A brief history of the hyperfunction theory of aging and future directions

Weismann, McCay, and early aging programs

Scientific discussion of the causes of aging dates back at least to the time of Aristotle in ancient Greece, and the first proposal that aging may be linked to development appeared in the 19^th century [4]. August Weismann initially put forward the idea that aging may have evolved for a purpose (i.e., is programmed), though he later abandoned this view. In the early 20^th century, however, the programmatic hypothesis that growth, development, and aging were interconnected was under scrutiny. Interestingly, Clive McCay’s breakthrough caloric-restriction experiments were designed to test the hypothesis that growth restriction could influence aging [5].

Over the following decades, however, the idea that aging results from a program or from continued developmental processes became less popular. The evolutionary theory of aging argued against a programmed (i.e., adaptive) aging process, predicting that such a program would be selected against [6]. The dominant view was that aging arises from the declining force of natural selection with age, the so-called “selection shadow” [7, 8]. In this model, both genetic variants with detrimental late-life effects, or variants beneficial early in life but harmful later, can become fixed in populations, contributing to aging; here, aging is not an adaptation and, in that sense, not programmed.

Although classical evolutionary theory explains why we age, it offers little mechanistic detail regarding how we age. In the 1970s, Kirkwood proposed the disposable soma theory, suggesting a trade-off between reproduction and somatic maintenance as a driver of species differences in aging [9]. An underlying assumption of the disposable soma theory was that aging is largely the result of damage accumulation [7]. It is noteworthy that Blagosklonny was at times critical of the disposable soma theory [10], and offered his hyperfunction theory as an alternative to it, given that it too combines evolutionary and mechanistic explanations [11].

The increasing emphasis on molecular damage in aging research was likely driven, at least in part, by advances in molecular biology and biochemistry. The explosion of molecular discoveries in the second half of the 20^th century revealed a vast and intricate number of cellular components and biological processes, which in turn led to a proliferation of theories linking aging to defects in each of these many processes. Because virtually any important biochemical or molecular process can malfunction and become harmful to cells, it is easy to conceive new damage-based theories of aging. Indeed, I speculate that almost any biological process can be shown to change with age if studied well enough.

This abundance of molecular detail reinforced the perception that aging is driven by stochastic damage and led to many theories and frameworks positing damage accumulation as the root cause of aging [3, 12]. It was against this trend that a new wave of programmatic theories emerged at the start of the 21st century.

From free radicals to the genetic regulation of aging

Like many scientists entering the field at the end of the 20^th century, I encountered the free radical theory of aging as the prevailing theory when I started studying biogerontology. Its central idea that oxidative damage gradually accumulates in multiple cell components and drives the process of aging is beautifully attractive in its simplicity and elegance [13]. It also explained species differences in aging by linking metabolic rate to mitochondrial reactive oxygen species (ROS) production [14].

During my PhD (2004), and assuming these processes were drivers of aging, I studied oxidative stress, DNA damage, telomeres, telomerase, and cellular senescence in human cells [15]. However, the groundbreaking work by Van Remmen et al. in 2003, showing that mice heterozygous for mitochondrial superoxide dismutase (SOD2) – despite higher levels of oxidative damage – do not have a shorter lifespan, was a major hit on the free radical theory [16]. My work and that of others showing that metabolic rate does not correlate with mammalian longevity further disproved the rate-of-living theory and, by extension, questioned free-radical-based explanations for species differences in aging [17]. If aging were simply passive damage accumulation from metabolic by-products, then this is not supported by these comparative analyses, at least for current measures of metabolic rate. Other studies, such as the observation that rats kept at lower temperatures eat 44% more food than controls and yet do not have shorter lifespans [18], provided further evidence against simplistic metabolic explanations for aging. Taken together, these results challenged the free radical theory of aging and made me question the idea that inevitable molecular damage is the driver of mammalian aging.

One observation that shaped my interest in genetic and programmatic mechanisms came from my PhD work with telomerase-immortalized cells. I found it striking that human cells can proliferate indefinitely in vitro as long as they express telomerase [19], despite the various forms of damage that they supposedly sustain. If unavoidable, cumulative intrinsic damage determines cellular aging, indefinite proliferation should not be possible. Moreover, unlike mouse cells, human cells do not spontaneously transform [20]. Therefore, at the fundamental level, human cells can overcome intrinsic damage and can remain functional indefinitely when key genetic programs allow it. Interestingly, Blagosklonny’s observation that growth stimulation can induce cellular senescence even in the absence of molecular damage was also an important inspiration for his programmatic theories [21–23].

Developmental biology also provides clues. Most physiological functions improve from zygote to adult, and cell transplantation experiments in mice suggest that some cell populations possess intrinsic longevity beyond that of the organism [24]. If aging is due to inevitable damage accumulation, why does such damage not accumulate during early embryonic stages? Perhaps repair mechanisms are progressively downregulated throughout development, but in that case age-related damage is no longer simply a product of passive deterioration but rather a consequence of genetic programs that regulate repair and maintenance mechanisms during development.

Similarly, strong correlations across mammals between developmental time, growth rate, and lifespan shaped my views – detailed below – on the developmental theory of aging [17]. In other words, how long it takes animals of a given species to develop is a very strong predictor of how long they will live afterwards – and presumably how fast they will degenerate. Furthermore, a mouse ages much faster than a human, regardless of environment; clearly, the genome encodes species-specific rates of aging, even between closely related species like humans and chimpanzees.

The fact that aging can be genetically manipulated in animal models further demonstrates that it is, at least in part, under genetic regulation. From my early days in the field, I found this fascinating, to the point of publishing – against the advice of more senior colleagues – a short opinion piece [25]. The significant impact of single gene manipulations on aging and longevity further supports the view that aging is not merely a process of passive deterioration. Instead, it points to coordinated biological processes in – at least to some degree – determining and regulating our longevity and aging rates. In mice, the growth hormone (GH) and insulin-like growth factor 1 (IGF-1) pathway, in particular, is a major regulator of aging [26]. Intriguingly, the wild-type alleles that promote normal levels of these signals appear to reduce lifespan, suggesting a quasi-program (non-adaptive program) that promotes aging [27].

By the early 21^st century, the field was ripe for a revival of programmatic theories. Although some forms of damage – such as DNA damage, which is clearly important in cancer and similar mutation-driven diseases – contribute to aging, the idea that aging primarily emerges from genetic programs, namely the same developmental programs that build an organism, became a powerful one.

Hyperfunction, quasi-programs, and software design flaws

Blagosklonny and I developed our ideas independently. As I completed my PhD, and unconvinced that oxidative damage, telomeres, or cell senescence were drivers of aging, I became increasingly interested in genetic programs. Early proposals focusing solely on neuroendocrine mechanisms did not convince me [28], and did not convince Blagosklonny either, though his father was a proponent of neuroendocrine aging processes [29]. Developmental regulation clearly extends beyond hormones, implying deeper programmatic mechanisms.

George Williams’s antagonistic pleiotropy theory was a major influence on both my work and Blagosklonny’s. Williams proposed that processes beneficial early in life but harmful later would be favored by evolution, an observation that set the stage for programmatic theories [30]. After reviewing the literature on antagonistic pleiotropy, developmental theories of aging, and programmatic mechanisms [28, 31, 32], I formulated my first model proposing aging as a continuation of a subset of developmental programs. Published in 2005, it focused largely on physiological processes that may represent cases of antagonistic pleiotropy due to a continuation or cessation of developmental processes whose “run-on” becomes detrimental later in life [33]. I worked in collaboration with my postdoctoral mentor, George Church, whose zoological background complemented my comparative perspective. We also conceived the relation between development and aging in the context of oxidative signaling, which plays key roles during development [34].

In 2006, Blagosklonny published his seminal conceptual paper proposing the quasi-programmed theory of aging, introducing the term hyperfunction [35]. Blagosklonny proposed “a quasi-program for aging, a continuation of the developmental program that is not turned off, is constantly on, becoming hyper-functional and damaging, causing diseases of aging.” While acknowledging that damage occurs with age, including damage caused by ROS, he argued that such damage plays a negligible role in determining lifespan. Instead, quasi-programs are the principal drivers of aging and limiting human lifespan [35].

A central component of Blagosklonny’s model is the target of rapamycin (TOR) pathway as a major driver of growth and, later, aging. From his first publication in 2006, he argued that TOR activation promotes hyperfunction, aging, and age-related diseases – a concept he expanded upon in later work [36]. He also proposed rapamycin as a promising therapy against aging and age-related diseases.

Blagosklonny was a prolific author, publishing many influential conceptual papers, including several on the hyperfunction theory, as reviewed by others in this special issue [23]. The fact that rapamycin is the most effective lifespan-extending drug discovered so far supports programmatic theories and Blagosklonny’s ideas [37, 38]. I believe, however, that there must be more to growth and developmental regulation than mTOR alone. Species differences – for example, why mice grow, develop, and age so much faster than humans – must involve additional pathways. Therefore, while mTOR may underpin programmatic processes at a high regulatory level, from my comparative biology perspective, I cannot see how mTOR alone could account for species differences in aging.

I first became aware of Blagosklonny’s work through his 2006 review. He first emailed me in June 2006 to compliment senescence.info, my website on aging. After his 2006 quasi-program paper was published, I requested a PDF from him later that year. We corresponded only occasionally. In 2008, he sent me a manuscript on TOR-driven program of aging, and I replied that although I agreed with much of it, I felt he placed too much emphasis on the mTOR pathway. His medical background meant that his hyperfunction theory had a strong human-pathophysiological focus, whereas my approach has emphasized cellular, molecular, genetic, and comparative perspectives. I believe we first met in person at the 3rd International Conference on Genetics of Aging in Sochi, Russia, in April 2014, and had a cordial discussion about our respective theories.

My recent work on programmatic theories, including the software design flaw hypothesis, emphasizes the genome and epigenome as master regulators of an informational architecture producing late-life dysfunction [39, 40]. Meanwhile, Blagosklonny’s framework focused more on pathology as a manifestation of hyperfunction. In retrospect, it is unfortunate that we never collaborated, as our perspectives are highly complementary.

More recently, I have also proposed that processes that reduce cancer risk early in life may represent adaptive processes that later contribute to aging. Declines in cell proliferation, growth, and plasticity from early development until adulthood may protect against cancer, but later impair regeneration, cell proliferation, and tissue homeostasis [41]. Like me, Blagosklonny also argued that, while molecular damage drives cancer, it does not drive aging [42].

The future of the hyperfunction theory

Although programmatic theories, such as hyperfunction and software design flaw, provide powerful conceptual frameworks for understanding the aging process, much work remains to be done. They are still outside the dominant geroscience paradigms, such as the “hallmarks” and “pillars” of aging, which has drawn criticism from me and others [43, 44]. The unfortunate consequence – I would argue – is that most aging studies focus exclusively on adult life, hindering efforts to connect aging to developmental processes. If aging processes follow trajectories set early in life, then studying the whole life course is imperative to elucidate aging mechanisms. Besides, if repair and maintenance mechanisms are downregulated during development, then studying early development may prove valuable for identifying rejuvenation therapies, as already demonstrated in partial reprogramming.

In addition, most attempts to develop interventions and translate findings from the biology of aging to the clinic have been grounded on damage-based paradigms, such as antioxidants and senolytics [45, 46], which have so far shown limited ability to broadly retard organismal aging. While such interventions might still prove effective for specific conditions, I would argue that developing effective interventions for aging will ultimately require a greater mechanistic understanding of its underlying causes, including programmatic theories.

At present, programmatic theories remain largely underappreciated and overlooked. Blagosklonny did not seem deterred by this; he was confident he was on the right track to develop a conceptual model that explains aging and most age-related diseases. He also noted the importance of conceptual and theoretical understanding in biology [47], something often overlooked in an era dominated by high-throughput approaches.

Programmatic theories still face many challenges. They remain somewhat abstract, despite attempts (including mine and Blagosklonny’s) to outline specific mechanisms, such as TOR and epigenetic regulation. Their direct relevance to human pathology is also far from established. A few aging changes, such as presbyopia and thymic involution, could reflect hyperfunction or programmatic design flaws [39, 48]; David Gems has also suggested that osteoarthritis may fit this category [49]. However, such cases are exceptions rather than the rule.

The majority of age-related degenerative changes lack clear mechanistic explanations. Except for cancer, which I view as the opposite side of aging and is driven largely by DNA damage and somatic mutations [41], I suspect most degenerative changes arise from programmatic mechanisms: the continuation or mis-regulation of developmental pathways that early in life change an embryo from a highly plastic and resilient organism to an adult with more limited cell plasticity and resilience, whose run-on later results in loss of plasticity, resilience, and the major features of what we call aging. How much of the aging phenotype is programmatic? I speculate that, apart from cancer and a few other mutation-driven pathologies, most degenerative changes are due to gene regulatory programs set during development that persist into later life and become maladaptive. In other words, the loss of function and repair observed across most organs with age is the product of growth and signaling networks operating beyond their adaptive window. These quasi-programmatic processes may, in turn, result in downstream damage to cells and tissues.

Testing programmatic theories remains challenging. If aging is encoded in human biology, it will be difficult to modify with current technology. The fact that GH and IGF-1 inhibition can slow aging in mice is remarkable and supports the hyperfunction theory [27]. Likewise, the life-extending effects of caloric restriction and rapamycin further support this notion [25, 33, 39]. These interventions have systemic effects, which explains their impact on broader aspects of the aging phenotype, yet presumably more specific processes exist and remain unknown. As such, I speculate that reactivating developmental mechanisms in aged tissues could restore plasticity and function, albeit at the possible cost of increased cancer risk [41]. Several growth and signaling pathways, such as hedgehog, hippo, and wnt signaling, decline from early development and merit further study. Their complex regulation complicates therapeutic manipulation, however.

It is important to mention that programmatic theories are falsifiable. For example, if mice genetically engineered to greatly reduce molecular errors (e.g., somatic mutations) were shown to live substantially (e.g., 2-3x) longer, and aging is slowed across organs, this would strongly challenge the hyperfunction theory. Likewise, robust life-extension and multi-organ aging retardation from interventions that reduce molecular damage without altering developmental signaling would also challenge the theory. A central challenge in geroscience, however, is that life-extending interventions remain modest, making it difficult to test mechanistic theories.

Partial reprogramming using Yamanaka factors to rejuvenate cells and reset the epigenome demonstrates that biological time can be reversed, at least in cells [50]. Indeed, many companies are currently pursuing rejuvenation technologies [51]. If partial reprogramming or other genetic interventions can reverse biological age in tissues or organisms, it would strongly support a programmatic model of aging. While I believe this is possible, Yamanaka factors are perhaps too blunt for organ rejuvenation and may increase cancer risk; tissue-specific factors that revert cells to young rather than embryonic states will, I assume, be needed. Elucidating tissue-specific rejuvenation factors and demonstrating that these are not only effective in retarding aging in vivo but also require regulatory resetting rather than damage repair alone would provide strong evidence for the hyperfunction theory.