RamanOmics Maps the Molecular Fingerprint of Senescent Cells in Aging Tissue
A new multimodal platform identifies senescent cells without labels, linking biochemical lipid signatures to transcriptional aging programs across tissues.
Summary
Senescent cells are hard to find inside intact tissue because most detection methods rely on a few molecular markers or require destroying the sample. Researchers combined label-free Raman light imaging, which reads the chemical makeup of cells, with single-nucleus RNA sequencing and spatial gene mapping in young (2-month) and old (26-month) mouse lung and skin. They found tissue-specific senescence programs: matrix remodeling and TGF-β signaling in lung, and epidermal differentiation genes in skin. They also found a shared lipid-related Raman signature (1,131–1,135 cm−1) in p21+ senescent cells, and used machine learning to build a multimodal barcode that identifies senescent cells in place. The same epidermal genes reactivated in senescent cells during mouse wound healing. The work is preclinical, but it points toward nondestructive senescence detection.
Detailed Summary
Senescent cells contribute to tissue aging, yet they are usually defined by isolated markers such as p21 or p16, which says little about their chemistry or how they differ between organs. Transcriptomics shows which genes are on, but not the lipid composition, protein folding, or matrix chemistry that ultimately govern cell function. This study asks whether the biochemical fingerprint of senescence can be read directly in intact tissue and tied to gene programs, which matters for any future effort to find and track senescent cells without destroying the sample.
The team built RamanOmics, which combines hyperspectral Raman imaging (600–1,800 cm−1, 873 spectral dimensions, 3 µm pixels), single-nucleus RNA sequencing, and STARmap in situ sequencing covering 890 genes (700 cell-type markers and 190 senescence/aging markers). They studied lung and skin from naturally aged mice at 2 and 26 months, with three animals per age group. Single-nucleus profiling captured 35,474 lung and 12,128 skin nuclei. Raman spectra and spatial transcriptomics were collected from the same sections and registered using the 791 cm−1 DNA Raman peak and DAPI staining, yielding 15,201 matched lung cells and 11,069 matched skin cells.
The transcriptional aging patterns differed by tissue. Old lung showed immune activation, vascular remodeling, and reduced epithelial renewal, with endothelial cells and T cells shifting most. Old skin showed the greatest changes in fibroblasts and interfollicular epidermis. Senescent cells, defined by Cdkn1a (p21), were most common in lung AT2 cells (0.31–0.51%) and endothelial cells (0.06–0.36%), and in skin interfollicular epidermal cells (0.53–1.55%). Senescent cells were also physically larger than their neighbors, and old skin showed a higher concentration of them in the epidermis.
Senescent cell programs also differed by age. In old lung, p21+ cells upregulated Serpine1, Dab2, and Mrc1 and downregulated Acoxl, Aldh1a1, and Prdx6, a fibrosis- and inflammation-prone profile. Young senescent cells instead showed DNA-damage-response genes, which the authors interpret as a more transient, reparative state. Old skin senescent cells downregulated lipid-metabolism genes (Cd36, Cidec) and a DNA-repair gene (Eepd1). Across skin, senescent cells shared Krt1, Krt10, Lor, and Sbsn expression.
On the Raman side, old lung had higher overall intensity than young lung (46.72 vs 43.27, P = 2.1 × 10−13), whereas p21+ senescent cells in old lung had lower intensity than in young lung (42.42 vs 45.43, P = 1.36 × 10−3). A conserved lipid-linked peak at 1,131–1,135 cm−1 marked p21+ cells across tissues. A random forest classifier trained on integrated spectral and transcriptomic features produced a multimodal barcode that classifies senescent versus nonsenescent cells. In a wound-healing model, senescent cells reactivated Krt10, Lor, and Sbsn alongside increased lipid-associated Raman signatures.
This is a technology and discovery study in mice, not an intervention trial. Senescence was defined mainly by p21, which the authors note marks a population distinct from p16+ cells, so the signature may not generalize to other senescent subtypes. With three mice per group, frozen sections, and a low senescent cell fraction, the barcode needs validation in other tissues, species, and human samples before any clinical use.
Key Findings
- Compared 2-month and 26-month mouse lung and skin (n = 3 mice per group), profiling 35,474 lung and 12,128 skin nuclei by snRNA-seq
- Matched 15,201 lung and 11,069 skin cells across Raman imaging and a 890-gene STARmap panel using the 791 cm−1 DNA peak plus DAPI registration
- Found a conserved lipid-linked Raman signature at 1,131–1,135 cm−1 marking p21+ senescent cells across tissues
- p21+ cells were most frequent in lung AT2 cells (0.31–0.51%) and skin interfollicular epidermis (0.53–1.55%)
- Old lung p21+ cells showed lower Raman intensity than young (42.42 vs 45.43, P = 1.36 × 10−3), while overall old lung intensity was higher (46.72 vs 43.27, P = 2.1 × 10−13)
- Old lung senescent cells upregulated Serpine1, Dab2, and Mrc1; old skin senescent cells downregulated Cd36, Cidec, and the DNA-repair gene Eepd1
- In mouse wound healing, senescent cells reactivated Krt10, Lor, and Sbsn with increased lipid-associated Raman signals
Methodology
Naturally aged (26-month) and young (2-month) mouse lung and skin were collected in OCT, with three animals per group for snRNA-seq and three for Raman imaging followed by STARmap-ISS on the same sections. Raman spectra (600–1,800 cm−1, 873 dimensions) were registered to STARmap cell segmentation, and p21+ cells served as spatial anchors. Differential expression used FDR <0.05, Tangram imputed whole-transcriptome expression, and a random forest classifier built the multimodal barcode. Spectral comparisons were made at single-cell level, and a wound-healing model tested whether the signature changes during repair.
Study Limitations
The work was done only in mice, with three animals per group, and senescent cells were defined mainly by p21 expression, which captures just one subpopulation. Senescent cells were rare (often under 2% of cells), and the Raman signature still needs validation in other tissues, in human samples, and for live in vivo use. Conflict-of-interest disclosures were not included in the supplied text.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: