Researchers at UC Berkeley have built a miniaturized lab device that uses blood serum from older donors to drive human fat and liver tissue grown from stem cells through multiple biological hallmarks of aging in under a week, a process that takes decades in living people.
Understanding how human tissues age has been hampered by a fundamental problem: the process is slow, and most of the available tools to study it are either animal-based or indirect. Mouse models have produced a large body of aging research, but mice and humans differ enough in physiology particularly in fat tissue and liver gene expression that findings frequently do not transfer well across species. Human clinical studies, meanwhile, can take years or decades to yield results. Researchers developing drugs to slow or reverse aging have lacked a rapid, human-relevant system in which to test them.
“Over $130 billion is spent on drug development each year in the United States, but over 90% end up failing in clinical trials,” said Andreas Stahl, the Ruth Okey Professor in the Department of Metabolic Biology and Nutrition at UC Berkeley and a co-leader of the study. “Pharmaceutical developers and regulators such as the US Food and Drug Administration are increasingly realizing that we need to change our drug development pipeline and make it more relevant to human biology.”
Much of the prior research on the relationship between blood and aging comes from parabiosis experiments in rodents a technique in which two animals are surgically joined so they share circulation. A 2005 study in Nature by Irina Conboy and colleagues showed that exposing older mice to blood from younger animals produced signs of tissue rejuvenation, while younger mice aged more rapidly when connected to older partners. Subsequent work by the same group found that diluting the plasma of old mice with a saline-albumin mixture had similarly rejuvenative effects on the brain, liver, and muscle. When a related procedure was tested in humans, the researchers reported reductions in inflammatory markers and changes in circulating proteins associated with aging. The underlying hypothesis that emerged from this work is that aging in mammals is partly regulated by proteins circulating in the bloodstream, which accumulate to counterproductive levels with age.
What remained unknown was whether and to what extent those findings from rodents translate to human tissue, and whether the blood-aging relationship could be reproduced outside a living body, in a controllable human model.
The Study was conducted by team co-led by Stahl and Irina Conboy, a former UC Berkeley bioengineering professor who now serves as Chief Science Officer at Generation Lab used human induced pluripotent stem cells (hiPSCs) as their starting material. These are adult cells that have been reprogrammed to a stem-cell-like state, giving researchers the ability to generate multiple tissue types from the same genetic source. The team differentiated hiPSCs into fat cells (adipocytes) and liver cells (hepatocytes), then loaded them into separate but connected chambers on a small microfluidic chip.
The chip was fabricated at UC Berkeley’s Biomolecular Nanotechnology Center and is designed to mimic the anatomical relationship between white adipose tissue and the liver two organs closely linked through the portal circulation, by which blood draining from fat tissue passes directly to the liver. Tiny channels allow fluid to flow between the two cell chambers, enabling researchers to simulate the inter-organ communication that happens in a living body.
To test aging, the researchers circulated culture medium containing five percent blood serum collected from older donors—defined as those older than 62 years—through the chip for four days. This was compared against chips receiving serum from young donors between 21 and 34 years old. Serum was matched by biological sex for each comparison group. The team then measured a range of aging-associated markers in both the fat and liver tissue compartments, including cellular senescence (using β-galactosidase staining and markers such as p16 and p21), oxidative DNA damage, inflammatory gene expression and protein secretion, and functional measures of lipid and glucose metabolism.
To validate how closely the chip’s output matched real human tissue aging, the team trained a custom machine learning model on gene expression data from hundreds of publicly available human tissue samples, then used it to assign a biological age to their chip-derived tissues.
Several candidate anti-aging interventions were also tested in the aged chip system: senolytics (drugs that selectively clear senescent cells), rapamycin (an immunosuppressant drug widely studied for longevity effects), oxytocin (a hormone also involved in metabolic regulation), a TGF-beta signaling inhibitor, and dilution of old serum with serum from young donors.
The result shows that the chips exposed to older serum showed increased cellular senescence across both tissue types, measured by multiple independent markers. Inflammatory secretory proteins associated with senescence including TNF and IL-6 were elevated. Oxidative DNA damage was robustly induced in fat tissue and showed a similar, though less pronounced, trend in liver tissue. Metabolic function was altered in both compartments: fat tissue displayed disrupted lipid handling and impaired insulin-stimulated glucose uptake, while liver tissue showed signs consistent with an insulin-resistant state. Critically, when aged fat tissue was connected to liver tissue that had not been exposed to old serum, the liver tissue began to acquire aging markers on its own, suggesting that age-related signals from fat propagate to the liver through the chip’s interorgan connection.
“The age of one organ propagates forward and establishes the age of another organ,” Irina Conboy noted. “And that’s why, typically, people show signs of aging all over their bodies.”
The machine learning model assigned tissues exposed to young serum a biological age in the range of people in their 30s, while those exposed to old serum matched gene expression profiles from people in their 50s. The model performed with 90 to 97 percent accuracy, depending on tissue type and biological sex.
Aging patterns differed between tissues treated with male versus female serum. Tissues receiving male serum showed stronger inflammatory responses and more pronounced aging markers; female serum produced more variable results, and the machine learning model performed less accurately on female tissue profiles. The researchers attributed this to the more complex hormonal landscape of female aging, including menopause. The study also identified 11 biomarkers not previously associated with aging in this tissue context, which the researchers describe as potential targets for future investigation.
Among the anti-aging interventions tested, oxytocin produced the broadest reductions in inflammation, senescence, and metabolic dysfunction in the aged chip. Rapamycin, which the researchers noted is widely used by people on an off-label basis for longevity, showed minimal rejuvenative effect in this system. The greatest improvements overall came from exposure to young serum, though the researchers note that some molecular changes associated with prior old-serum exposure persisted even after switching to young serum a finding Stahl described as a “clear memory” of prior age exposure.
Regarding DNA damage accumulation, Irina Conboy offered a revised interpretation of what the data suggest: “People assume that it takes time to accumulate damage that it is a random process. But what this paper teaches us is that DNA damage happens all the time, and cells repair this damage, unless they are exposed to old blood serum. The old serum seemingly reduced the capacity for repair.”
The team concluded that combining heterochronic human serum serum from donors of different ages with a hiPSC-based microphysiological system provides a platform for rapidly modeling human tissue aging, identifying mechanisms and biomarkers relevant to the process, and testing potential anti-aging interventions. They described the ability to induce measurable aging hallmarks in four days as a meaningful compression of a process that takes decades in living people. The system, they argue, could help address the translational gap between animal-model aging research and human biology.
The research team has filed a patent for the technology and is pursuing commercialization.
This is preclinical, in vitro research. The chip contains only two tissue typesfat and liver and does not recapitulate the full complexity of the human body. All tissues derive from a single hiPSC line, which may not capture the genetic diversity of the human population. Sample sizes for individual assays were relatively small, typically ranging from three to fifteen chip units per condition, depending on the measurement. The serum donors were drawn from specific age windows rather than a continuous age distribution.
The findings on anti-aging interventions including oxytocin and rapamycin are derived from this in vitro model and have not been tested in human clinical trials for the outcomes reported here. The conclusion that rapamycin is ineffective as a longevity therapeutic cannot be drawn from this study alone; the chip models only fat and liver tissue aging under a specific set of experimental conditions. Similarly, the apparent benefit of oxytocin in the chip does not constitute clinical evidence for its use in humans.
The machine learning model showed lower accuracy for female tissue profiles, and the researchers’ interpretation that this reflects hormonal complexityis a hypothesis rather than a demonstrated mechanism. The observation that aging propagates from fat to liver via the chip’s fluidic connection is consistent with known physiology, but whether the specific signals involved in the chip replicate those operating in living humans has not been fully established.
References
Lin Qi, Yuchen He, Alexandra Sviercovich, Xiaoyue Mei, Erzhen Chen, Yihan Xia, Michael J. Conboy, Irina M. Conboy, and Andreas Stahl. “Human microphysiological systems of aging recreate the in vivo process expediting evaluation of anti-geronic strategies.” Nature Biomedical Engineering, 2026. DOI: 10.1038/s41551-026-01618-6
