Scientists hone tools to measure aging and rejuvenation interventions

Scientists hone tools to measure aging and rejuvenation interventions

It takes stamina: Rolling Stones front man Mick Jagger has been performing for 62 years.Credit: Gonzales Photo / Alamy Stock Photo

At the end of August, Copenhagen will welcome the who’s who in longevity research for the tenth Aging Research and Drug Discovery Meeting. Top of the agenda are interventions aimed at slowing or even reversing the aging process. But if the aim is to persuade the US Food and Drug Administration and other regulators to approve drugs, diets or supplements, the field needs to agree on the best biomarkers or tools to measure the changes that best characterize aging in humans.

“This is the question that I get most often: which aging biomarker to use? Because everyone would like to use the biomarkers to quantify aging,” said longevity researcher Vadim Gladyshev from Harvard Medical School during the launch in March of the Biomarkers of Aging Consortium, a new project set up by leading academics, clinical practitioners and biotech companies to collaborate in testing and identifying reliable biomarkers.

Gladyshev, who is director of redox medicine, Brigham and Women’s Hospital, says that it is still unclear which biomarker is best to use for evaluating pro-longevity interventions. To that end, the consortium compiled a list of the most commonly used biomarkers, including those from ’omics — genomics, epigenomics, transcriptomics, proteomics or metabolomics — as well as AI-derived blood-based clocks and other clinical markers. The plan is that each proposed biomarker will be assessed by a panel of longevity scientists, pass through regulatory bodies, and emerge with a set of guidelines for the community to use and refine.

One of the challenges to measuring aging in humans is that there is so much heterogeneity: people age at different rates, and so do body tissues within one person and even cells within a given organ. For the many interventions tested around the world — be it supplements, drugs, caloric restriction, plant-rich diets, intermittent fasting, metabolic interventions, young plasma, senolytics or transient reprogramming — people will vary in how they respond. “We need to develop a common and standardized way to measure the impact of these interventions in different people,” said consortium member Vittorio Sebastiano, Stanford School of Medicine, during the launch event. A paper submitted to Cell reviewing the potential and challenges of blood-based composite biomarkers will be published at the end of August.

“How do we measure biological aging in humans? We need this; it’s a prerequisite for trials,” says Sebastien Thuault, chief editor of Nature Aging. In June, China set up its own Aging Biomarker Consortium to answer fundamental questions about what underpins healthy aging and the transition to pathological states, and how to best quantify changes in biological age for those wanting to test aging interventions.

Consensus may be hard to achieve, however. There is a community that is skeptical of the use of biomarkers, mostly because, until recently, efforts to define a single measurement for biological aging have been largely unsuccessful. “Should we be looking at gene expression, physiological function, regenerative capacity?” says Sebastiano. To add to the complexity, clocks can potentially measure events at the single cell, tissue or whole organism level. What’s needed is a holistic view of these changes. For that reason, Kejun (Albert) Ying, from the Gladyshev lab, released ClockBase, an online visual tool that incorporates all these data into a single platform for estimating biological age. With machine learning, Lim mined more than 2,000 DNA methylation datasets and nearly 200,000 mouse and human samples. The resulting age-profiling tool is open for the academic drug discovery community to access and test new interventions.

In 2013, the publication of pioneering work on possible molecular biomarkers of aging by Steve Horvath at the University of California, Los Angeles, and separately by Gregory Hannum, at the University of California, San Diego, represented a pivotal moment for the field. They proposed in their work that DNA methylation at specific cytosine-phosphoguanine (CpG) sites within the human genome can be used to predict an person’s chronological age. Hannum, who is now principal bioinformatics scientist at Exact Sciences, mathematically modeled changes in 71 CpG markers in DNA from white blood cells from a single cohort of 656 individuals. Horvath, now principal investigator at Altos Labs in Cambridge, UK, introduced the pan-tissue clock. He used an Illumina DNA methylation microarray to spot CpG sites gained or lost in 8,000 human samples from healthy tissues and cell types at different ages. With machine learning algorithms, he selected 353 CpG sites that could work as an age predictor and called it an epigenetic ‘clock’.

Second-generation clocks add into the mix blood biochemistry and lifestyle choices that may influence CpGs. One of these is PhenoAge, a composite epigenetic marker developed by Morgan Levine while at the David Geffen School of Medicine, University of California, Los Angeles. The test incorporates nine blood markers regularly used in annual physicals, such as creatinine and C-reactive protein, to predict biological age: it can tell if a person is physiologically younger or older than their chronological age. Levine, now principal investigator at Altos Labs, developed this new phenotypic age estimator combining blood results with DNA methylation at 513 CpG sites. This clock correlates with aging tested in different tissues and cell types, and in addition identifies pathways — for instance, pro-inflammatory ones — that when activated are associated with accelerated aging.

Levine has now developed Systems Age, a more reliable version of the clock that captures aging in 11 different physiological systems, including, heart, lung, kidney and liver, in a single aging score. Levine used machine learning to link DNA methylation to the clinical chemistry and functional measures from each system. The result is a single aging measure that requires a single blood draw. The marker is described in a bioRxiv paper still under review. “When it comes to predicting general aging-related outcomes, we have a new group of methylation measures that seem to be better than most existing epigenetic clocks,” Levine says.

The GrimAge clock, developed in Horvath’s lab, is also a composite maker: it meshes DNA methylation with a detrimental environmental exposure — smoking. Horvath combed the genome of thousands of people to identify patterns of CpG changes due to tobacco smoking. Using mathematical algorithms, he combined the results with a selection of plasma proteins associated with mortality or disease — for instance, plasminogen activator inhibitor. This yielded 1,030 CpG sites that could predict mortality risk, an improved DNA methylation clock that predicts “time to death.”

Because second-generation biomarkers factor in lifestyle exposures that influence aging, they have led to a booming business in startups selling mail-order biological age tests direct to consumers. For instance, Elysium Health’s saliva test is based on the clock developed by Levine, and MyDNAge’s blood or urine test licensed Horvath’s clock. Many other companies offer biological age estimates, though none are yet approved by the US Food and Drug Administration or other regulators.

Their reluctance is not surprising: still today, a single best measure of aging does not exist. What companies developing geroprotective interventions need is the equivalent of a speedometer, something to quantify how fast a person is aging. A one-of-a-kind cohort could provide this critical information.

In Dunedin, New Zealand, a longitudinal study has been tracking the health of 1,037 babies born in 1972–1973, with 94% still taking part. Researchers led by Terrie Moffitt calculated each individual’s rate of aging using 19 indicators of organ fitness, including cardiovascular, metabolic, renal, hepatic, immune, periodontal and pulmonary systems, at ages 26, 32, 38 and 45. Artificial intelligence (AI) modeling allowed them to come up with a third-generation DNA methylation measure: DunedinPACE. “This is a clock that tracks only how fast people decline,” says Moffitt. Because all the individuals in the cohort are the same age and still healthy, any differences in deterioration are purely down to aging, and not disease. “We trained the algorithm on people aged 26 to 50s before they got any disease. They are healthy but aging.” This DNA methylation quantification of biological aging eliminates statistical bias due to disease and cuts the noise that results from exposures, such as childhood vaccines, that affect the methylome and differ between generations.

DunedinPACE as a measure of whole-body aging is being validated in 44 cohorts in 13 countries. And although the tool was trained on people up to age 45, it works just as well in studies of older adults such as the Finnish Twin Study on Aging, the US Framingham Study, the Lothian birth cohorts in the United Kingdom, the Leiden Longevity Study in the Netherlands and others. Moffitt has licensed the epigenetic biomarker to TruDiagnostics as part of a direct-to-consumer blood test kit.

Longevity clinics are some of the methylation clocks’ fastest adopters. They want responsive biomarkers that measure biological age and its change to interventions. Internal Medicine specialist Andrea Maier, founder of Chi Longevity, a clinic based in Singapore, says that to help their patients, biomarkers need to reflect within 4–6 months whether lifestyle, drug or supplement interventions are working. “The DNAmAge are currently the most meaningful biomarkers of aging,” says Maier.

The DNAmAge is the median of Hannum, Horvath, GrimAge and PhenoAge clocks. “None of the clocks [on its own] is ‘best’.” To refine the age prediction further, Maier includes Blood Age, a clock developed by Hong Kong-based Deep Longevity that makes AI-driven recommendations based on a panel of standard blood tests. Another is GlycanAge, a blood-based test that measures glycation, a reaction between sugar molecules and proteins that ramps up with age and drives low-level inflammation. The workup also includes a gut microbiome test from The NU, a company located in Leiden, The Netherlands, where Maier is the chief medical officer.

For Horvath, the more people mining the clocks for answers, the more robust the AI and machine learning predictions become. To that end, he set up the Clock Foundation, a UCLA spinout and non-profit organization that makes epigenetic aging clocks freely accessible to researchers and physicians and available for a fee to drug development companies conducting proof-of-concept clinical trials. For instance, the TRIIM-X trial, a pilot study by Stanford spinout Intervene Immune, is evaluating a treatment for thymus regeneration. The study will test multiple agents in combination, including human growth hormone, metformin and the male hormone dehydroepiandrosterone, and evaluate epigenetic and immunosenescence biomarkers in healthy men aged 50–65. Ultimately, identifying the most powerful biomarkers should help guide regulators, clinicians and venture investors.

For drug developers conducting preclinical studies, a pan-mammalian clock could prove immensely valuable. If an intervention reverses epigenetic age in a mouse, for example, it will likely work in humans too. In a herculean effort, more than 200 scientists formed the Mammalian Methylation Consortium to test more than 11,000 arrays from the across 59 tissue types and 185 mammalian species. The study results, published in Nature Aging, show that even in species with different lifespans — from long-lived whales to short-lived mice — the DNA methylation gains and losses in specific chromatin regions follow a predetermined sequence that tracks aging and tissue deterioration. Of the millions of CpG methylation changes, not all are conserved. “That’s where machine learning helps to hone in the couple of hundred CpGs that are highly conserved [across mammalian species],” says Horvath.

Several startups that focus on dogs as a model for human aging have tested the pan-mammalian clock and found it to be equally accurate as in humans. Loyal, based in the San Francisco Bay Area, is collecting saliva from 1,600 dogs for its X-Thousand Dogs study. Matt Kaeberlein, who is co-director of the University of Washington’s Dog Aging Project, notes there are many parallels between human and canine aging. A Harvard startup, Rejuvenate Bio, co-founded by George Church, is testing rejuvenation in dogs using gene therapy.

For those pursuing interventions, a single formula to estimate age is a drug developer’s holy grail. But aging remains an unwieldy and pleiotropic phenomenon, and pinning it down to one measure will likely require a multi-pronged biomarker effort. Until then, even the definition of aging for the purpose of interventions is up for debate.

Read More

Leave a Reply