Methylation and the aging clock

Josh Mitteldorf, Progress in Methylation-based Aging Clocks Translation: Evgenia Ryabtseva, portal "Eternal Youth" http://vechnayamolodost.ru
For links, see the original article.

As it was already written in an article a year ago, we will be able to effectively test anti-aging therapeutic approaches only when we find an objective way to measure the rate of aging. In the absence of such a method, we are left with a standard epidemiological approach: testing therapy on thousands of people and waiting for the death of some of them. In the same article, it was suggested that the methylation-based aging clock would turn a page in the history of epidemiology.

Six years ago, a biological statistician from the University of California at Los Angeles, Steve Horvath, realized the potential value of the aging clock and set himself the task of learning how to measure a person's age by analyzing methylation (the presence of methyl groups) on DNA in cells of different organs. He used statistical pattern recognition software to find the relationship between a person's age and their DNA methylation profile. Methyl tags are the most well-studied of the epigenetic markers that regulate the activation and inactivation of genes, as well as ensuring the active state of various gene complexes at various stages of an organism's life.

Age is an important prognostic factor, since the probability of developing diseases that kill most people depends very much on age. In fact, the risk of developing cancer, heart disease and Alzheimer's disease increases exponentially with age.

One of the statistical results obtained with the help of Horvath's watch has a very deep meaning, which experts in the field of aging accepted very reluctantly: Horvath's watch was created on the basis of statistical methods that take into account only chronological age. The algorithm has been optimized to obtain the best estimated value of a person's calendar age. It is obvious that the age shown by the calendar is a good prognostic risk factor for human death. For Americans over the age of 40, the probability of death doubles every 8 years.

It is expected that if the Horvath clock readings correlate well with age, and age correlates well with mortality, then the Horvath clock readings should correlate with mortality. (This statement has no mathematical guarantees beyond the strength of both correlations.) An interesting twist is this: Horvath's clock readings correlate more strongly with mortality than with age. The algorithm of the clock is made taking into account the chronological age, so their mathematical component takes into account only calendar years. However, the algorithm itself predicts mortality better than calendar age.

Based on this, it can be assumed that the additional accuracy of the methylation-based clock is due not to mathematics, but to biology. In other words, the methylation process is interconnected with the biological aging process. Changes in the methylation profile do not just occur during aging, they are interrelated with some factors that cause an increase in the risk of death, or, in other words, directly with aging.

More recent developments in the field of aging clocks not only confirm the existence of this relationship, but also emphasize its strangeness.

2017: Chzan's Watch

Yan Zhang of the German National Institute of Oncology in Heidelberg, after analyzing historical blood samples of 406 people who died over a 15-year period and 1,000 demographically comparable controls, developed another method for calculating mortality risk based on the methylation profile.

His group identified 58 DNA sites with a strong association between methylation and mortality. For 49 out of 58 sites, less pronounced methylation is associated with an increased risk of death, while the other 9 sites are characterized by an inverse relationship. (Stronger methylation corresponds to reduced gene expression. In this case, this may mean that age-related mortality is more due to the activation of genes that destroy the body than the suppression of the activity of genes that protect us from death.)

None of these 58 sites were taken into account in the development of earlier aging clocks (Horvath and Hannum clocks). What follows from this? Age is more closely related to mortality than any other biological indicator, and in fact, as we age, the risk of death increases exponentially. At the same time, Czan's group was searching for methylation sites most strongly associated with mortality risk, and Horvath's group was interested in methylation sites most closely related to chronological age, and the sites identified by them did not overlap at all. In fact, less than half of the sites identified by Chzan (23 out of 58) had no statistically significant correlations with age at all.

"The newly developed epigenetic clock (DNAm age) is attracting more and more attention, as the results of an increasing number of studies demonstrate the feasibility of their use for assessing biological aging and, accordingly, health status and mortality risk. It is interesting that the identified DNA methylation changes do not coincide with the previously identified CPGs sites, whose methylation changes are used to determine age." [Zhang]

This can be partly explained by the fact that Czan conducted his study with older people (median age – 62 years) with a higher risk of death, and the Horvath clock, with which the comparison was made, was designed to assess the age "from the cradle to the grave". According to Chzan, on average, methylation levels were estimated 8.2 years before death.

Czan's computational algorithm for mortality risk determined how many of the 10 most important methylation sites are in the "worst" quartile for the tested population. (The "worst" quartile is the highest 25% for some and the lowest for other methylation sites.) 5 points on this scale corresponds to a 7–fold increase in the risk of mortality. This makes the Czan scale one of the most powerful risk indicators to date. For comparison, the body mass index value of 35 corresponds to obesity, and the risk factor of death in this case is only 1.36. Hemoglobin A1C and high-density lipoprotein are traditional indicators of the health status of older people, while their levels have only minor associations with age-standardized mortality. C-reactive protein and interleukin-6 are markers of inflammation, and their risk coefficients are 1.6 and 1.9, respectively. According to these standards, the Chzan scale is a big step forward.

It is generally believed that the methylation process is under the programmatic control of the body. There are two reasons why methylation may have a strong association with mortality. First, some changes in the methylation profile may indicate an acute reaction to some life-threatening stress. Secondly, certain changes in methylation may be a component of the body's inherent death program associated with age. It may well be that both of these mechanisms work to some extent, but perhaps the first one is more pronounced, since only 23 of the 58 identified sites have significant correlations with age.

Another interesting fact: methylation sites associated with smoking are a better indicator of mortality risk than the fact of smoking itself. (More on this below.)

2018: Levine Watch

Morgan Levine, who worked with Horvath at the University of California, last year developed a second-generation watch based not only on chronological age indicators, but also on data on mortality and morbidity. Levine's watch was optimized retrospectively, taking into account age-related diseases that developed much later than blood samples were obtained for analysis.

The work of Levine and her group consisted of 2 stages. First, they developed a set of parameters called "phenotypic age" and, in addition to chronological age, includes 9 modifying factors that contribute to an increased risk of death. Below is a list of these factors:

• Albumin: proteins dissolved in blood plasma, including hormones and other signaling molecules.
• Creatinine: a product of vital activity excreted by the kidneys, its high concentration indicates impaired kidney function, however, it can be caused by severe physical exertion.
• Glucose: blood sugar levels increase with type 2 diabetes mellitus and decreased tissue sensitivity to insulin.
• C-reactive protein: is a marker of systemic inflammation.
• Relative number (%) of lymphocytes: the most numerous types of leukocytes.
• Average red blood cell volume (RDW): The average size of red blood cells.
• Red blood cell volume distribution index (RDW): standard deviation of the previous indicator.
• Alkaline phosphatase: the level of this enzyme increases in liver diseases, including cancer and hepatitis.
• The total number of leukocytes of all types.

The list turned out to be quite unexpected. It includes not the most popular parameters, but parameters that have a statistical association with mortality. The presence of glucose and C-reactive protein in the list is not surprising (although glucose could be replaced with glycated hemoglobin, the level of which is relatively stable, while the concentration of glucose varies from hour to hour). It would also be expected to see high-density lipoproteins and interleukin-6 in this list, and especially unexpected was the great significance of the distribution of red blood cells by volume. It is measured as the standard deviation of the average volume of individual red blood cells. It turns out that the presence of small red blood cells in the bloodstream is a symptom of diabetes mellitus, whereas high values of the distribution of red blood cells by volume are associated with cancer and heart disease. There is also a moderate strength association between this indicator of dementia in Alzheimer's disease.

Also interesting is the fact that the number of red blood cells has a positive, and the number of lymphocytes (one of the subpopulations of leukocytes) has a negative relationship with the risk of age-related diseases. What are white blood cells that are not lymphocytes? These include neutrophils, eosinophils, monocytes and basophils. An increase in the number of these cells is a harbinger of a deterioration in health. Neutrophils are the most numerous cells of the above. They are a component of innate immunity that provides protection against cancer and infections. Lymphocytes, on the other hand, include natural killers, as well as T and B cells. Natural killers also belong to innate immunity, whereas T- and B-lymphocytes are components of the adaptive immune system, but the state of populations of all these cells reflects the state of health and prospects for a long life.

All of these parameters were collected by Levin and her colleagues at their "phenotypic age". After that, the researchers moved on to the second stage, which consisted in searching for methylation sites that most strongly correlated with the values of the "phenotypic age" indicator they developed. 513 sites were included in their calculations. Thus, "PhenoAge" ("phenotypic age") is an indicator calculated from the 9 blood test indicators listed above and chronological age. And "DNAm PhenoAge" (DNAm – methylated DNA) is an aging clock based on DNA methylation, developed on the basis of a blood test "PhenoAge".

As a result, DNAm PhenoAge indicators correlate with chronological age by only 75% (compared to 94% for the original Horvath watches). However, DNAm PhenoAge predicts mortality and morbidity much better than the chronological age and the original Horvath clock. As expected, the methylation-based aging clock obtained with PhenoAge does not predict the probability of death as well as PhenoAge itself. This is expected, since the DNAm PhenoAge clock is designed to estimate the PhenoAge indicator and only indirectly predict mortality.

50 sites instead of 500

The first stage in the development of the watch is to compile a list of individual methylation sites for subsequent assessment of the strength of their relationship with age. If you create a watch based on several sites at the top of the list, you get the strongest correlation and the most accurate age estimate. However, this accuracy may be illusory. When selecting several parameters from a list of hundreds of thousands of them, exceptions, sharp deviations and statistical deviations will inevitably be obtained. An increase in the number of sites included will reduce the excessive influence of any of the sites on the resulting indicator. That is, if some of the correlations turn out to be statistical errors, the overall average value will still be fairly accurate. The Croat, in general, took a more conservative path, partly sacrificing accuracy in favor of reliability.

The future of the GrimAge Aging Watch

Methylation analysis turned out to be the most accurate method of assessing age and predicting the development of age-related diseases, significantly surpassing other approaches in accuracy. However, even more accurate results can be achieved by adding to the methylation of other facts known to us about a person, not other parameters of the blood test, but lifestyle factors. This approach formed the basis of a new generation of aging watches – GrimAge.