH2 overexpression alone can improve aged liver phenotypes in vivo over a short period. The reported reversal of steatosis, fibrosis, glucose intolerance, and age-associated liver gene expression makes this the most translationally interesting finding in the paper.
Critique
The biggest limitation is that the primary discovery model is replicative aging of fibroblasts, where loss of proliferation is a dominant feature. That creates a risk that the screen preferentially identifies factors that restore proliferation or cell-cycle activity rather than deeper, generalizable rejuvenation. The authors partly address this by analysing cells within cell-cycle phases and by showing changes in proteostasis, mitochondria, and lysosomes, but the enrichment of cell-cycle modules remains a major interpretive issue.
The term “rejuvenation” may be too strong for some of the data. The perturbations improve selected aging-associated markers, but they do not reset all hallmarks. The paper reports little or no significant change in DNA damage, telomere length, methylation age, or fibroblast identity markers. This suggests partial functional remodeling rather than comprehensive cellular rejuvenation.
The in vivo EZH2 result is promising but short-term. Three weeks of liver-specific overexpression improved several phenotypes, but this does not establish durability, safety, lifespan effect, or long-term cancer risk. EZH2 is a chromatin regulator associated in other contexts with cancer biology, so the lack of short-term cancer-like transcriptional signatures is reassuring but not sufficient. Longer studies with tumour surveillance, dose-response testing, reversibility, and multi-tissue assessment are essential.
The paper’s cancer-safety argument is useful but incomplete. Comparing transcriptomic signatures with cancer models does not exclude increased long-term tumour risk, clonal expansion, epigenetic instability, or pre-malignant effects that emerge only after months. This is especially relevant because the interventions include factors such as E2F3 and EZH2, which can promote proliferation or chromatin states linked to malignancy in other contexts.
Another weakness is that the liver experiment focuses mainly on EZH2, so the broader claim that TRDP identifies generally useful rejuvenating transcription-factor perturbations still needs more in vivo validation. It would be stronger if E2F3, STAT3 repression, and ZFX repression were tested in additional tissues, including post-mitotic or low-turnover cell types where cell-cycle restoration is not the central issue.
Bottom line
This is a strong discovery-platform paper with an unusually interesting in vivo proof of concept. Its most important contribution is showing that a systematic Perturb-seq-based approach can identify single transcription-factor perturbations that partially reverse aging-associated cellular and tissue phenotypes. The main caution is that the findings show partial, short-term rejuvenation-like remodeling, not yet proven durable rejuvenation, broad organismal benefit, or long-term safety.