When does ageing become disease?

Our understanding of the molecular process of ageing is growing, as such our understanding of risk from specific ‘age-related’ diseases is becoming clearer says Professor Jamie Timmons.

In humans, chronological age defines the time since you were born. Whether you are considered old, within your community, depends on many factors. For example, when I grew up in Glasgow a 60-year-old man was very old, while where I sit today, writing this article, is Stockholm - where a 60-year-old man is positively middle-aged. An alternative definition for old-age, articulated by Sanderson and Scherbov (Sanderson & Scherbov, 2013) defines when, for example, you have 10 years of life expectancy left. Thus, the average life expectancy varies from community to community, reflecting a mix of genetics, environment, behaviours and stochasticity.

In epidemiological studies, particular those focused on the molecular mechanisms of ‘growing older’, loss of function and the emergence of ‘biomarkers’ of disease, even in young middle-aged ‘healthy’ adults, are often presented as diagnostics for human ageing (Belsky et al., 2015). From my perspective, this is almost certainly misleading as it implies that health, disease and longevity are all interchangeable synonyms for ageing.

If we wish to identify a definitive ‘ageing’ molecular programme (e.g. biological age), one that is independently informative for future health and life span then it is critical that we clearly define what is meant by the term ‘ageing’ and appropriately develop an assay that measures this parameter. We also have to consider if the developed diagnostic, while statistically significantly related to biological age, is sufficiently sensitive and specific enough to be considered a useful diagnostic (most will fail this final criteria e.g. telomere assays).

The other major consideration relates to how a novel diagnostic of ‘biological age’ would be used. If it were to be used as an independent diagnostic of longevity then it would be combined with other factors and behaviours that determine life-span, such as smoking and obesity. One could imagine the generation of an integrated risk ‘score’ utilised to determine insurance premiums for healthcare or to calculate pension requirements. These may seem controversial examples, but in reality our chronological age (birth year) and behaviours are already judged and used for these purposes. Why not have a more accurate ‘diagnosis’ of the contribution ‘age’ makes to these decisions?

In a medical setting ‘biological age’ could revolutionise medical screening.

For example, if you are a poor ‘biological age’ (for your chronological age) then your breast-cancer or prostate-cancer screening might be scheduled 5-10 yr earlier than average. If you have a youthful ‘biological age’ score at 50 years of age (chronological age) then screening could be delayed by several years. This could have a profound effect on the effectiveness of medical screening (which has its caveats). Shifting the cost-benefit ratio for a multitude of screens for medical conditions (i.e. disease) is certainly a priority given the uncertain effectiveness of existing strategies. I would argue one of the reasons behind this is the use of chronological age – which is a major epidemiological factor for chronic disease burden – as a tool to schedule screening. Epidemiological ideas rarely scale to n=1 medicine.

Thus, genomic technologies are facilitating unrivalled molecular analysis directly in humans allowing for the production of diagnostics and models of ‘ageing’ and the underlying biochemistry. The use of genotyping, DNA methylation analysis and transcriptomics has allowed the discovery of several models^9,2,6,10 which have prognostic or diagnostic performance when age-related diseases or longevity^5,10,11 are considered. However, to interpret and validate molecular models built from human clinical materials, one has to also consider the different aspects of human ageing, and try and distinguish it from ‘age-correlated disease’. This is where the majority of existing models of ‘biological age’ fail because they represent composite ‘signatures’ of age-related changes and the underlying pathophysiology present in the clinical cohorts utilised to build and replicate the models^{3, 6}.

Variation in the human transcriptome (RNA) has proven particularly powerful for identifying the huge variations in human physiology and physiological responses to environmental influences⁷. So it is not surprising it has been used to develop diagnostics of human ageing, including our own model (see below). In fact, when you closely examine the linear regression modelling used to create, for example, a model that relates blood RNA levels to age, and then to clinical disease, you find that ‘adjusting’ for covariates during the model building isn’t always helpful. This is a typical epidemiological strategy for claiming, incorrectly, that a factor is independently related to a phenotype and it actually introduces information into the ‘assay’ while removing genuine association with ageing phenotypes.

[caption id="attachment_51906" align="alignnone" width="620"] Genomic testing can more accurately predict a person's 'true' biological age.[/caption]

 

For example, Peters et al⁶ found that the concentration of hundreds of RNA molecules in the blood significantly correlated with ageing phenotypes (impaired cognitive function or memory for example) but after adjusting their model for various laboratory and clinical variables, they actually created a model that no longer related to this ageing phenotype. The remaining RNA model strongly associates with blood pressure and RNA that related to the protein synthesis machinery. One could conclude that this RNA diagnostic reflects some aspect of cardiovascular disease and perhaps even drug treatments they take (commonly used drugs like metformin modulate the protein synthesis machinery⁴ ) but it is unlikely to reflect a person’s ‘biological age’.

While you can’t use chronological age to diagnose the health status of an individual – the relationship between chronological age and disease is an epidemiological one – existing RNA or DNA methylation assays represent composites of ageing, disease and drug-treatment and not chronological age. We believe that ‘biological’ age will determine when you show clinical symptoms of disease and that we need an assay which accurately reflects your underlying ‘rate of ageing’ or ‘biological age’. Which ‘age associated’ disease an individual then develops will depend on their genetic, epigenetic and environmental risks factors (and stochasticity).

In a new study from our laboratory¹⁰ we generated what we believe to be the first reliable tool to define ‘biological age’ in humans; one that is very distinct from chronological age and one that does not correlate with simple life-style related disease. This latter point is important as while, for example, the prevalence of Type II diabetes increases with age, an individual can become diabetic at just about any age because as a disease it is largely driven by life-style factors (diet and exercise).

[caption id="attachment_51905" align="alignnone" width="363"] RNA markers can distinguish young and old.[/caption]

To produce this new diagnostic of ‘biological age’ we had the hypothesis that we can find a set of RNAs in the tissue that was diagnostic for telling tissue from healthy old from healthy young people apart. In our study healthy old people were living a normal sedentary lifestyle, did not have type II diabetes and importantly had good fitness levels. By applying machine learning to this ‘special’ healthy ageing cohort, we found 150 RNA markers which then were validated using fully external validation processes in 7 independent data-sets including muscle, skin and brain. The ROC plots show very high true positive (the method calls older samples 'old') and while it not so surprising this has been achieved across multiple muscle data sets, it was a surprise that the same marker genes work in each tissue.

In fact we could see that these 150 RNAs were either up or down regulated in tissue from healthy old people and we reasoned that activation of this gene expression ‘programme’ may help explain why these 65 year old people achieved good health despite living a sedentary life style. In fact, when we then applied the 150 RNA assay to a group of 70 year old people (people with the same chronological age) we found that their ‘biological age’ score varied dramatically and for those that failed to switch the gene expression pattern “on” as much died sooner and had a greater decline in organ function (kidney). In fact, the same RNA diagnostic was differentially regulated in blood RNA samples, and again if you failed to ‘activate’ the expression programme, you were more likely to belong to the group that demonstrated impaired cognitive function¹⁰.

Is this a diagnostic for Alzheimer’s disease? No, not yet.

To develop a new ‘one-off’ diagnostic for Alzheimer’s disease you need to compare your ‘molecular’ diagnostic with a clinical gold standard. Unfortunately, the current clinical diagnosis becomes definitive after the patient dies in the pathology lab, and thus you are relying on a diagnostic process that is perhaps 90% or so accurate (evidence of amyloid plaque build-up is not good enough on its own to diagnose cognitive disease).

What we have developed is a novel tool to help Alzheimer’s disease research. We can study people decades before they develop cognitive decline and ascertain if they are at greater risk given that it seems cognitive decline is a physiological capacity driven by ageing. Alternatively, when combined with other clinical and biochemical read-outs our ‘biological age’ test could improve current Alzheimer’s disease diagnostic procedures.  In particular there is a great need to carry out mass-screening of the older population to find those most relevant to enter into clinical trials that examine age-related diseases. If you can enrich the trial population, and increase the event rate, you make these trials more economically viable by reducing the size and potentially the duration of the studies.

In conclusion, how biomedical studies define the biology ageing must change, particularly when relying on model systems to evaluate anti-ageing strategies. Equally, molecular changes observed in older people with disease (and concurrent drug treatment) can’t inform us about the biology of ageing and terms such as ‘inflammation-ageing’ are naïve. Clearly, some ‘young’ people have diabetes and/or inflammation and hence neither will ever be diagnostic for ageing.  In short, measuring symptoms of age-associated disease is not the same as measuring ageing.

[caption id="attachment_51904" align="alignnone" width="486"] It is now possible to assess if someone is likely to develop Alzheimer's.[/caption]

The medical research challenge now is to understand how each persons risk profile for (prevalent) ‘disease’ interacts with their ‘biological’ age to determine which age-related illnesses they are most likely to suffer from (first). Clearly, for conditions like ‘cancer’ earlier identification through precision medicine can benefit the individual. How as a society we fund this or as an individual cope with knowing ‘our number’, is a much longer discussion.

The social challenge is broader. If we can now put a number on an individual’s ‘biological age’ then arguably we can apply the logic articulated by Sanderson and Scherbov, that defining how we treat and support an individual as they age might be better reflected on their project life expectancy than their chronological age. This means personalised health surveillance plans or personalised pension plans. Whether this is acceptable or embraced is an entirely different question.

The author:

Jamie Timmons is a professor of Genetics and Molecular Medicine at King’s College London. His research focuses on molecular predictors of insulin sensitivity and human muscle ageing.

References

1 Belsky DW, Caspi A, Houts R, Cohen HJ, Corcoran DL, Danese A, Harrington H, Israel S, Levine ME, Schaefer JD, Sugden K, Williams B, Yashin AI, Poulton R & Moffitt TE (2015). Quantification of biological aging in young adults. Proc Natl Acad SciE4104–E4110.

2 Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda S, Klotzle B, Bibikova M, Fan J-B, Gao Y, Deconde R, Chen M, Rajapakse I, Friend S, Ideker T & Zhang K (2013). Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 49, 359–367.

3 Horvath S (2013). DNA methylation age of human tissues and cell types DNA methylation age of human tissues and cell types. Genome Biol 14, R115.

4 Larsson O, Morita M, Topisirovic I, Alain T, Blouin M-J, Pollak M & Sonenberg N (2012). Distinct perturbation of the translatome by the antidiabetic drug metformin. Proc Natl Acad Sci 109, 8977–8982.

5 Marioni RE, Shah S, McRae AF, Ritchie SJ, Muniz-Terrera G, Harris SE, Gibson J, Redmond P, Cox SR, Pattie A, Corley J, Taylor A, Murphy L, Starr JM, Horvath S, Visscher PM, Wray NR & Deary IJ (2015). The epigenetic clock is correlated with physical and cognitive fitness in the Lothian Birth Cohort 1936. Int J Epidemiol 44, 1388–1396.

6 Peters MJ et al. (2015). The transcriptional landscape of human age. Nat CommunIn submission.

7 Phillips BE, Williams JP, Gustafsson T, Bouchard C, Rankinen T, Knudsen S, Smith K, Timmons JA & Atherton PJ (2013). Molecular Networks of Human Muscle Adaptation to Exercise and Age ed. Gibson G. PLoS Genet 9, e1003389.

8 Sanderson WC & Scherbov S (2013). The Characteristics Approach to the Measurement of Population Aging. Popul Dev Rev 39, 673–685.

9 Sebastiani P, Solovieff N, Dewan AT, Walsh KM, Puca A, Hartley SW, Melista E, Andersen S, Dworkis D a, Wilk JB, Myers RH, Steinberg MH, Montano M, Baldwin CT, Hoh J & Perls TT (2012). Genetic signatures of exceptional longevity in humans. PLoS One 7, e29848.

10 Sood S, Gallagher IJ, Lunnon K, Rullman E, Keohane A, Crossland H, Phillips BE, Cederholm T, Jensen T, van Loon LJC, Lannfelt L, Kraus WE, Atherton PJ, Howard R, Gustafsson T, Hodges A & Timmons JA (2015). A novel multi-tissue RNA diagnostic of healthy ageing relates to cognitive health status. Genome Biol 16, 185.

11 Zhang WB & Pincus Z (2015). Predicting all-cause mortality from basic physiology in the Framingham Heart Study. Aging Celln/a – n/a.