Introduction

Age‌‌ is the biggest risk factor for most diseases, including high-impact conditions such as cancer, cardiovascular disease, and neurodegenerative disease [1]. The proportion of aged individuals is increasing globally, and this is resulting in economic, healthcare, and societal strains. Consequently, there is a heightened need to understand the aging process to ensure better health in older age [2]. Aging has been examined from a number of perspectives, ranging from evolutionary biology to human epidemiology and, more recently, basic biology. It is the study of the latter, the biology of aging, that currently holds the potential to generate novel treatments. In this Essay, we argue that research into the basic biology of aging needs to adapt concepts and approaches from human life-course epidemiology in order to adequately address the needs of numerous, diverse, aging populations.

Nutrient signaling and aging

The past four decades have seen a profound shift in how we view aging, from seeing age as a non-modifiable risk factor for many diseases, to understanding aging as a plastic process that can be targeted for improved health in older age. The study of the mechanisms behind aging and its plasticity has, since the beginning, been tightly linked to nutrient signaling pathways (Fig 1). A classic example is the significant extension of Caenorhabditis elegans life span as a consequence of reduced activity of age-1 or daf-2, two genes that encode core components of the nutrient-responsive insulin/insulin-like growth factor (IGF) signaling (IIS) pathway [3–5]. IIS is a neuroendocrine signaling network defined by the structure of the extracellular ligands (which resemble insulin) and their cognate cell-surface receptors (which are related to the insulin receptor). Experimental evidence has accumulated over decades demonstrating that reduced IIS activity can decelerate aging. This effect is strikingly well conserved across large evolutionary distances, and is observed in worms, flies, and mice. Indeed, genetic epidemiology studies indicate that at least some components of this network affect human aging. For example, FOXO3A is one of the few genes that have been robustly associated with human longevity [6–9]. FOXO3A encodes one of the members of the Forkhead box O (FoxO) transcription factor family in humans, which seems to be at the core of the IIS pathway’s effects on longevity [4,10,11]. This suggests that IIS manipulation is a promising way to slow human aging.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. The impact of nutrient signaling on aging.

A simplified summary of some of the known nutrient signaling pathways, including insulin/IGF, mTORC1, ERK, and AMPK, and how they broadly impact healthy aging. This diagram is based on previously published schematics, omitting pathway cross-regulation for clarity [12–14]. Dotted arrows denote parts of pathways with missing components for simplification. AMPK, AMP-activated protein kinase; ERK, extracellular signal-regulated kinase; FOXO, Forkhead box O; IGF, insulin-like growth factor; mTORC1, mechanistic target of rapamycin complex 1; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3003794.g001

Aging is marked by the degeneration of several processes, including proteostasis, mitochondrial quality control and activity, metabolic and immune function, and other interconnected aspects of cellular and physiological function, altogether comprising the “hallmarks of aging” [15]. Importantly, several of these hallmarks are targeted in concert by direct modulation of IIS, and are collectively responsible for the substantial changes in the rate of aging under IIS-reduced conditions (Fig 1) [16–20]. The responses are tissue-specific, often coordinated by a number of downstream transcription factors and other effectors that act in a tissue-restricted manner [18,21–24]. Conversely, changing the intracellular IIS activity in specific tissues and cell types is sufficient to extend life span of the whole organism [25–28].

The IIS pathway does not act alone, but is integrated with several others, at both the endocrine and intracellular level, all of which coordinately control animal physiology (Fig 1). For example, IGF1 production can be regulated though the somatotropic axis in mammals to mediate part of the effects of growth hormone (GH) [29]. Inside the cell, receptor activation may result in the activation of both the Akt and extracellular signal-regulated kinase (ERK) pathways, which in turn interact with other intracellular nutrient- and energy-signaling pathways such as those centered around mechanistic target of rapamycin complexes 1 and 2 (mTORC1 and mTORC2) and AMP-activated protein kinase (AMPK) [30]. There is evidence that these pathways also drive aging; for example, GH may have key importance in mammals [31], and ERK has been documented to be relevant in flies, mice, and potentially in humans [23,32–34]. Several of these components can be targeted by drugs to induce a longer and healthier life span: mTORC1 can be targeted by rapamycin‌‌ [35], ERK by trametinib [32], and AMPK by metformin [36]. Interestingly, endocrine signals coordinately regulate the activity of these intracellular pathways, and the interventions that target these upstream signals often provide the largest and most robust effects; examples include targeting Drosophila insulin-like peptides in flies, or GH in mammals [31,37,38]. Viewed from this perspective, it is unsurprising that combination drug treatments aimed at kinases that would be coordinately regulated downstream of the endocrine signal(s) can have substantially additive effects on life span [34].

Once the components of the network started being explored as drug targets, and the questions of when treatments could be effective were raised, some curious observations emerged. For example, transient rapamycin treatment in mice and fruit flies has a long-term beneficial effect; notably, these long-term effects of transient rapamycin treatment could not be explained by prolonged mTORC1 inhibition [27,39]. Similarly, short-term treatment with an mTORC1 inhibitor in humans, aside from improving vaccine efficacy in older individuals, may also have a long-term beneficial effect on susceptibility to infection [40,41]. The benefits of transient treatment are seen not only for mTORC1 inhibitors: activation of the transcription factor FoxO restricted to early adulthood was sufficient to extend life span in flies [26]. Additionally, reducing early adulthood protein synthesis in Drosophila extends life span by lowering the levels of juvenile hormone (JH) at this stage; this could be via reducing IIS, as JH potentially mediates life span extension by IIS inhibition [42–45]. These surprising findings add to a growing body of evidence suggesting that interventions in early life, particularly those impacting nutrient signaling, can have an effect on subsequent aging. The way they do so is poorly understood; however, we can define this lasting state of the body as having “physiological memory.” This term is intentionally broad, highlighting that the animal’s physiology seems to be altered in the long term, while being agnostic as to the underlying mechanisms.

The fundamental role of the nutrient signaling network is to mediate responses of cells to nutritional environments. It has long been known that dietary regimens, such as dietary restriction and intermittent fasting, can robustly extend life span in multiple model organisms and, while the relationship is complicated and multi-factorial, nutrient signaling is believed to be at the core of how feeding regimens can slow aging [46–48]. Recent findings have shown that such dietary regimens can have long-term impacts that persist after administration; late-life mortality analysis indicated that mice are affected by past diets despite a dietary change, with ad libitum diet leaving a stronger memory than dietary restriction [49]. Work in C. elegans has also shown that removal of the food source Escherichia coli can extend life span, even if limited to the first 10–15 days of adulthood [50]. Intermittent fasting can also extend life span when restricted to early adulthood, as seen both in C. elegans and in Drosophila [51,52]. While the exact mechanisms are yet to be identified, it was shown that in Drosophila they are independent of the S6K-specific branch of mTORC1 signaling, informing the direction for future work [51]. Further work on fruit flies has shown that other dietary changes restricted to development or early adulthood can still have long-term effects on life span [53–55]. For example, short-term exposure to a high sugar diet in early adulthood shortens subsequent life span in a manner dependent on foxo, the Drosophila orthologue of FOXO3A [54]. Similarly, the life span-extending effects of short-term methionine restriction are also dependent on FoxO-induced methionine sulfoxide reductase A expression [55].

Overall, work with model organisms clearly shows the potential of short-term alterations of nutrient signaling (pharmacological, genetic, or dietary) to slow aging. These long-term effects may be dependent on a physiological memory engrained in the system. But is this physiological memory applicable to human aging? To explore this, we will turn to some key observations in human epidemiology.

Evidence for physiological memory in human disease epidemiology

Substantial epidemiological evidence supports the notion that the physiology and health of an aged adult are influenced not only by their current environment but also by exposure to specific insults or conditions throughout their life. Most surprising is the influence of the in utero and early-life environment on metabolic health in late adulthood, which has been documented for decades and has culminated in the developmental origins of adult disease (DOD) hypothesis. This field began with striking work by Barker and colleagues [56,57], who documented a strong negative correlation between the weight of an infant at birth and the likelihood of them developing type 2 diabetes (T2D), hypertension, and death from cardiovascular disease by 75 years of age. This suggests that the nutritional environment experienced during fetal development could influence the health of the individual decades after birth. The key idea that past environmental exposures can contribute to health outcomes in late adulthood took shape in a life-course approach to epidemiology [58], which spurred on research into the underlying biological mechanisms, as well as societal implications, with a focus on metabolic health.

The correlations between in utero nutrition and late-life metabolic health have been demonstrated in several other longitudinal studies, many of which make use of exceptional historical events that exposed pregnant mothers to abnormal conditions for a specific amount of time. For example, in the winter and spring of 1944, the Netherlands experienced a famine, and among the children of mothers who were pregnant during this period there was a strong correlation between the restriction of calories in utero and either increased or decreased likelihood of late-life obesity and cardiovascular disease, depending on whether the famine occurred at early or late gestation, respectively [59]. These correlations were reported in multiple other studies following individuals who experienced famines across the globe [60–64]. Longitudinal studies following the effects of extreme nutritional events are not limited to famines; between 1940 and 1953, sugar was rationed in the United Kingdom, limiting consumption to the current daily recommended amounts. Exposure to sugar rationing in utero significantly reduced the likelihood of developing hypertension or T2D, and this effect was intensified if rationing continued up to 12 months after birth [65,66]. These remarkable correlations were, importantly, replicable in controlled animal studies, strengthening the conclusion that late-life metabolic health can be affected by developmental nutritional exposures [67,68].

The DOD theory establishes that a physiological memory can be formed during development and impact health in the long term. However, information on adults that have experienced a short-term but profound metabolic change, or have undergone interventions that alter metabolic health, is not as extensive. Nevertheless, a few studies have tracked the long-term effects of diabetes or obesity, even when these conditions are controlled or reversed, identifying a way in which short-term metabolic dysfunction can be cemented in what has been termed “metabolic memory” (which we suggest to be a specific instance of a broader phenomenon of physiological memory). Such a metabolic memory formed in adulthood is seen in individuals with obesity who lose weight: these individuals often experience weight re-gain (often termed a “yo-yo” effect) [69,70], which is likely driven by long-term and irreversible physiological changes caused by obesity. Using single-cell sequencing approaches, Hinte and colleagues recently found that, despite weight loss, adipose tissue from previously obese humans and mice seemingly retains a transcriptional memory of the metabolic state of obesity, as well as persistent pro-inflammatory signatures and impaired adipocyte functionality and metabolism, which potentially primes individuals who previously had obesity to weight re-gain [71]. This obesity-induced and post-weight-loss-retained pro-inflammatory phenotype, as well as insulin insensitivity, has also been observed in other studies, and could not be resolved by preventing hyperphagia, indicating that the yo-yo effect may indeed be caused by a physiological memory of obesity [72]. Additional evidence of such memory, including metabolic memory of T2D despite glucose level correction, has been thoroughly reviewed by others [73]. Such long-term maintenance of pro-inflammatory phenotypes or IIS disruption would be expected to influence aging significantly [74].

While there is a tight link between metabolic disease and aging, tracking the formation and persistence of metabolic or broader physiological memory in individuals that have a healthy early adulthood is key to extend this concept to the broad treatment or prevention of aging. One such work will be the legacy study of the CALERIE phase 2 trial [75], where healthy, middle-aged individuals without obesity experienced moderate caloric restriction for 2 years, with positive outcomes, such as improvements in cardiometabolic health markers [76]. Ten to fifteen years following the end of treatment, participants will be reassessed to monitor the long-term effects of this short-term intervention [77]. Results from this legacy study may reveal whether physiological memory can be formed in healthy individuals, and whether this can meaningfully impact health and aging in the long term.

Altogether, there is abundant evidence suggesting that humans can retain a physiological memory, including that of past diets and of the metabolic states they lead to. This tallies well with the recent observation of long-term effects on aging from manipulations of nutrition or of nutrient signaling pathways in model organisms, and strongly suggests that early-life environments can have a profound effect on the biology of human aging, regardless of how transiently they are experienced. But how is this memory formed and retained?

Possible mechanisms of physiological memory

Nutrient signaling pathways, and other similar regulators of animal physiology, are thought to act in a homeostatic manner, providing continuous adjustments to maintain function in a changing internal and external environment. Their responses and actions are essentially thought of as short-term, so it is important to understand how they are able to cause a long-term change in the rate of aging.

One major contender is the epigenome, including but not limited to DNA methylation and histone modifications. Indeed, it is becoming increasingly clear that nutrient signaling and metabolism have a complex and significant impact on epigenetic remodeling [78,79]. Epigenetic modifications can affect the expression of large sets of genes, enabling metabolic pathways to be more robustly manipulated than if individual genes are targeted alone. Additionally, epigenetic modifications have staying power, which can be carried across generations of cell division [80] and sexual reproduction. Indeed, transgenerational inheritance has been demonstrated in multiple model systems, where parental epigenetic changes are transferrable and impact progeny aging over multiple generations [81–84]. Importantly, there is ample evidence that nutrient signaling can alter the epigenome. For example, fasting leads to chromatin remodeling via mTORC1 and RNA polymerase I inhibition [85], and AMPK directly phosphorylates histone H2B, which facilitates the expression of AMPK-responsive genes [86]. While there is still much to be explored, nutrient signaling can both directly and indirectly alter the epigenome [87–94], thus giving these pathways the potential to have lasting effects beyond their activation or inhibition.

Epigenetic changes formed during development could explain the DOD phenomenon [95,96]. More recent work, in the context of metabolic disease, supports the idea that this form of physiological memory can also be generated in adulthood [97]. For example, the difference in H3K4me3, H3K27me3, H3K27ac, and H3K4me1 signatures seen between adipocytes from animals with obesity compared with lean animals is partially maintained even after weight loss, and may explain the predisposition of mice that have previously had obesity to regaining weight [71]. Furthermore, there is substantial evidence that epigenetic regulation underlies metabolic memory in other contexts as well (reviewed previously [73]). Importantly, mounting evidence shows that the epigenome can strongly impact longevity [78,93,98]. Hence, it is possible that the epigenetic mechanisms demonstrated to underlie metabolic memory are the same as those underlying other forms of physiological memory that can affect aging. Indeed, epigenetic changes that affect gene accessibility seem to hold memory of life span-extending interventions [26,99]. The memory effect of FoxO activation in Drosophila early adulthood, which can extend life span significantly, seems to be dependent on the presence of functional switch/sucrose non-fermentable (SWI/SNF) and imitation SWI (ISWI) complexes, which are responsible for opening chromatin and enabling transcription of target genes [26].

Additional hypotheses implicating epigenetic modifications as the source of physiological memory can be proposed to explain other recent findings. For example, it is possible that the life span-extending effects of methionine restriction in early adulthood, which upregulates levels of the methyl-group donor S-adenosyl methionine (SAM), can be mediated by long-term changes in histone methylation [55,100,101].

Epigenetically encoded physiological memory need not be the only mechanism by which early-life events can influence longevity. On a cellular level, post-translational modifications and behaviors of proteins act as a short-term as well as long-term and trans-generational memory store. Mnemons, or protein condensates, occur in multicellular organisms, and have been thoroughly studied in Saccharomyces cerevisiae, enabling cells to remember mating cues, stressors, or nutritional environments. Many such protein condensates are RNA-binding proteins, suggesting that the formation of condensates can have broad and long-lasting effects on the translation of specific proteins [102]. Moreover, aging may be partially caused by the accumulation of irreversible damage, which can become a physiological memory. For example, damage by advanced glycation end-products and reactive oxygen species (ROS) can begin in early adulthood and lead to long-term changes that impact longevity [103,104]. Exposure to inflammation, particularly chronically, can also cause cumulative damage that can accelerate aging [105]. Additionally, the gut microbiome, which is tightly linked to the progression of aging, can have its composition altered in the long term by dietary conditions experienced in the past, suggesting that it may hold part of the whole body’s physiological memory [106–108].

Is the physiological memory that impacts whole-body aging held in a specific tissue? Identifying its location could be key to effective manipulation. The adipose tissue is consistently identified as a source of physiological memory in multiple organisms, including humans [71]. FoxO activation in the Drosophila fat body alone was sufficient to extend life span, even when restricted to early adulthood, and while this organ can also function as a hepatic tissue, the liver does not seem to be responsible for physiological memory in mice [26,49,109]. Additionally, memory may be retained in the adipose tissue through changes in its composition. High sugar or fat diets, diabetes, and obesity, even in the short term, can affect the immune cell content and the inflammation of this tissue, forming a potential physiological memory that can impact whole-body aging [71,110–112]. The gut may also be responsible for the maintenance of physiological memory, as it is in Drosophila. Here, the memory of rapamycin treatment seems to be held in a manner dependent on the upregulation of lysosomal α-mannosidase V expression and persistent induction of autophagy [27]. While individual tissues may be important, it is entirely possible that robust longevity and healthspan are dependent on the holistic maintenance of good health in multiple organs, with a memory potentially encoded in each tissue having an important role.

Additionally, vital to understanding physiological memory is knowing when it is possible to cement it in the body. For example, while the correlation mentioned earlier between exposure to sugar rationing and metabolic health is clear, if children experienced rationing for longer than 12 months after birth, there was no additional protection against developing T2D or hypertension, indicating that a significant reduction in formation of metabolic memory occurs after the first year [65,66]. This perhaps indicates a reduction in the system’s plasticity, despite the body still developing. Similarly, ROS have different effects on life span depending on when they are modulated: in C. elegans, even just a two-day difference in age can affect whether an antioxidant has a positive or negligible effect on life span [113]. If we consider an animal’s entire life course, when is the system most plastic and able to shape aging? And at which points is any memory formed still reversible? The answers to these questions may be context dependent: the effects of life span-extending or -shortening events may depend on when the relevant biological processes take place. An example of this comes from work in Drosophila showing that reducing protein synthesis in early adulthood alone extends life span, possibly because this is when translation is at its peak in fruit flies [45].

Much work has yet to be done to truly understand the core mechanisms behind physiological memory, and only a handful of studies have functionally linked these molecular mechanisms to longevity. This is in part because the biology of aging, as a research discipline, has yet to pay attention to the untapped potential of studying aging from a life-course perspective.

Towards a life-course understanding of the biology of aging

Understanding how physiological memory is formed and the consequences it has for health in older age should be a key part of understanding the biology of human aging, for several reasons. First, given the growing global obesity epidemic affecting multiple generations, and the current view that obesity accelerates aging, it is very likely that nutrition and dietary habits throughout life are having and will continue to have a significant impact on how we age [66,114–116]. Importantly, metabolic programing arising from past dietary habits and metabolic disease may persist regardless of whether a healthy metabolic state is eventually achieved [71]. Hence, unless strategies are specifically developed to combat this programming, the effects of metabolic disease on the aged population will linger, despite our best efforts. Second, the diversity in nutritional and other environments that individuals and populations are exposed to during their lifetime is highly likely to lead to disparities, often linked to socioeconomic status, which could affect late-life health and aging [117–120]. Are particular life experiences going to make some interventions targeting aging ineffective in certain human populations?

Additionally, a detailed mechanistic understanding of the points in the life course that can profoundly shape the biology of aging may open up avenues for potentially easy, effective, cheap, and broad prophylactic treatments. Indeed, there is mounting evidence that pharmacological interventions that target aging may not have the same efficacy across the life course [121]. Knowing when they are effective may alleviate not only the unwanted effects of pharmaceuticals but also the economic burden they impose if administered chronically. Dietary, lifestyle, and behavioral interventions that may extend life span are also unlikely to be maintained for a lifetime, and their implementation as well as efficacy may vary drastically between people in different socio-economic circumstances [122]. Thus, a biological understanding of when they are most effective, and in which individuals, will help improve impact and minimize costs. Furthermore, there is the possibility of targeting physiological memory in later life, either by reversing physiological memory that would negatively impact life span, or mimicking that which would lead to healthy aging.

We think that nutrient signaling pathways are likely to have an important role in programing animal physiology towards particular aging outcomes. This is because we are starting to see evidence that they can program life span and older-age health in animal models; because they respond to environmental cues that we know from epidemiological studies to be relevant to programing of human aging and age-related diseases; and because they provide a plausible mechanistic link between past environments and health in older age, for example, by their known ability to alter the epigenome. The aging field has identified multiple, highly interconnected nutrient signaling pathways that independently impact the rate of aging, modulating a number of cellular and physiological processes in a number of tissues and organs. Now, to the many diagrams summarizing the mechanisms of how these pathways act on aging, the temporal dimension of physiological memory should be introduced (Fig 2). In essence, we propose that to understand animal aging, we need to move towards a more dynamic understanding of the biological mechanisms whereby different environmental exposures, internal processes, and longevity-promoting mechanisms integrate across the life-course in model organisms.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. A life-course dimension to the mechanism of aging.

Temporal changes in environmental exposures, such as diet, can lead to relatively transient modulation in the activity of pathways that promote aging, such as nutrient signaling pathways. This can be sufficient to form physiological memory and impact subsequent aging, health, and longevity.

https://doi.org/10.1371/journal.pbio.3003794.g002

How is physiological memory formed and retained in an animal? What kind of nutritional and signaling states can be imprinted? Are there points along an animal’s life course when memory is most readily formed and retained? Can the understanding of physiological memory be used to tailor anti-aging therapy? We believe answering these questions will be key to advancing our understanding of biogerontology, and will require the targeted efforts of researchers interested in the biology of aging.