A person’s body does not age uniformly; instead, some parts wear out before others. A new study in the December 6 Nature online now puts numbers to this, and pins down the proteins involved. Researchers led by Tony Wyss-Coray at Stanford University, Palo Alto, California, profiled the plasma proteome of more than 5,000 people to uncover changes across the lifespan. They focused on proteins produced by specific organs, using these to measure the relative age of each organ. In one in five people, one organ was deteriorating faster than the rest of their body. This correlated with greater risk of disease in that system.

  • Plasma proteome study quantified the aging profile of different organs in the body.
  • One in five adults have an organ with accelerated aging, amplifying disease risk.
  • Synaptic, vascular, and myelin proteins can portend cognitive decline.

For brain, the authors linked a set of 49 proteins to rapid aging and dementia risk. The proteins predicted cognitive decline better than did plasma p-tau181. Many of them are expressed in synapses, myelin, and the vasculature, suggesting that defects in these systems could foreshadow decline. The data could point toward new biomarkers and therapeutic targets, Wyss-Coray told Alzforum.

Luigi Ferrucci at the National Institute on Aging, Bethesda, Maryland, called the paper a landmark study that lays a foundation for translating plasma proteome studies to the clinic. “The specificity and predictive values obtained are impressive, and appear to be superior to some of the biomarkers used currently in clinical practice,” he wrote to Alzforum. Keenan Walker, also at NIA, saw multiple applications. “In addition to providing individuals with a readout of organ-specific health, the organ age scores will likely be useful for monitoring the therapeutic or adverse effects of interventions or exposures on multiple organ systems in a cost-effective manner,” he wrote (comments below).

How to Build Organ Aging Clocks. Using plasma proteome data (left), researchers developed aging profiles for 11 organs (middle). A person's organs had different physiological ages, which correlated with disease (right). [Courtesy of Oh et al., Nature.]

Previously, Wyss-Coray and colleagues had reported that proteins from body fluids of young mice rejuvenated old mice, and proteins from old mice prematurely aged the young (May 2014 conference news; Sep 2016 news; May 2022 news). A small study suggested the same could be true in people (Dec 2017 conference news). Which proteins were responsible?

To identify aging proteins in human plasma, joint first authors Hamilton Se-Hwee Oh and Jarod Rutledge analyzed blood samples from five cohorts. These were the LonGenity Ashkenazi Jewish cohort at Albert Einstein College of Medicine in New York, the Stanford Aging Memory Study, the Stanford Alzheimer’s Disease Research Center cohort, the Knight-ADRC cohort at Washington University in St. Louis, and the Covance study of lifetime health. The first four studies were longitudinal, the fifth cross-sectional; together they comprised 5,676 participants.

Using the SomaScan commercial assay, the authors measured nearly 5,000 proteins, then cross-referenced these with data from the Genotype-Tissue Expression project, an RNA-Seq study of human tissues (GTEx Consortium, 2020). This netted 856 proteins that were predominantly expressed by a single organ or tissue. The researchers profiled 11 systems: brain, heart, kidney, pancreas, liver, lung, intestines, arteries, muscle tissue, immune cells, and adipose tissue.

Using samples from about 1,400 people in the Knight-ADRC cohort, they applied machine learning to define how the proteins specific for each organ changed over the lifespan. These data enabled the authors to calculate an organ “age gap,” that is, the difference between a given organ’s physiological and chronological age, for each person in the other cohorts. They also analyzed changes in all non-organ-specific proteins, referring to this as “organismal aging.”

Surprisingly, organ age gaps only weakly correlated with a person’s organismal aging profile, at r=0.29. Instead, different organs within a person aged at different rates. Moreover, about 20 percent of participants had one organ with extreme aging, defined as more than two standard deviations beyond the norm for their chronological age. Extreme organ age correlated with disease in that system. For example, kidney aging was linked to diabetes, obesity, and hypertension; heart aging was linked to heart attacks and atrial fibrillation.

Brain aging associated with cerebrovascular disease. Curiously, Alzheimer’s disease was most closely linked to overall organismal aging, rather than to a specific organ, suggesting that many systems contribute to it. Partly, this may be because the organismal aging protein set includes many vascular, metabolic, and immune proteins that are found in brain, but are not specific to it.

To home in on the proteins responsible for cognitive slippage, the authors looked for brain-specific proteins that correlated with worsening CDR scores. They identified 49, many of them active in synapses, myelin, and extracellular matrix. The set included the synaptic proteins complexin and neurexin, oligodendrocyte proteins carnosine dipeptidase 1 and LancLike Glutathione S-transferase 1, and ECM proteins tenascin R and heparan sulfate-glucosamine 3-sulfotransferase 4. This “CognitionBrain” protein set predicted decline, with people one standard deviation higher on the scale having a third higher risk of sliding by two points on the CDR over the next five years.

The CognitionBrain protein set predicted decline better than did plasma p-tau181, baseline CDR, or an AD polygenic risk score (Apr 2022 news). Combining CognitionBrain and plasma p-tau181 boosted overall predictive accuracy, suggesting the plasma proteins capture distinct information that is separate from amyloid and tau pathology.

Finally, the authors looked for early changes by examining plasma proteins associated with worse cognition in otherwise healthy people. This picked out a distinct set of brain, artery, and non-organ-specific proteins, including pleiotrophin, transgelin, and WNT1. Most of these proteins were expressed by the vasculature’s smooth muscle cells, pericytes, and fibroblasts, and many interacted with extracellular matrix. Some of these proteins, including sclerostin, frizzled related protein, and matrix gla protein, were previously shown to contribute to vascular calcification (Callegari et al., 2014; Köhler et al., 2021; Qureshi et al., 2015). The authors believe the findings shed light on the biology of the earliest phase of cognitive decline. Other studies in mice and people have linked changes in vascular and myelin proteins with aging and AD, as well (Jun 2023 news; Jul 2023 news).

Wyss-Coray said the next step is to replicate the findings in larger and more diverse cohorts. In ongoing work, he has identified similar organ aging patterns in U.K. Biobank samples from 50,000 people. His lab is doing functional studies to determine how specific proteins affect aging.—Madolyn Bowman Rogers


  1. This is wonderful work. It truly lays out the basis for translation to the clinic of plasma proteomic analysis, a powerful method that so far has been used almost exclusively for research purposes (some application in oncology excluded).

    The authors measure a large number of proteins (~5k), in plasma samples of five separate cohorts, using an aptamer technology. From the proteins assessed and quantified, they select a subgroup that is specifically overrepresented in specific organs. They do this using data from a high-quality database (GTEX) of gene expression from specific organs.

    Then, using organ-specific subsets, they derive organ-specific aging clocks, and then they go on to demonstrate that these organ-specific aging clocks specifically predict emerging pathology in those organs. The specificity and predictive value obtained are impressive and appear to be superior to some of the biomarkers used currently in clinical practice.

    Overall, all the organ-specific clocks predict mortality, although with different strength. Only a minority of individuals (<2 percent) appear to have accelerated “aging” in multiple organs. This is somewhat surprising because previous studies have found that multiple organ function and rate of decline with aging are correlated, especially in older persons.

    The analysis of the brain-specific clock is particularly elaborated and sophisticated and, to work properly, required tuning a clock on both organ-specific and chronological age. After this analysis, the predictivity of various brain-specific phenotypes is remarkable.

  2. In a commendable effort, Oh and colleagues identified a set of organ-specific proteins that were then used to compute organ-specific age and age gaps. By taking a set of proteins that were specifically expressed in brain tissue and mapped onto cognitive function, the authors identified a brain-specific proteomic signature that was associated with AD dementia.

    The protein-based organ age scores could provide potential value for both research and clinical practice. For example, in addition to providing individuals with a readout of organ-specific health, the organ age scores will likely be useful for monitoring the therapeutic or adverse effect of intervention or exposures on multiple organ systems in a cost-effective manner.

    Beyond demonstrating that the molecular brain aging signature is associated with AD dementia, the authors identified a subset of brain-specific proteins as part of the cognition-optimized brain aging model. This subset of proteins was used to calculate the CognitionBrain age gap, a measurement that proved to be much more predictive of AD than the Brain age gap.

    Importantly, the authors show that the CognitionBrain age gap predicted AD risk independent of plasma pTau181, age, and AD polygenic risk score. These findings suggest that the CognitionBrain age gap may provide incremental value for dementia prediction beyond that of traditional AD/dementia biomarkers. Proteins that had the largest positive weights for the CognitionBrain age score included several synaptic proteins, such as complexin 1 (CPLX1) and complexin 2 (CPLX2), which my group recently identified as midlife biomarkers of 25-year dementia risk (Walker et al., 2023).

    Of particular importance is the set of analyses that examined how non-CNS organ-specific aging related to cognitive decline and Alzheimer’s disease. By demonstrating that the age gaps derived from artery- and pancreas-specific proteins and optimized for cognition predicted AD risk, the authors provide evidence for the role of these organ systems in AD/dementia pathogenesis. Early vascular dysfunction was especially implicated in this study, as the artery age gap predicted conversion to MCI over a 15-year follow-up period and did so more strongly than did brain age. These findings support previous work that has identified vascular dysfunction as an early feature of—and likely risk factor for—late-onset AD (Iturria-Medina et al., 2016).


    . Proteomics analysis of plasma from middle-aged adults identifies protein markers of dementia risk in later life. Sci Transl Med. 2023 Jul 19;15(705):eadf5681. PubMed.

    . Early role of vascular dysregulation on late-onset Alzheimer's disease based on multifactorial data-driven analysis. Nat Commun. 2016 Jun 21;7:11934. PubMed.

  3. This new study provides one of the most comprehensive approaches to date on how aging intersects with many disorders for which it acts as a major risk factor. One major strength is Oh et al.'s leveraging of a large-scale protein quantification approach with existing gene expression datasets to create fresh insight. Aptamer-based relative quantification of nearly 5,000 plasma proteins was performed in more than 5,000 individuals from cohorts across several U.S. centers.

    To gain organ specificity for the plasma proteins, the authors mapped these proteins to organ-specific gene expression enrichments gleaned from the GTEx project. This allowed organ age gaps in biological and chronological age to be determined for each organ under study using machine-learning approaches. Interestingly, while there are extreme organ age gaps that associate with some diseases, Alzheimer’s disease most strongly associates with the overall set of organismal aging proteins in this model, suggesting that AD may be a disorder involving multiple organ systems.

    A wealth of information from the identified plasma protein organ-specific sets can be mined to further probe specific drivers for aging processes not previously linked to certain diseases. To extract additional insights linking brain aging to AD specifically, the authors used cognitive phenotypes in their cohorts and brain age gap protein sets to train a more sophisticated model they term the CognitionBrain aging model. This brain age gap model was able to provide predictive power for several important AD-related measures. Several of the individual proteins comprising the model were previously associated with cognition and AD. Of note, some are enriched in oligodendrocytes, a cell type shown to be sensitive to the effects of young blood in old mice through parabiosis (Ximerakis et al., 2023).

    Of particular interest to my group is the subset of brain proteins in the model involved in the biology of the extracellular matrix, which we find is regulated by certain systemic mediators in the context of brain aging (Castellano et al., 2017; Ferreira et al., 2023). 

    Overall, the study sets up many exciting directions for the field. How malleable are the aging signatures for specific organs in the face of challenges, lifestyle changes, or therapeutic interventions? Are some organs more resistant to aging reversal than others? With the relative ease of longitudinal blood draws, these and many other answers seem within reach, especially with the ever-expanding coverage offered by proteomic platforms and development of new computational approaches to model the complexity of aging. 


    . Heterochronic parabiosis reprograms the mouse brain transcriptome by shifting aging signatures in multiple cell types. Nat Aging, March 9, 2023

    . Human umbilical cord plasma proteins revitalize hippocampal function in aged mice. Nature. 2017 Apr 19; PubMed.

    . Neuronal TIMP2 regulates hippocampus-dependent plasticity and extracellular matrix complexity. Mol Psychiatry. 2023 Sep;28(9):3943-3954. Epub 2023 Nov 2 PubMed.

  4. This study provides nice insights into aging at the molecular level, gained through proteomics applications. That the investigators were able to generate such a powerful dataset, and analyze the mass of data, shows how far proteomics technology and machine learning have truly come. The ongoing maturation of the approaches used here will advance our knowledge of aging.

    Though the analyses were unbiased and agnostic, the interpretation of their place in aging and biology are speculative but reasonable. Overall I tend to agree with the authors’ interpretations. Some of the speculative points, if accurate, to me pack quite a wallop.

    One such point had to do with the fact that while a large number (~20 percent) had disproportionate aging in one organ, far fewer (<2 percent) had disproportionate aging in more than one organ. Adding to this additional context that disproportionate organ aging predicts organ-related disease, I wonder whether a particular stress within an organ accelerates the subsequent aging of the organ.

    If one extrapolates from data that indicate advancing age demands functional compensation, one could wonder whether an organ-specific stress compounds the amount of compensation that is triggered by aging itself. This might force the organ to reach its limit of compensation, and to transition from compensated to decompensated aging.

    The manuscript identifies proteins and, by extension, pathways that I suspect will provide insight into the aging-Alzheimer’s disease nexus. Studies like this one, of course, are correlation-focused, and inferring causation from correlation is tricky. Does disease/stress cause aging, or does aging cause disease/stress? Maybe it is both ways.

    Also, while I was very impressed that the authors were able to infer organ specificity of the measured plasma proteins, I wonder how confident we can be about the origin of those proteins. To this point, the manuscript briefly addresses the issue of protein levels going in opposite directions between brain and plasma, and gives the example of Aβ levels going down in Alzheimer’s patient CSF while cortex plaque burden increases. Explanations for this general principle, as is the case with Aβ compartmentalization, may be more complex than we imagine.

    Regardless, this is a very well-done and interesting study.

Make a Comment

To make a comment you must login or register.


News Citations

  1. In Revival of Parabiosis, Young Blood Rejuvenates Aging Microglia, Cognition
  2. Young Blood a Boon for APP Mice
  3. Cerebrospinal Fluid from Youngsters Boosts Memory in Old Mice
  4. Blood, the Secret Sauce? Focus on Plasma Promises AD Treatment
  5. Paper Alert: Massive GWAS Meta-Analysis Published
  6. Transcriptomics Confirm Vascular Changes in Alzheimer’s Brain
  7. Atlas of Mouse Brain Aging Finds Biggest Changes in White Matter

Paper Citations

  1. . The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020 Sep 11;369(6509):1318-1330. PubMed.
  2. . Increased calcification in osteoprotegerin-deficient smooth muscle cells: Dependence on receptor activator of NF-κB ligand and interleukin 6. J Vasc Res. 2014;51(2):118-31. Epub 2014 Mar 7 PubMed.
  3. . The Human Phenotype Ontology in 2021. Nucleic Acids Res. 2021 Jan 8;49(D1):D1207-D1217. PubMed.
  4. . Increased circulating sclerostin levels in end-stage renal disease predict biopsy-verified vascular medial calcification and coronary artery calcification. Kidney Int. 2015 Dec;88(6):1356-1364. Epub 2015 Sep 2 PubMed.

Further Reading

Primary Papers

  1. . Organ aging signatures in the plasma proteome track health and disease. Nature. 2023 Dec;624(7990):164-172. Epub 2023 Dec 6 PubMed.