Nytt verktyg mäter cellers biologiska ålder
Forskare vid Karolinska Institutet och Stockholms universitet har utvecklat ett nytt verktyg som uppskattar en cells biologiska ålder
genom att analysera dess genaktivitet. I tidskriften Advanced Science visar de också hur verktyget kan användas för att hitta substanser som påverkar cellers åldrande.
Åldrande är en viktig riskfaktor för många sjukdomar, men det saknas fortfarande metoder som på ett tillförlitligt sätt kan mäta hur snabbt celler åldras.
I den nya studien presenterar forskarna Pasta, ett fritt tillgängligt verktyg som beräknar en cells biologiska ålder utifrån vilka gener som är aktiva eller inaktiva.
För att utveckla verktyget analyserade forskarna genaktivitetsdata från mer än 17 000 vävnadsprover från friska människor. Därefter testades modellen på flera oberoende datamängder.
– Tidigare verktyg för att mäta biologisk ålder fungerar ofta bara för en vävnad eller en typ av data, vilket begränsar deras användbarhet. Vi byggde ”Pasta” för att fungera brett, över många vävnader, celltyper och laboratorietekniker, så att alla forskargrupper kan tillämpa det på data de redan har. Med det här verktyget kan vi följa hur celler förändras över tid och få ledtrådar om vilka mekanismer som driver åldrandet, säger försteförfattaren Jérôme Salignon, forskare vid institutionen för medicin, Huddinge, Karolinska Institutet som har lett studien tillsammans med Federico Pietrocola, senior forskare vid institutionen för cell- och molekylärbiologi, Karolinska Institutet, samt Christian G. Riedel, professor vid Stockholms universitet och senior forskare vid institutionen för medicin, Huddinge, Karolinska Institutet.
Forskarna såg att celler med hög biologisk ålder ofta hade ökad aktivitet i gener som är kopplade till DNA-skador och stress i cellen. Verktyget kunde också skilja mellan äldre, så kallade senescenta celler, och mer ungdomliga stamcellsliknande celler.
I nästa steg användes Pasta för att analysera mer än tre miljoner genprofiler från offentliga databaser där celler exponerats för tusentals läkemedel och genetiska förändringar. Analysen identifierade substanser och biologiska signalvägar som verkade öka eller minska cellernas biologiska ålder. Några av resultaten bekräftades sedan i laboratorieförsök på mänskliga celler.
– Att på ett tillförlitligt sätt bestämma cellers biologiska ålder har länge varit en stor utmaning. Nu kan vi göra det utifrån genuttrycksdata – en typ av data som redan genereras rutinmässigt i väldigt många forskningsstudier. Jag tror att vi därmed går in i en ny era där biologisk ålder kan användas som ett experimentellt mått inom många olika typer av studier. Pasta öppnar helt nya möjligheter att förstå mekanismerna bakom åldrandet och att systematiskt söka efter gener och substanser som kan påverka det, säger Christian Riedel.
Jérôme Salignon fortsätter:
– Verktyget kan hjälpa oss att hitta kandidater för framtida behandlingar av åldersrelaterade sjukdomar och cancer. I laboratorieexperiment validerade vi två nya kandidater: pralatrexate, som accelererade cellulärt åldrande, och piperlongumine, som gjorde cellerna mer ungdomliga. Men våra resultat bygger på cellförsök, så mer forskning behövs innan de kan omsättas i behandlingar för patienter.
Publikation
”Pasta, a Versatile Transcriptomic Clock, Maps the Chemical and Genetic Determinants of Aging and Rejuvenation”, Jérôme Salignon, Maria Tsiokou, Patricia Marqués, Enriqueta Rodríguez-Diaz, Hazel Ang, Federico Pietrocola, Christian G. Riedel, Advanced Science, online 27 juli 2026, doi: 10.1002/advs.76740
https://advanced.onlinelibrary.wiley.com/doi/10.1002/advs.76740
ABSTRACT
With the growing burden of age-related diseases, understanding and modulating the aging process has become a priority. Transcriptomic aging clocks (TACs) can track biological age but remain limited by platform dependence, tissue specificity, or restricted accessibility. To address this, we developed Pasta, a robust and broadly applicable human TAC, built using a novel ‘age-shift’ learning framework. Pasta accurately predicted relative age across diverse tissues and data types, including bulk and single-cell RNA-Seq as well as microarray data. Its predictions aligned with senescent and stem-like cellular states and relied on model coefficients enriched for p53 and DNA damage response pathways. Pasta’s age scores correlated with tumor grade and patient survival in several cancer types, indicating potential clinical relevance. Applied to over three million transcriptomes from the Connectivity Map L1000 dataset, Pasta identified both established and previously unrecognized age-modulatory compounds and genetic perturbations, highlighting mitochondrial translation and mRNA splicing as key determinants of cellular propensity for aging and rejuvenation, respectively. Experimental validation confirmed pralatrexate as a potent senescence inducer and piperlongumine as a rejuvenating agent in human cells. Together, these findings establish Pasta as a versatile and accessible tool for aging research and therapeutic discovery.
Introduction
The growing prevalence of chronic non-communicable diseases in later life has positioned aging as a major driver of pathology. This, coupled with advances in understanding the fundamentals of aging and health [1, 2], has intensified interest in pharmacological and lifestyle strategies to modulate aging [3]. Unlike chronological aging, which progresses at a constant pace, biological aging is modulated by endogenous and environmental factors that can accelerate, decelerate, or even reverse aging processes [4]. This highlights biological age as a latent but measurable trait that reflects current and future health, captures individual variability in aging rates, and offers a means to evaluate geroprotective interventions.
Over the past decade, a range of computational tools—collectively termed ‘aging clocks’—have been developed to quantify biological age and estimate aging rates. These models draw from diverse molecular and clinical hallmarks across multiple ‘omics’ layers, including epigenetic, transcriptomic, metabolomic, proteomic, and metagenomic data [5]. Among them, epigenetic clocks based on genome-wide CpG methylation patterns have been widely used due to their high precision in estimating chronological and biological age, strong predictive value for morbidity and mortality, and the relative stability of such DNA methylation (DNAm) marks, which enable the capture of cumulative aging effects [6]. However, DNAm clocks also have limitations, including reduced sensitivity to rapid or transient changes [7] and limited biological interpretability, as CpGs are not always clearly linked to gene function [8]. In addition, single-cell DNAm approaches remain costly, technically demanding, and lack standardized protocols, restricting their utility in resolving aging at single-cell resolution [6].
Transcriptomic aging clocks (TACs) address these limitations: they sense transient shifts, yield interpretable gene and pathway outputs, and can be applied to extensive bulk [9], single-cell [10, 11], spatial [10, 12], genetic [13, 14], and chemical [13, 15] perturbation datasets. Notably, ongoing efforts include constantly growing transcriptomic perturbation resources, such as the Connectivity Map (CMAP) L1000 [13], Perturb-Seq datasets [14], or the Tahoe-100 dataset [15]. Applying TACs to such datasets could reveal regulators of aging and rejuvenation with translational potential. For example, compounds that raise the biological age of cancer cells may enhance therapeutic efficacy [16], and gene perturbations that accelerate aging in iPSC-derived neurons can improve models of late-onset disease [17]. Conversely, agents that lower age in normal cells may promote regeneration; improving stem cell fitness [18], iPSC protocols [19], or partial reprogramming cocktails [20]. While a handful of age-modulatory perturbations have recently been described [21], systematic approaches to identify them remain limited. A recent study showed the potential of such an approach by mining the Gene Expression Omnibus (GEO) [22] database using aging clocks to find novel age-modulatory perturbations [23]. Another landmark and untapped resource in this context is the CMAP L1000 dataset [13], which contains over three million transcriptomes from 248 cell lines exposed to more than 14 000 gene perturbations and 30 000 compounds. Identifying both general and cell line-specific age-modulatory perturbations in this dataset could inform the development of more effective age-related translational strategies.
Human TACs have been developed in blood [24], muscle [25], fibroblasts [26], skin [27], and multi-tissue settings [28-33], and more recently at single-cell resolution using blood [34], immune cells [35, 36], and PBMCs [37]. Despite this progress, no TAC has achieved widespread adoption comparable to epigenetic clocks, due to limitations in performance, availability, and platform compatibility. Gene expression sensitivity to environmental cues enables detection of transient states but also introduces variability and noise, requiring robust modeling across heterogeneous datasets. Most human TACs, including single-tissue models and the multi-tissue RNAAgeCalc [29], are trained on single datasets, leading to overfitting and limited generalization. Consistently, MultiTIMER [30], trained on multiple datasets, outperforms RNAAgeCalc [29] despite being trained on fewer samples (∼3000 vs ∼9600). Another key barrier is availability: DeepQA [32], and Shokhirev & Johnson ’s [28] TACs offer no usable software, while RAPToR [31] requires users to generate custom references without default human datasets. Moreover, most multi-tissue TACs [28-30, 32] require raw data reprocessing, batch correction, or model retraining, limiting their use to bulk RNA-Seq data and reducing reproducibility. While the multi-species tAge [33] clock circumvents these issues, it was trained on only four human tissues, leaving it unclear how well it performs in other human tissues.
To address these limitations, we present Pasta (Predicting Age-Shift from Transcriptomic Analysis), a ready-to-use transcriptomic aging clock (TAC) applicable across multiple tissues and experimental platforms. Pasta showed improved performance over MultiTIMER [30] and tAge [33]—the current leading multi-tissue human TACs—across diverse datasets, generated biologically meaningful age estimates in mouse tissues profiled using different platforms, and revealed a substantial contribution of p53-related genes to its output. Beyond age prediction, Pasta effectively distinguished between senescent, quiescent, and stem cells, in both static and dynamic settings, indicating its capacity to capture bidirectional cell state transitions from stemness to senescence. Clinically, we observed that the Pasta age score could stratify tumor grade and survival for several cancers, which may reflect its ability to detect senescence- and stemness-related features in tumors. When applied to the CMAP L1000 dataset [13], Pasta identified potential age modulators such as pralatrexate and piperlongumine, which we experimentally validated, and it pinpointed molecular features associated with cell-line responsiveness to pro- or anti-aging interventions. Taken together, Pasta offers a robust and interpretable framework for biological age estimation across tissues, platforms, and species, and for systematic discovery of genetic or chemical modulators of aging and rejuvenation, with potential relevance to translational research, including in oncology [16], neurodegeneration [17], and regenerative medicine [18-20].
Läs hela artikel open source free
https://advanced.onlinelibrary.wiley.com/doi/10.1002/advs.76740
Nyhetsinfo
www red DiabetologNytt