Perspective https://doi.org/10.1038/s43587-020-00015-1

Asynchronous, contagious and digital aging

Thomas A. Rando1,2,3 ✉ and Tony Wyss-Coray 1,3,4 ✉

Organismal aging is often characterized as a steady, monotonic decline of organ and tissue function. However, recent studies indicate spatial and temporal variations of aging rates across the lifespan. We consider these variations from the perspective of underlying cellular changes. Cells in certain tissues may age earlier and produce signals that accelerate the aging of other cells, locally or distantly, acting as drivers for organismal aging and leading to a lack of synchronous aging between tissues. As cells adopt new homeostatic states, cellular aging can be viewed, at least in part, as a quantal process we refer to as digital aging. Analog declines of tissue function with age may be the sum of underlying digital events. Cellular aging, digital or otherwise, is not uniform across time or space within organisms or between organisms of the same species. Advanced systems-level and single-cell methodologies will refine our understanding of cell and tissue aging, and how these processes integrate to produce the complexities of individual, organismal aging.

volutionary theories of aging have sought to explain why we from features of cellular aging that might not have been predicted age, with concepts evolving over the decades from genetic from first principles. and deterministic to more physiological and adaptive theo- E1 ries . Among the most widely accepted theories currently are those Lifespan versus aging involving accumulation, antagonistic pleiotropy and One of the reasons for a limited understanding of the determinants physiological trade-offs between the mortal soma and the immor- of individual aging is that the broader research field of aging and tal germline (Box 1)2–4. Such theories provide a basis for under- has relied heavily on studies in model organisms in which standing aging as a result of processes whose manifestations are lifespan, and not aging itself, has been the predominant focus and subject to evolutionary selection for reasons other than aging primary benchmark. Studies of the determinants of lifespan within traits themselves. However, these theories shed little light on a species have been highly productive because of the numerous variations of the rate of aging within individuals of a species available short-lived model systems9,10, as well as the relative ease (Fig. 1). In this Perspective, we will discuss challenges to of the assay; namely, counting the number of animals that are alive the studies of aging and recent findings that highlight features versus the number that are dead. This has led to a growing list of of organismal aging that pertain to how individuals rather than genes, drugs and environmental factors (including diet) that can species age. influence the median and maximum lifespans of certain species11–13. Throughout, we will refer to aging as the underlying processes In contrast, the identification of the determinants that lead that lead to the cellular, tissue and organismal phenotypes associ- to an acceleration of aging in one member of a species and the ated with that phase of metazoan life known, in the vernacular, as slowing of aging in another has proved to be more challenging. old age. Specifically, these manifestations are the hallmark of that While lifespan is a measure that incorporates the impacts of aging, phase of life that follows the period of development and postnatal the determinants of longevity are distinct from those that influence growth, that follows a period of relative homeostasis during adult the rate at which an organism ages14. Furthermore, lifespan studies life associated with reproductive maturity, and that is character- are conducted in laboratory settings that allow for prolonged peri- ized primarily by a decline in structure and function. This third ods of protected aging15, just as human societies do, thereby mask- phase is the raison d’être for the field of geroscience5, and is also ing many increased risks of dying as a result of functional declines characterized in humans by a dramatic increase in the incidence with age that would be evident in the wild. Therefore, it is possible of age-related diseases, including cancer, cardiovascular diseases for measured lifespan to be uncoupled, at least to a large extent, and neurodegenerative diseases. Although the manifestations of from the consequences of aging—the recording of the death of a aging are a feature of later life, their underlying cellular and molec- person provides no information as to how functional that person ular bases certainly begin earlier and, arguably, at least for some was over the previous decades. While there is clearly an interre- molecular changes, even during development6. However, such lationship between aging and longevity, they are not inextricably molecular events can be identified as potential underpinnings of linked16. Evidence in has provided support aging phenotypes only in retrospect, and all of these occur even in for this idea by demonstrating that genetic interventions that extend organisms with negligible senescence7. As such, no unified theory of lifespan may do so without comparably altering the relevant aging the molecular basis of cellular aging has emerged, primarily because phenotypes17. As such, it is imperative that quantitative measures of of the myriad of interacting contributions8, and we do not attempt aging itself—so called biomarkers—be increasingly used in order to to define aging, its potential causes or the time when the underly- understand the complex processes that underlie the manifestations ing changes begin (Box 2). We focus primarily on how the mani- of organismal aging. In addition, quantitative measures of aging will festations of aging, at the organismal level, reflect variants in the allow more rapid study of interventions to slow the rate of aging rate of tissue aging across space and time, and how they may arise of long-lived species, including humans, in which longevity studies

1Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, CA, USA. 2The Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, USA. 3Glenn Center for the Biology of Aging, Stanford University School of Medicine, Stanford, CA, USA. 4Wu Tsai Neurosciences Institute, Stanford University, Stanford, CA, USA. ✉e-mail: [email protected]; [email protected]

Nature Aging | VOL 1 | January 2021 | 29–35 | www..com/nataging 29 Perspective NAture Aging

Box 1 | Evolutionary theories of aging

As the renowned geneticist and neo-Darwinian evolutionary the- switch later in life because the expression of these genes early in orist Teodosius Dobzhansky noted by the title of his 1973 essay, life has actually been selected for based on their benefcial efects. “Nothing in biology makes sense except in the light of evolution”84. Hence, the pleiotropy is the diferent traits in early and late life, Tis is particularly germane to the biology of aging since all mod- and the antagonism is the benefcial versus detrimental efects of ern evolutionary theories of aging begin by positing that aging those traits. Williams himself proposed a theoretical example: a traits are largely, if not entirely, outside of evolutionary selective mutation may favor calcifcation of bone during development but, pressure. Tis is based on the principle that the forces of natural in a diferent somatic environment (that is, the aged soma), leads selection diminish with age because of the myriad causes of ex- to calcifcation of the connective tissues of arteries. Te disposable trinsic mortality in the wild. Tus, the challenge has been to as- soma theory of Kirkwood4 takes a more physiological perspective cribe the features of aging to evolutionary processes that could, at to propose that an organism is able to shif allocation of its own least in theory, account for their existence. Early theories centered resources to either reproduction or to processes that maintain the around the idea that aging is genetically programmed and that soma (for example, growth, homeostasis, nutrient storage and so aging is somehow adaptive for the species. Tis notion has been on). In that sense, it is not a purely genetic theory, as those that increasingly discounted as evidence has failed to provide support came before were, and does not invoke genomic or the and as subsequent theories have suggested plausible accounts for efects of specifc alleles to explain cell and tissue . While aging phenotypes based on processes that are under the infuence previous theories certainly set aside the germline as a specifc case of natural selection and those that are not. separate from the soma in terms of aging, the disposable soma Te mutation accumulation theory, which is attributed to theory directly incorporates the germline as a critical component Medawar2, postulates that mutations accumulate in the germline of the equation about allocation of resources. Te theory rests on over generations, and that these mutations are largely, if not the notion of physiological trade-ofs in the context of broader entirely, insignifcant during development, growth and early genetic programs that support somatic maintenance and the adulthood (and are thus outside of evolutionary selection) but failure of those programs, which accounts for aging phenotypes. manifest themselves as deleterious in the post-reproductive Tis failure varies from species to species depending on the period. Tis theory posits that the expression of these deleterious environment and the need to allocate resources to the soma versus genes either increases with age or is switched on later in life. the germline. Of course, these are all theories that will evolve with To date, few examples have been reported that support this increasing evidence and further understanding of the relationship pattern. In some senses, the antagonistic pleiotropy theory of between genetic programs, the response of those programs to Williams3 expands on this idea of late-life deleterious efects of environmental infuences, and the susceptibility to functional gene expression but posits that there is no increase over time or decline in later life that we recognize as aging.

are extraordinarily difficult and expensive. This, then, leads to the rates of aging) as determined by lifespan22. Identifying which factors concept of the rate of aging. (especially if not genetic and not environmental) determine individ- ual rates of aging remains a major challenge to the field. However, Rates of aging equally interesting and provocative is the question of whether there Studies of the rate of aging have been entwined with the study of are different rates of aging within a single individual, and what fac- variations in lifespan across species. The notion of differences of tors determine those rates (Fig. 1). We will expand on these ques- rates of aging among species is related to the rate of living theory tions below. of aging, which posits that the maximum lifespan of a species is At the level of an individual, a rate of aging can be determined determined, inversely, by its metabolic rate18. According to this the- by the rate of change of virtually any anatomical structure or physi- ory, species with a high metabolic rate age rapidly and die sooner, ological process (for example, bone density, renal function and so whereas species with a low metabolic rate age more slowly and die on for humans), as there are almost no aspects of anatomy or physi- later. While generally correlated, there are many exceptions (nota- ology that do not exhibit declines as an individual ages23. It is from bly, flying species, such as birds and bats, and a wide variety of euso- such observations that the notion of biological age as opposed to cial animals). Within the animal kingdom, maximum lifespans vary chronological age emerges; measurement of any such biomarker can by more than five orders of magnitude19, with the current world identify an individual as being either biologically older or younger record holder being a 507-year-old quahog clam20. Assuming that than the mean for members of the species of the same chronological nearly all animals in a protected environment will age before dying, age. Clearly, readily obtainable physiological measures or molecular it is almost tautological that members of a species whose maximum biomarkers are essential to studies of biological age. A vast num- lifespan is 100 days age more rapidly than those from a species ber of biomarkers, ranging from DNA methylation24,25 to whose maximum lifespan is 100 years. In contrast, comparisons of length26,27 to plasma composition28 to facial morphologies29, rates of aging among individuals within a given species shine light have been used to calculate biological age. Of course, any one of on variations that are independent of the maximum lifespan, and these measures produces a snapshot, whereas it is the changes of the genetic determinants thereof, of that species. Within any given these biomarkers over time that reveal trajectories and rates of species, there will be individuals that age more rapidly and those aging. With such measures of age-related changes, composite scores that age more slowly. In humans, this is brought into sharp relief can be generated to provide an overall rate of organismal aging to in older populations. For example, among centenarians, for whom predict, for example, health trajectory or . a main focus has been their relative propensity to avoid diseases Based on the idea that measuring changes in biomarkers from of aging and age-related declines, there is tremendous variability one age to another allows for the calculation of a rate of aging, there in functional status21. Furthermore, even for genetically identical are several key aspects to those measurements that will determine organisms maintained in identical environments, there remains how relevant that calculated rate is for members of a species. A cen- large variability of aging trajectories (presumably related to different tral feature is the definition of the starting point. It is essential to

30 Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging NAture Aging Perspective

Cells Organs Individuals Cell 1

Heart Cell 2 Person 1

Cell 3 Brain Person 2 Cell 4

Kidney

Person 3

Fat

Person 4

Fig. 1 | Lifespan as an integration of rates of aging at multiple levels. Graphical representation of rates of aging with boxes representing aging in individual cells, organs and persons. Cellular stress results in distinct homeostatic cellular states. These digital cellular states (as described later in the text) are integrated to give rise to distinct rates of aging across organs, ultimately leading to analog functional readouts that differ among organs and thus among individuals. The x axis represents time/age and the y axis represents the magnitude of change for a given box.

Box 2 | What is aging? the rate of change. As a trivial example, if height were used as a biomarker in humans, a starting point of age 5 would lead to the erroneous conclusion, regardless of what endpoint was used, that Few questions evoke greater debate and disagreement in the height increases during aging, whereas we know that height in fact feld of the biology of aging than ‘what is aging?’. Tis is followed decreases over the decades of old age. This point highlights a second closely in terms of contentiousness by the highly related ques- key aspect of measurements of rates of change of biomarkers, and tions ‘what causes aging?’ and ‘when does aging begin?’. For the that is that the accuracy of the rate, and especially changes in rate, purposes of this Perspective, we have chosen to focus on pheno- will increase as the number of longitudinal measurements increase. types of aging as enumerated in the text. But is this aging? Tis In fact, as discussed below, some of the recent findings that shed is a semantic issue, but a central one. For example, is Alzheimer’s light on unexpected temporal patterns of aging were discerned only disease the clinical syndrome of dementia, or is Alzheimer’s dis- because of multiple, serial biomarker measurements. Finally, cau- ease the pathology of amyloid plaques and neurofbrillary tan- tion should be exercised when using biomarkers measured imme- gles? Clearly it is possible to have a syndrome of dementia with- diately before death, particularly in humans, because of the many out these specifc pathological changes, and clearly there can be confounding aspects of comorbidities that might yield highly vari- widespread pathology without the clinical syndrome. As such, able results and thus skew trajectories. for sporadic Alzheimer’s disease (unlike the rare genetic forms), the cause and the onset remain enigmatic. For aging, the defni- Variations of individual aging rates across time and space tion is even more problematic, and answers to questions about Using serial changes in biomarkers, whether molecular, anatomi- the cause and onset are even more elusive for one simple rea- cal or physiological, allows for a definition of a rate of aging for an son—aging is universal (or nearly so). individual. This also allows for an assessment of whether that rate Clearly, phenotypes of aging arise as a result of underlying of aging is constant over time, or at least monotonic. In general, the molecular processes, but unlike genetic diseases that are decline of tissue function is commonly illustrated as gradual and unequivocally due to genetic mutations, there is no obvious or steady23, perhaps with a more rapid decline towards the end of life. comparable starting point. Rather, we consider the processes However, is this how aging actually proceeds? Ultrastructural studies that lead to cellular aging phenotypes to be ones of system in C. elegans almost two decades ago hinted that it might not30. The failures—failures that can arise from dysfunction of diverse and authors noted not only vastly different rates of aging of tissues and interconnected organelles and other subcellular structures, which cells in individual worms, which they ascribed largely to stochastic in turn can arise from diverse and interconnected molecular events, but also that the onset of degenerative changes was neither defects at the levels of DNA, RNA, protein and metabolites. Tis linear nor synchronized. Mathematical modeling of longevity curves Perspective maps onto the notion of interacting layers of aging in flies also suggest at least two phases of aging, with a sharp tran- 85 that has been proposed . Te identifcation of proximal causes sition in between states characterized by intestinal degeneration31. of aging phenotypes is tractable. Te search for the cause of aging This discontinuous behavior of aging is becoming clearly evident in is akin to the search for the beginning of a circle. Te search recent large-scale omics studies in mice and humans. Transcriptional for the inception of aging is akin to the search for the onset of changes in 17 organs across ten ages from neonatal to 30-month-old homeostasis. So, we have taken the easy way out and avoided mice showed very different trajectories of aging with varied onset these questions (for now). and amplitudes of changes32. In humans, a meta-analysis of >300 brains and 17 brain regions showed broad nonlinear changes in know that the biomarker is no longer changing along the trajectory gene expression that varied across brain regions with age33. Similarly, of development and growth but has reached some steady state as a DNA methylation, which shows a remarkable correlation with aging starting point, particularly when only two ages are used to assess across species, was reported to be 24 times faster during the first

Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging 31 Perspective NAture Aging

Development Reproductive phasePost-reproductive phase

Cellular aging

Fat

Spleen

Kidney

Thymus

Heart

Fig. 2 | Asynchronous aging of different tissues and organs. Differential rates of aging at the cellular level lead to organ aging at different rates and at different stages of life. Aging may start early in life in every tissue, but different tissues may age faster than others. The different rates of aging between tissues depicted here for illustrative purposes are based on transcriptomic changes across 17 organs in aging mice, as described recently32. few years of life compared with after puberty34. Inflection points in be certain organs and tissues that tend to be the most susceptible to DNA methylation indicative of non-monotonous changes were also functional decline with age, which is the actual driver of organismal observed in a 20-year follow-up Swedish aging cohort of 845 indi- aging for any given member of that species is highly variable. viduals around the ages of 70 and 85 years35. A recent study of some If the most rapidly aging tissues are the main drivers of organ- 3,000 in the blood of over 4,000 healthy people concluded ismal aging, it naturally begs the question of ‘what are the intrinsic that markers of aging do not in fact suggest a monotonic aging pro- changes to the cellular and extracellular matrix constituents that cess, but instead show periods of accelerated change and periods drive the aging of those tissues?’. Within any tissue, changes in the of slower change28. Peripheral blood mononuclear cells exhibit two cellular composition could be a result of changes in the proportions distinct peaks of rapid epigenomic changes in humans: one around of resident cells (such as, for example, an increase in the proportion 40 years of age and another around 60–70 years of age36. Likewise, a of fibroblasts or resident macrophages) or due to intrinsic changes longitudinal study of the transcriptome and DNA methylation state to subpopulations without any change in cellular proportions. of skin in 121 humans between 21 and 76 years of age revealed four Analysis of bulk populations of cells may fail to reveal such a dis- distinct phases of aging highlighted by the activity of different bio- tinction, but single-cell analysis has the potential to reveal the rela- logical processes37. Interestingly, functional MRI measurements of tive contribution of these two distinct, but of course not mutually brain connectivity seem to support these asynchronous molecular exclusive, processes of tissue aging. In a groundbreaking single-cell measurements, showing nonlinear changes and transition points study, RNA isolated from individual cardiomyocytes of young and around the ages of 40 and 70–75 years of age38. old mice showed broad mosaicism in gene expression among cells In addition to these variations in rates of organismal aging within individual mice and increased differences in expression over time, the above studies also provide clear evidence for the lack between cells45. Isotope-based dating of DNA and protein at the of uniformity of underlying aging processes across space; namely, ultrastructural level in various tissues in mice revealed extensive cel- among tissues in an individual (Fig. 2). Undoubtedly, for any given lular mosaicism within aging tissues46. Cells of the same type have species, there are certain tissues and organs whose functions decline vastly different ages based on DNA dating (including postmitotic earliest and most rapidly. For example, in humans, the cardiovascu- cells), with some as old as the animal, and subcellular structures lar system exhibits a much greater functional decline with age than within individual neurons or pancreatic β cells have different ages the gastrointestinal system23, at least based on physiological param- based on protein dating. These observations are supported by a rap- eters measured, as noted above, on absolute scales. In Drosophila, idly growing number of single-cell RNA sequencing (scRNA-seq) gut and skeletal muscle dysfunction appear to be primary drivers studies showing greater cell-to-cell variability in gene expression in of aging and mortality39–42. As mentioned above, transcriptomic human T cells and pancreatic cells, as well as differences in cellular studies strongly agree that tissues, organs and even subre- mutation rates with age47,48. A recent mouse aging cell atlas, Tabula gions within an organ such as the brain age at different rates32,33. Muris Senis, further underlines this finding across 23 tissues and This is consistent with findings from studies in rats of organ-specific organs49. In addition, the atlas, as well as the related tissue RNA bulk changes in the transcriptome and proteome with age43. In addition, study32, demonstrates a change of the composition of the cellular tissues within human and mouse individuals show unique DNA constituents. Notably, multiple tissues exhibited an increase in the methylation rates25,34. Thus, spatially, there is clear heterogeneity of numbers and percentages of pro-inflammatory immune cells as a aging rates among tissues, and the aging of members of one spe- function of age—a conclusion independently drawn in single-cell cies may be driven by functional loss of tissues that are different transcriptomic surveys of seven tissues in aging rats and three from those in another species. However, is it the case that, within tissues in aging mice50,51. a species, the relative rates of aging of different tissues and organs are the same from one individual to another? In one pioneering Digital aging versus analog aging human study, a multi-omics analysis of aging biomarkers suggested With regard to intrinsic changes to cells as a driver of tissue aging, that different individuals manifest different ageotypes, mean- a particular cell type in a tissue could change gradually over time ing that the tissue that ages most rapidly differs from one person in a process of analog aging, analogous to how we characterize to another44. Therefore, for any given species, although there may organismal aging. Indeed, in the scRNA-seq study mentioned above,

32 Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging NAture Aging Perspective

Birth Adult Aged Death

Fig. 3 | Digital aging. Cells from the same organ, born at the same time, show distinct temporal patterns of functional decline. These functional changes can be viewed as occurring in digital steps corresponding to discreet states of dysfunction (blue and yellow), resulting in different homeostatic organ states over time. Cells do not necessarily follow the same sequences of digital steps or acquire the same levels of dysfunction before they die, shown here as the adoption of brown, apoptotic features. parenchymal cells in multiple tissues exhibited progressive changes therapy is clearly aimed at enhancing healthspan by cells in a par- in gene expression with age49. However, at the cellular level, the pos- ticular state, not all cells in an aged tissue56. Aging is associated sibility of a certain amount of digital aging needs also to be con- with changes in the population of immune cell infiltrates51, and sidered (digital in the sense of discrimination by discrete units, not caloric restriction can reverse those trends50. This is not an effect necessarily binary units) (Fig. 3). Whereas the progression of tissue on the aging states of the parenchymal cells but certainly results aging suggests an analog process with such measures as physiologi- in age-related changes at the tissue level as a consequence of cell cal function, gene expression levels and protein aggregates chang- composition. Whereas transcriptional signature changes are com- ing gradually, we know that there are at least some discrete state mon among various lifespan extension interventions57, it remains to transitions of cells in a tissue that can underlie this apparent analog be determined by single-cell analysis whether these changes reflect decline. For example, for the thymus, heart and other tissues52,53, cell composition changes of a tissue as opposed to distributed gene there is a progressive decline in cell number with age due to apop- expression changes among resident cells. totic cell death. The death of a cell can be considered one such digital readout; cells are either alive or dead. Likewise, cellular senescence Is aging contagious? is another digital cellular state in the sense that cells can be classified Given the complex heterogeneities of cell and tissue aging in any by various markers as being either senescent or non-senescent54. The single individual and the notion of the most rapidly aging tissues extent to which the age-dependent phenotype of a tissue is due to being the driver of the aging of that organism, do those more rapidly the burden of senescent cells provides another example of a gradual aging tissues accelerate the aging of other tissues in the body? Does process being determined by the relative levels of cells in one state the aging of one cell affect the age of another cell? Is aging conta- or another. Indeed, based on scRNA-seq data, the proportion of gious? The notion of the aging process spreading from one cell to cells expressing the senescence marker p16 more than doubled in a another is highlighted, again, by the field of cellular senescence. The number of tissues of old mice compared with young mice49. None of secretome of senescent cells has been shown to induce senescence these transitions occurs instantaneously, and there are clearly inter- of neighboring cells58. In that sense, there can be cellular leaders that mediate states. Nevertheless, such cellular states are routinely deter- accelerate the aging of other cells in the tissue. mined as discrete and quantal rather than continuous. Even in the The notion of cell-to-cell spreading of cellular dysfunction is of case of senescence, where multiple different senescent states have course not limited to the biology of senescence. This is becoming an been defined55, they are still modeled as discrete. With this model increasingly recognized phenomenon in the pathogenesis of neu- taken to its extreme, the analog decline of the function of a tissue can rodegenerative diseases59. In many diseases, including Alzheimer’s be modeled as a result of the gradual accumulation of cells that have disease, Huntington’s disease and Parkinson’s disease, a cardinal transitioned from a functional to one or more dysfunctional states. feature of the pathology is intracellular aggregation of proteins. Further single-cell studies will delineate the relative contribution of While seemingly a cell-intrinsic phenomenon, one of the curious analog versus digital cellular aging as drivers of tissue aging. features of the pathology of these diseases is the apparent spread of This notion of aging by quantal steps can clearly be extended the cellular abnormalities to anatomically connected brain regions. to the subcellular world to explore proximal causes of age-related The general concept of this kind of spreading proteinopathy from changes in cellular phenotypes. Proteins can be appropriately folded one cell to another, locally, arises from the biology of prions and or misfolded, RNA can be correctly spliced or mis-spliced, and DNA prion diseases60. Of course, some prions are truly contagious, in the can have single base changes. The extent to which cellular aging is sense of being transmissible between individuals or across species, due to digital versus analog changes at the molecular level is certain but the spread within the central nervous system of an individual to depend on which cellular phenotype is under consideration in suggests cell-to-cell spread61. As with senescence, this phenomenon the same way that different tissues may exhibit age-related changes could represent the conversion of cells from one state (free from as a result of digital cellular aging to different degrees. aggregates) to another (aggregate-laden) since protein aggregation Likewise, it will be interesting to consider whether interven- can be self-propagating. As protein aggregation is one of the key tions that extend lifespan or healthspan act by shifting the balance features of cellular aging, it is intriguing to consider the possibility between digital states at the cellular level. As an example, senolytic of aged cells achieving a sufficiently dysfunctional state as a result of

Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging 33 Perspective NAture Aging protein aggregation, then conferring an aging signal to nearby cells Conclusion through non-cell autonomous regulation of proteostasis62. As the numbers of distinct measures and biomarkers of aging If aging is indeed contagious, is the spread restricted to neighbor- increase and as the technologies to quantify them develop, the pic- ing cells or might it spread to distant tissues via the systemic circu- ture that emerges of the patterns of aging will continue to become lation? Based on early work from our laboratories that ushered in a increasingly more complex. Clearly, these temporal patterns cannot new era of the use of the technique of parabiosis in aging research, be discerned from lifespan studies, and distinct patterns of aging it is clear that systemic factors originating from distant tissues in the among different tissues (as illustrated in Fig. 1) cannot be discerned body are able to either promote or reverse cell and tissue aging phe- from samples of whole model organisms. Likewise, individual pat- notypes63–65. These findings, as well as many follow-up studies66–68, terns of aging cannot be understood from pooled samples. We posit including the demonstration that plasma infusions alone are suffi- that the increased complexity that has emerged (and will continue cient to exert these effects65,69, have unequivocally demonstrated that to emerge) will redirect attention away from singular and linear factors in the blood are able to communicate information from one or causes of aging and towards a more unified model of both intrinsic more source tissues to other tissues throughout the body. These could (for example, oxidative damage, protein aggregation and genomic potentially accelerate, delay or even reverse the rate of aging of other instability) and extrinsic (for example, inflammation, hypoxia and tissues in the body. Indeed, scRNA-seq studies of brain endothelial nutritional deprivation) stresses impinging on cells whose responses cell aging showed that infusion of aged plasma can accelerate while are dictated by genetically determined homeostatic response. Such a young plasma can reverse aging as determined by analysis of the tran- model provides a framework for incorporating findings that high- scriptome70. These studies highlight the fact that cellular aging does light phenotypes of aging that deviate from models of uniform and not occur independent of influences that are both local and systemic. synchronized processes throughout an organism.

Aging and disease: chicken and egg? Received: 11 August 2020; Accepted: 3 December 2020; Together, these findings highlight another emerging concept in aging Published online: 14 January 2021 research. It has long been known that the major risk factor for the vast majority of chronic diseases is age itself71. Without a doubt, the aging process renders cells and tissues more susceptible to the patho- References physiological processes that underlie these diseases. However, the 1. Flatt, T. & Partridge, L. Horizons in the evolution of aging. BMC Biol. data on the acceleration of aging by blood-borne factors also lead to 16, 93 (2018). the converse question: is having a chronic disease a major risk fac- 2. Medawar, P. B. An Unsolved Problem of Biology (H. K. Lewis, 1952). tor for accelerated aging? That is, are diseases risk factors for aging? 3. Williams, G. C. Pleiotropy, natural selection, and the evolution of senescence. Evolution 11, 398–411 (1957). As a disease engulfs a tissue, it is likely to lead to local and systemic 4. Kirkwood, T. B. Evolution of . Nature 270, 301–304 (1977). changes that, themselves, may promote dysfunction on neighboring 5. Kennedy, B. K. et al. Geroscience: linking aging to chronic disease. Cell 159, and distant cells. Epidemiological data do suggest that individuals 709–713 (2014). with various chronic diseases such as cancer, diabetes and HIV/AIDS 6. Gladyshev, V. N. Te ground zero of organismal life and aging. Trends Mol. are more likely to develop comorbidities later in life72–74, consistent Med. 27, 11–19 (2020). 7. Rose, M. R. Adaptation, aging, and genomic information. Aging 1, with the notion that one disease (or its treatment) may accelerate 444–450 (2009). aging processes more generally, thus rendering individuals more sus- 8. Gladyshev, V. N. Aging: progressive decline in ftness due to the rising ceptible to other diseases, just as aging itself does. In this sense, and deleteriome adjusted by genetic, environmental, and stochastic processes. related to the notion of the spread of aging described above, aging Aging Cell 15, 594–602 (2016). begets aging. How this process is modulated within tissues and across 9. Tissenbaum, H. A. & Guarente, L. Model organisms as a guide to mammalian aging. Dev. Cell 2, 9–19 (2002). tissues to prevent an exponential increase in the rate of aging over 10. Piper, M. D. W. & Partridge, L. Drosophila as a model for ageing. Biochim. time may also relate to the ability of cells to adopt new, relatively sta- Biophys. Acta Mol. Basis Dis. 1864, 2707–2717 (2018). ble homeostatic states even in the face of extrinsic stresses. 11. Singh, P. P., Demmitt, B. A., Nath, R. D. & Brunet, A. Te genetics of aging: a Targeting cellular and tissue drivers of organismal aging, or their vertebrate perspective. Cell 177, 200–220 (2019). 12. Gonzalez-Freire, M. et al. Te road ahead for health and lifespan means of spreading their aging messages to nearby and distant cells interventions. Ageing Res. Rev. 59, 101037 (2020). and tissues, holds the promise of extending healthspan by slow- 13. Fontana, L., Partridge, L. & Longo, V. D. Extending healthy life span—from ing the overall rate of aging. If successful, this would not only slow yeast to humans. Science 328, 321–326 (2010). the structural and functional declines that are the quintessential 14. Hayfick, L. Entropy explains aging, genetic determinism explains longevity, features of organismal aging, but would also reduce the incidence and undefned terminology explains misunderstanding both. PLoS Genet. 3, e220 (2007). of age-related diseases. Pharmacological and dietary interven- 15. Goodell, M. A. & Rando, T. A. Stem cells and healthy aging. Science 350, tions that extend lifespan in model organisms and may do so by 1199–1204 (2015). slowing aging rates are increasingly being developed for testing 16. Brett, J. O. & Rando, T. A. Alive and well? Exploring disease by studying in humans75. However, it is likely that, for humans, a considerable lifespan. Curr. Opin. Genet. Dev. 26, 33–40 (2014). degree of personalized approaches will be necessary. Even among 17. Bansal, A., Zhu, L. J., Yen, K. & Tissenbaum, H. A. Uncoupling lifespan and healthspan in Caenorhabditis elegans longevity mutants. Proc. Natl Acad. Sci. highly genetically similar strains of yeast or mice, dietary restric- USA 112, E277–E286 (2015). 76–78 tion is equally likely to shorten lifespan as to extend it . A more 18. Pearl, R. Te Rate of Living (University of London Press, 1928). thorough understanding of the molecular processes of aging, the 19. Rando, T. A. Stem cells, ageing and the quest for immortality. Nature 441, cellular basis of tissue aging and the local and systemic communi- 1080–1086 (2006). cation networks that can accelerate or decelerate the rate of aging 20. Butler, P. G., Wanamaker, A. D., Scourse, J. D., Richardson, C. A. & Reynolds, D. J. Variability of marine climate on the North Icelandic Shelf will certainly increase the number and diversity of interventions in a 1357-year proxy archive based on growth increments in the bivalve to extend healthspan. Furthermore, the ability of interventions not Artica islandica. Palaeogeogr. Palaeoclimatol. Palaeoecol. 373, 141–151 only to slow the rate of aging but to actually rejuvenate cells and tis- (2013). sues seemingly by reversing aging processes expands the spectrum 21. Ailshire, J. A., Beltran-Sanchez, H. & Crimmins, E. M. Becoming of healthspan-promoting treatments79–83. Coupled with technologi- centenarians: disease and functioning trajectories of older US adults as they survive to 100. J. Gerontol. A Biol. Sci. Med. Sci. 70, 193–201 (2015). cal advances to measure aging phenotypes and rates of aging at the 22. Kirkwood, T. B. et al. What accounts for the wide variation in life span of cellular, tissue and organismal levels, this will also probably increase genetically identical organisms reared in a constant environment? Mech. the pace of the translation from the laboratory to the clinic. Ageing Dev. 126, 439–443 (2005).

34 Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging NAture Aging Perspective

23. Khan, S. S., Singer, B. D. & Vaughan, D. E. Molecular and physiological 59. Davis, A. A., Leyns, C. E. G. & Holtzman, D. M. Intercellular spread of manifestations and measurement of aging in humans. Aging Cell 16, protein aggregates in neurodegenerative disease. Annu. Rev. Cell Dev. Biol. 34, 624–633 (2017). 545–568 (2018). 24. Bocklandt, S. et al. Epigenetic predictor of age. PLoS ONE 6, e14821 (2011). 60. Prusiner, S. B. Biology and genetics of prions causing neurodegeneration. 25. Stubbs, T. M. et al. Multi-tissue DNA methylation age predictor in mouse. Annu. Rev. Genet. 47, 601–623 (2013). Genome Biol. 18, 68 (2017). 61. Guo, J. L. & Lee, V. M. Cell-to-cell transmission of pathogenic proteins in 26. Heidinger, B. J. et al. Telomere length in early life predicts lifespan. Proc. Natl neurodegenerative diseases. Nat. Med. 20, 130–138 (2014). Acad. Sci. USA 109, 1743–1748 (2012). 62. Morimoto, R. I.Cell-nonautonomous regulation of proteostasis in aging and 27. Steenstrup, T. et al. and the natural lifespan limit in humans. Aging disease.Cold Spring Harb. Perspect. Biol. 12, a034074 (2020). 9, 1130–1142 (2017). 63. Conboy, I. M. et al. Rejuvenation of aged progenitor cells by exposure to a 28. Lehallier, B. et al. Undulating changes in human plasma proteome profles young systemic environment. Nature 433, 760–764 (2005). across the lifespan. Nat. Med. 25, 1843–1850 (2019). 64. Brack, A. S. et al. Increased Wnt signaling during aging alters muscle stem 29. Chen, W. et al. Tree-dimensional human facial morphologies as robust aging cell fate and increases fbrosis. Science 317, 807–810 (2007). markers. Cell Res. 25, 574–587 (2015). 65. Villeda, S. A. et al. Te ageing systemic milieu negatively regulates 30. Herndon, L. A. et al. Stochastic and genetic factors infuence tissue-specifc neurogenesis and cognitive function. Nature 477, 90–94 (2011). decline in ageing C. elegans. Nature 419, 808–814 (2002). 66. Ruckh, J. M. et al. Rejuvenation of regeneration in the aging central nervous 31. Tricoire, H. & Rera, M. A new, discontinuous 2 phases of aging model: system. Cell Stem Cell 10, 96–103 (2012). lessons from . PLoS ONE 10, e0141920 (2015). 67. Salpeter, S. J. et al. Systemic regulation of the age-related decline of pancreatic 32. Schaum, N. et al. Ageing hallmarks exhibit organ-specifc temporal β-cell replication. Diabetes 62, 2843–2848 (2013). signatures. Nature 583, 596–602 (2020). 68. Katsimpardi, L. et al. Vascular and neurogenic rejuvenation of the aging 33. Isildak, U., Somel, M., Tornton, J. M. & Donertas, H. M. Temporal changes mouse brain by young systemic factors. Science 344, 630–634 (2014). in the gene expression heterogeneity during brain development and aging. 69. Castellano, J. M. et al. Human umbilical cord plasma proteins revitalize Sci. Rep. 10, 4080 (2020). hippocampal function in aged mice. Nature 544, 488–492 (2017). 34. Horvath, S. DNA methylation age of human tissues and cell types. Genome 70. Chen, M. B. et al. Brain endothelial cells are exquisite sensors of age-related Biol. 14, R115 (2013). circulatory cues. Cell Rep. 30, 4418–4432 (2020). 35. Li, X. et al. Longitudinal trajectories, correlations and mortality associations 71. Niccoli, T. & Partridge, L. Ageing as a risk factor for disease. Curr. Biol. 22, of nine biological ages across 20-years follow-up. eLife 9, e51507 (2020). R741–R752 (2012). 36. Marquez, E. J. et al. Sexual-dimorphism in human immune system aging. 72. Keating, N. L., Norredam, M., Landrum, M. B., Huskamp, H. A. & Meara, E. Nat. Commun. 11, 751 (2020). Physical and mental health status of older long-term cancer survivors. J. Am. 37. Holzscheck, N. et al. Multi-omics network analysis reveals distinct stages Geriatr. Soc. 53, 2145–2152 (2005). in the human aging progression in epidermal tissue. Aging 12, 73. Kirkman, M. S. et al. Diabetes in older adults: a consensus report. J. Am. 12393–12409 (2020). Geriatr. Soc. 60, 2342–2356 (2012). 38. Edde, M., Leroux, G., Altena, E. & Chanraud, S. Functional brain 74. High, K. P. et al. HIV and aging: state of knowledge and areas of critical need connectivity changes across the human life span: from fetal development to for research. A report to the NIH Ofce of AIDS Research by the HIV and old age. J. Neurosci. Res. 99, 236–262 (2020). Aging Working Group. J. Acquir. Immune Defc. Syndr. 60, S1–S18 (2012). 39. Biteau, B., Hochmuth, C. E. & Jasper, H. JNK activity in somatic stem cells 75. Longo, V. D. et al. Interventions to slow aging in humans: are we ready? causes loss of tissue homeostasis in the aging Drosophila gut. Cell Stem Cell 3, Aging Cell 14, 497–510 (2015). 442–455 (2008). 76. Schleit, J. et al. Molecular mechanisms underlying genotype-dependent 40. Clark, R. I. et al. Distinct shifs in microbiota composition during responses to dietary restriction. Aging Cell 12, 1050–1061 (2013). Drosophila aging impair intestinal function and drive mortality. Cell Rep. 12, 77. Rikke, B. A., Liao, C. Y., McQueen, M. B., Nelson, J. F. & Johnson, T. E. 1656–1667 (2015). Genetic dissection of dietary restriction in mice supports the metabolic 41. Demontis, F. & Perrimon, N. FOXO/4E-BP signaling in Drosophila efciency model of . Exp. Gerontol. 45, 691–701 (2010). muscles regulates organism-wide proteostasis during aging. Cell 143, 78. Liao, C. Y., Rikke, B. A., Johnson, T. E., Diaz, V. & Nelson, J. F. Genetic 813–825 (2010). variation in the murine lifespan response to dietary restriction: from life 42. Owusu-Ansah, E., Song, W. & Perrimon, N. Muscle mitohormesis promotes extension to life shortening. Aging Cell 9, 92–95 (2010). longevity via systemic repression of signaling. Cell 155, 699–712 (2013). 79. Rando, T. A. & Chang, H. Y. Aging, rejuvenation, and epigenetic 43. Ori, A. et al. Integrated transcriptome and proteome analyses reveal reprogramming: resetting the aging clock. Cell 148, 46–57 (2012). organ-specifc proteome deterioration in old rats. Cell Syst. 1, 224–237 (2015). 80. Villeda, S. A. et al. Young blood reverses age-related impairments in cognitive 44. Ahadi, S. et al. Personal aging markers and ageotypes revealed by deep function and synaptic plasticity in mice. Nat. Med. 20, 659–663 (2014). longitudinal profling. Nat. Med. 26, 83–90 (2020). 81. Wyss-Coray, T. Ageing, neurodegeneration and brain rejuvenation. Nature 45. Bahar, R. et al. Increased cell-to-cell variation in gene expression in ageing 539, 180–186 (2016). mouse heart. Nature 441, 1011–1014 (2006). 82. Ocampo, A. et al. In vivo amelioration of age-associated hallmarks by partial 46. Arrojo e Drigo, R. et al. Age mosaicism across multiple scales in adult tissues. reprogramming. Cell 167, 1719–1733 (2016). Cell Metab. 30, 343–351 (2019). 83. Sarkar, T. J. et al. Transient non-integrative expression of nuclear 47. Martinez-Jimenez, C. P. et al. Aging increases cell-to-cell transcriptional reprogramming factors promotes multifaceted amelioration of aging in variability upon immune stimulation. Science 355, 1433–1436 (2017). human cells. Nat. Commun. 11, 1545 (2020). 48. Enge, M. et al. Single-cell analysis of human pancreas reveals transcriptional 84. Dobzhansky, T. Nothing in biology makes sense except in the light of signatures of aging and somatic mutation patterns. Cell 171, 321–330 (2017). evolution. Am. Biol. Teacher 35, 125–129 (1973). 49. Tabula Muris Consortium. A single-cell transcriptomic atlas characterizes 85. Zhang, R., Chen, H. Z. & Liu, D. P. Te four layers of aging. Cell Syst. 1, ageing tissues in the mouse. Nature 583, 590–595 (2020). 180–186 (2015). 50. Ma, S. et al. Caloric restriction reprograms the single-cell transcriptional landscape of Rattus norvegicus aging. Cell 180, 984–1001 (2020). Acknowledgements 51. Kimmel, J. C. et al. Murine single-cell RNA-seq reveals cell-identity- and This work was supported by the Glenn Foundation for Medical Research, Nan Fung Life tissue-specifc trajectories of aging. Genome Res. 29, 2088–2103 (2019). Sciences and the NOMIS Foundation. 52. Goronzy, J. J. & Weyand, C. M. T cell development and receptor diversity during aging. Curr. Opin. Immunol. 17, 468–475 (2005). 53. Tower, J. Programmed cell death in aging. Ageing Res. Rev. 23, 90–100 (2015). Competing interests 54. Gorgoulis, V. et al. Cellular senescence: defning a path forward. Cell 179, The authors declare no competing interests. 813–827 (2019). 55. Hernandez-Segura, A. et al. Unmasking transcriptional heterogeneity in senescent cells. Curr. Biol. 27, 2652–2660 (2017). Additional information 56. Robbins P. D. et al. Senolytic drugs: reducing senescent cell viability to extend Correspondence should be addressed to T.A.R. or T.W.-C. health span. Annu. Rev. Pharmacol. Toxicol. https://doi.org/10.1146/ Peer review information Nature Aging thanks Steven Austad, Vadim Gladyshev and annurev-pharmtox-050120-105018 (2020). Jing-Dong Jackie Han for their contribution to the peer review of this work. 57. Tyshkovskiy, A. et al. Identifcation and application of gene expression Reprints and permissions information is available at www.nature.com/reprints. signatures associated with lifespan extension. Cell Metab. 30, 573–593 (2019). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 58. Acosta, J. C. et al. A complex secretory program orchestrated by the published maps and institutional affiliations. infammasome controls paracrine senescence. Nat. Cell Biol. 15, 978–990 (2013). © Springer Nature America, Inc. 2021

Nature Aging | VOL 1 | January 2021 | 29–35 | www.nature.com/nataging 35