Design principles of gene circuits for longevity

Abstract

Aging is a dynamic process driven by cellular damage and disruption of homeostatic gene regulatory networks. Traditional studies often focus on individual genes, but understanding their interplay is key to unraveling aging mechanisms. This review explores gene circuits that influence longevity, highlighting feedback loops’ role in maintaining cellular balance. The Sir2-HAP circuit in yeast serves as a model to explore how mutual inhibition between pathways influences aging trajectories and how engineering stable fixed points or oscillations within these circuits can extend lifespan. Feedback loops critical for maintaining homeostasis are also reviewed, highlighting how their destabilization accelerates aging. By leveraging systems and synthetic biology, strategies are proposed that may stabilize these loops within single cells, enhancing their resiliency to aging-related damage.

Unraveling gene regulatory networks

While human lifespan has increased over time, this has not necessarily been accompanied by a corresponding improvement in the health of the elderly population[1,2]. Reducing the healthcare burden of aging requires novel approaches that are rooted in a deeper understanding of the biology of aging[3,4]. The aging process is driven by the accumulation of cellular and genetic damage, stemming from a complex interplay of intrinsic biological mechanisms and external environmental influences[4,5]. Prior aging studies have focused on individual genes or pathways in isolation and measuring lifespan as a static endpoint assay[6–9][10,11]. As a result, how aging-related genes interact with one another and how these gene regulatory networks (see Glossary) (GRNs) operate dynamically to drive aging remain significant unanswered challenges.

GRNs consist of nodes, symbolizing genes or regulatory elements, and edges, depicting the interactions or regulatory connections between these nodes (Figure 1). Highly connected nodes at the center of a GRN are the major orchestrators of a cell’s response to stimuli. The dynamics of these nodes can often be explained by focusing on a few key local interactions, i.e. sub-graphs. This simplification facilitates mathematical modelling and permits simulations of dynamics under different parameter regimes. Network motifs are recurrent sub-GRNs, typically including up to four nodes, with characterized behaviors[12,13] Network motifs can be as simple as positive autoregulation which ensures sustained activity of a node (Figure 1, right - top). In contrast, mutual inhibition between two nodes can lead to two distinct cell fates (Figure 1, right - middle), where the system stabilizes in one of two states based on initial conditions but can also be extended to support quadrastable states[14] .The negative feedback loop is a motif especially critical for ensuring homeostasis, activated by detecting deviations from a set point and triggering mechanisms that counteract those changes (Figure 1, right - bottom). These motifs are observed in many GRNs and are reinforced by redundant and compensatory pathways to increase resiliency to perturbations.

Figure 1. Representation of the Aging Process through Sub-Networks in Gene Regulatory Circuits.

Dynamic systems analysis can be used to model the complex interplay of gene regulatory networks (GRNs) in the aging process. GRNs are composed of nodes (genes or regulatory elements, depicted as circles) and edges (interactions or regulatory relationships between these nodes, depicted as lines). Damage accumulation and changes in gene regulation are key factors driving aging, and the GRNs that respond to such perturbations can be broken down into smaller, more manageable sub-networks (sub-GRNs). These sub-GRNs typically consist of one to four nodes (circles) and their relationships (arrows) with other nodes. Sub-GRNs that occur consistently within GRNs are termed network motifs, which give rise to important temporal behaviors, such as sustained activity, bimodality, and oscillatory behavior. These dynamical behaviors then lead to a specific response or change in the cellular state.

Decoding the emergent behavior of aging-related GRNs sets the stage for rational design of new interventional strategies to mitigate age-related diseases and promote healthy longevity. However, the intricate nature of aging-related processes cannot be fully understood through traditional reductionist methods. Instead, systems-level approaches designed to analyze nonlinear dynamics of gene circuits are required. In addition, such network-based approaches can be naturally integrated with synthetic biology to reveal the design principles of pro-longevity strategies. This review focuses on a series of recent studies in the model organism Saccharomyces cerevisiae (baker’s yeast). S. cerevisiae is a powerful model organism for studying aging as it shares evolutionarily conserved molecular mechanisms associated with longevity, is highly amenable to genetic manipulation, and, due to its relatively short generation time, can reveal aging-related processes over multiple generations in a relatively short period. Yeast display several of the canonical hallmarks of aging including dysregulation in nutrient sensing [15,16], mitochondrial dysfunction [17–19], loss of proteostasis [20–22], genomic instability[23–27], and others[28]. These studies demonstrate how a dynamic systems approach can bring new insights to aging biology, followed by a discussion of potential applications of this approach to different aging-related networks in yeast and mammals.

An example pipeline for the assembly and verification of a GRN

In decoding the regulatory relationships of a GRN, researchers can infer not only the direct interactions between nodes but also speculate on their emergent and dynamic behaviors. For instance, a negative feedback loop might suggest oscillatory dynamics between nodes A and B, where the system alternates between an A^highB^low state and the reverse (Figure 1 right – bottom). Such insights are most accurately derived from a combination of methods. This section provides an example pipeline for constructing GRNs that incorporate both static and dynamic information.

Generating novel or utilizing publicly-available bulk or single-cell (sc) RNA-sequencing data is an excellent starting point due to its high-dimensionality. For example, bulk RNA-sequencing performed before and after a perturbation can be analyzed using Weighted Gene Co-Expression Network Analysis (WGCNA) to identify gene modules based on co-expression patterns that become differentially expressed in response to a perturbation[29]. These modules can then be input into pathway analysis algorithms such as GSEA (Gene Set Enrichment Analysis)[30] to determine enriched biological pathways and processes.

If single-cell (sc) RNA-sequencing data is available, tools like SCENIC and SCENIC+ add an additional layer of information by incorporating transcription factor (TF) binding and target gene interactions to generate co-expression modules specific to subpopulations within heterogeneous samples[31,32]. For example, by analyzing the co-variability between cis-regulatory elements and gene expression, researchers identified GRNs that encode cell-type-specific gene expression in the fly brain[33]. Gene expression trajectories can be inferred using tools such as Monocle[34] and scVelo[35] , elucidating how dynamics differ during branching events that lead to the differentiation of distinct cell types. Pathway analysis or tools such as MetaFlux[36] predict pertinent cellular processes for distinct cell types.

To summarize, at this point in the pipeline, one has the following information: (1) differentially regulated genes and regulons at specific timepoints, (2) key regulators of cell trajectories and gene expression dynamics, and (3) pathways associated with these timepoints. This information can provide the starting point for generating ordinary differential equations (ODEs) describing the system. For example, transcription factors identified through SCENIC+ as key regulators can be modeled as activators or repressors in ODEs, with their target gene expression levels described by production and degradation rates inferred from Monocle or scVelo. Additional nodes representing larger metabolic pathways can be incorporated with the appropriate regulatory interactions deduced from co-expression of genes within pathways. Once a GRN is assembled, the dynamic behavior of GRNs can be simulated with COPASI[37] or PySB[38] for ODE-based simulations, or use stochastic frameworks like StochPy[39] to capture cell-to-cell variability in gene expression.

Researchers can then follow up on central regulators by single-cell imaging. Either through endogenous tagging, fluorescent reporters, dyes, or biosensors, the dynamics of these nodes can be revealed through single-cell time-lapse imaging[40] . For example, to capture the dynamics of a TF identified by SCENIC+, one can use as a proxy a promoter recognized by that TF driving expression of a fluorescent protein. For nodes that encompass multi-protein kinases that are largely regulated through post-translational modifications, biosensors are a good option[41] . Researchers can also use dyes that report on metabolic activity, such as Tetramethylrhodamine (TMRM or TMRE), which exhibits increased fluorescence with higher mitochondrial membrane potential. The data collected from imaging the dynamics of these processes in response to a perturbation can be used to constrain the parameters of the ODE, refining the model. Once the model is optimized it can reveal critical information about the stability of cellular states—for example, whether oscillations within the GRN represent a stable limit cycle or transient fluctuations between unstable points, and which specific cell states are maintained over time.

Dynamic systems analysis to gene circuits in cellular aging

The replicative aging of budding yeast S. cerevisiae has proven to be a genetically tractable model for aging of mitotic cell types in mammals and has led to identification of many conserved genes that influence longevity[42]. In contrast, the study of chronological aging in S. cerevisiae, the lifespan of a non-dividing cell, provides complementary insights especially regarding post-mitotic cells such as neurons[43]. This paper focuses on gene circuits that extend replicative aging to explore its contributions to mitotic cell aging.

Over the past seven years, the integration of high-throughput dynamic measurements with computational modeling has been employed to investigate the network-driven dynamics of yeast aging[44–50]. This work revealed that genetically identical yeast cells age through two different trajectories with distinct phenotypic changes - one with ribosomal DNA (rDNA) silencing loss and nucleolar decline (Figure 2, designated as mode 1 aging) and the other with heme depletion and mitochondrial decline (Figure 2, mode 2 aging). The divergent progression toward mode 1 vs mode 2 aging is governed by a mutual inhibition circuit of Sir2 and HAP, in which the lysine deacetylase Sir2 mediates rDNA silencing to maintain nucleolar stability [25,51] whereas the heme-activated protein (HAP) complex controls gene expression required for heme biogenesis and mitochondrial function[52].

Figure 2. Dual aging trajectories and engineered longevity interventions in S. cerevisiae.

The replicative aging of Saccharomyces cerevisiae proceeds through two distinct aging trajectories: mode 1, characterized by nucleolar decline and ribosomal DNA (rDNA) silencing loss, and mode 2, marked by mitochondrial decline and heme depletion. The balance between these two aging modes is governed by a mutual inhibition circuit between Sir2, a lysine deacetylase involved in rDNA silencing, and HAP, a complex regulating heme biosynthesis and mitochondrial function. Genetic interventions designed to overexpress Sir2 resulted in a new aging mode (mode 3). Genetically rewiring the Sir2-HAP circuit into a negative feedback loop led to longevity within the intermediate Sir2 and HAP level (oscillating between Mode 1 and Mode 2) and thus significant lifespan extension. This was also evident in cases of limited or periodic glucose availability. These findings demonstrate how genetic manipulations can modulate aging trajectories in yeast, offering a framework for studying longevity interventions that could be further investigated in other organisms.

A mathematical model composed of two ordinary differential equations revealed that the mutual inhibition between Sir2 and HAP gives rise to multiple different stable fixed points (or steady states) on the Sir2-HAP landscape[46]. The aging process of each single cell can thus be viewed as divergent progression toward these steady states – progression toward the low Sir2 high HAP state corresponds to mode 1 aging with rDNA silencing loss and nucleolar decline, whereas trajectories toward the low HAP states correspond to mode 2 aging featuring mitochondrial deterioration[46].

This dynamical model not only serves to understand the natural aging processes, but can also guide the design of interventional strategies for longevity. The main objective of such interventions is to prevent or at least delay progression toward the detrimental ending states of either mode 1 or mode 2 aging, linked to rapid damage accumulation and cell death. This task is particularly challenging because elevating either Sir2 or HAP, the two opposing longevity factors, tends to push the cell toward the alternative aging path, which ultimately leads to cellular decline and death. Therefore, effective interventions must navigate this delicate balance to avoid inadvertently accelerating the aging process by tipping the system too far in either direction. From a dynamical systems perspective, this requires the creation of a new stable fixed point or a limit cycle around the healthy, young state with intermediate Sir2 and HAP levels. Under guidance of modeling, this has been realized experimentally by genetic engineering and synthetic biology[46,49].

Genetic perturbations of Sir2 and HAP were performed to create a new stable fixed point for longevity within the Sir2-HAP space. It was found that a two-fold overexpression of Sir2 generated a new, long-lived aging mode that concluded at a state characterized by intermediate levels of Sir2 and HAP (Figure 2, mode 3 aging). This observation was reproduced by the model, in which 2-fold elevation in Sir2 level can partially counteract the inhibition from HAP, leading to the emergence of a new stable fixed point, which corresponds to experimentally-observed mode 3 aging. The model further predicted that an additional elevation in HAP under this condition can stabilize this new stable fixed point and enrich mode 3 cells in an aging population. This prediction was validated experimentally, resulting in a ~60% increase in lifespan[46].

In a separate study, to create a limit cycle for longevity, the endogenous Sir2-HAP circuit was genetically rewired into a negative feedback loop, which can theoretically produce sustained oscillations in Sir2 and HAP levels. To accomplish this experimentally, the native promoter of SIR2 was replaced with a HAP-inducible, CYC1 promoter, to introduce positive transcriptional regulation of SIR2 by HAP. To enable transcriptional inhibition of HAP by Sir2, a construct containing the HAP4 gene (encoding a major HAP component) was inserted under a constitutive promoter into the rDNA region, which is subject to transcriptional silencing mediated by Sir2. The rewired strain exhibited oscillations in Sir2 and periodic cycling of rDNA silencing and heme biogenesis during aging, without a prolonged commitment to either rDNA silencing loss (mode 1) or heme depletion (mode 2). This resulted in an 82% increase in lifespan - a record for yeast lifespan extension by interventions[49] (Figure 2). This work established, for the first time, the causal connection between gene network architecture and cellular longevity.

The stable fixed point for longevity can also be created through modulating environmental glucose levels. Previous studies showed that glucose limitation can increase the activities of both Sir2 and HAP[53–57][58] and extend lifespan in yeast and other model organisms[59], raising the possibility that such environmental alterations could give rise to a longevity stable fixed point similar to the one created by gene overexpression. To test this, the effect of glucose concentrations from 5% to 0.02% were systematically evaluated on yeast aging and found that the changes in glucose level modulate Sir2 and HAP activities with different dose response relationships and thereby influence the balance of the two factors and the fate decision of aging - decreasing glucose level biases the fate commitment toward Mode 1 aging. Intriguingly, 0.1% glucose leads to a subtle equilibrium between Sir2 and HAP at intermediate levels, resulting in the emergence of a longevity stable fixed point and therefore an optimal lifespan extension effect compared to other glucose levels tested. Moreover, it was shown both by theory and experiments that periodic oscillations of external glucose levels can also enable a dynamic stabilization of the system around intermediate Sir2 and HAP levels, leading to an extended lifespan without creating the longevity stable fixed point[50]. Efforts are ongoing to expand the Sir2-HAP circuit to more effectively simulate yeast aging trajectories, including identification of novel modes of aging. This includes leveraging sequencing and imaging approaches, as discussed in the previous section, and elucidating whether there is bifurcation in the dynamics of additional pathways associated with aging (see Outstanding Questions). The lifespan-extending effects of SIRT1, the mammalian homolog of Sir2, have also been observed in mice when SIRT1 is activated either pharmacologically[60] or through brain-specific over-expression[61] in mice on a standard diet, suggesting its relevance in mammalian aging models.

Outstanding Questions.

• As illustrated by the divergence between nucleolar silencing and heme biosynthesis in yeast, how do the interactions between the aging-related GRNs discussed here and other networks shape distinct aging trajectories, and what regulatory mechanisms underlie these interactions?

• What is the cell type-specific resiliency of negative feedback loops, and which compensatory pathways are the first to deteriorate during aging across different tissues?

• How does cellular heterogeneity within tissues contribute to aging, and can targeting specific cell populations slow down or reverse tissue aging?

• For longer-lived organisms, how can we integrate single-cell technologies to predict the dynamic activities of GRNs and their eventual breakdown during aging?

In addition to work on the Sir2-HAP circuit in aging, there have been a growing number of studies that used systems approach to unravel emergent quantitative properties of aging. For example, Uri Alon’s group developed a paradigmatic model to simulate the dynamics of damage accumulation and cell death. In their model, the aging process can be described as a competition between accelerating damage accumulation and saturating damage removal, in which cell death occurs when the damage level exceeds certain thresholds. This model can help quantitively explain characteristic dynamic features of aging, such as the Gompertz law[62], and can recapitulate aging dynamics and survival curves across different organisms[63–65]. Murat Acar’s group used microfluidics-based time-lapse imaging to monitor replicative aging in yeast cells and revealed emergent single-cell properties during aging, such as transcriptional noise reduction[66] and lifespan scalability[67] (see reviews[68,69] for details). A more comprehensive review on systems analysis of aging across different scales from single cells to human physiology is provided by Cohen et al.[70].

A general design principle of gene circuits for longevity

A general theme from the studies discussed above can be summarized as follows: Aging drives a deviation of important functional proteins from their “healthy” ranges, leading to cell deterioration and eventually cell death. Gene circuits embedded around these key factors can either accelerate or decelerate this process. Among these circuits, negative feedback loops function to dynamically counteract the effects of aging and thereby could serve as a general design principle for longevity. In many cases, such feedback loops can give rise to oscillations in the level of key functional factors, enabling a dynamic homeostasis around the healthy state. This section discusses different pathways that contain endogenous negative feedback loops with a potential pro-longevity function. Speculation is offered on how aging may disrupt these feedback circuits, leading to functional decline in aged cells. Strategies are proposed to enhance the robustness and resilience of these circuits that may mitigate the effects of aging.

Proteasome feedback circuit

Loss of protein homeostasis is a conserved hallmark of aging[4,71], associated with neurological age-related pathologies, such as Alzheimer’s disease and Parkinson’s disease[72]. Studies in model organisms revealed that damaged/misfolded proteins accumulate in aging cells, causing cell deterioration and age-related pathologies[73,74]. The proteasome is a multi-subunit proteolytic complex that degrades proteins marked by the attachment of small ubiquitin peptides which functions as the primary intracellular machinery responsible for removing damaged proteins in aging[75].

Proteasomal activity is a conserved mechanism for protein degradation in eukaryotes and its capacity is modulated in response to the level of damaged proteins through intricate feedback mechanisms. For example, in yeast, the abundance of the proteasome is regulated by a negative feedback loop[76]. In this case, Rpn4 is a transcriptional activator required for expression of the genes encoding proteasomal components. However, it is also a target for degradation by the assembled, active proteasome and therefore is extremely short-lived (Figure 3). Under normal physiological conditions this feedback circuit maintains homeostasis between the proteasome and damaged proteins. An increase in the amount of damaged proteins, which compete with Rpn4 for the existing proteasome pool, can reduce Rpn4 degradation and thereby elevate the Rpn4 protein level. Rpn4, in turn, upregulates the amount of the proteasomes to remove excess damaged proteins. Once damaged proteins are cleared, Rpn4 is rapidly degraded by the proteasomes, bringing the proteasome pool back to the its basal level.

Figure 3. Regulation of the proteasome feedback loop and its deterioration with age.

The proteasome is a multi-subunit complex responsible for degrading damaged proteins, maintaining protein homeostasis. Rpn4, a key transcriptional activator, upregulates the expression of proteasome genes and is degraded by the active proteasome, forming a negative feedback loop. As damaged proteins accumulate, Rpn4 degradation decreases, which in turn increases proteasome expression to restore balance. However, during aging, the transcriptional capacity of proteasome genes decreases due to disruptions in regulatory elements like Rpn4 and transcriptional activators such as Hsf1. This breakdown leads to reduced proteasome activity, causing further protein damage accumulation and ultimately cell death. Interventions are proposed to enhance the resiliency of the proteasome feedback loop. These include engineering Rpn4 and proteasome gene promoters to increase their transcriptional capacity and introducing an alternative negative feedback loop (blue dashed lines) that is less sensitive to age-related transcriptional declines, potentially leading to sustained proteostasis which may delay aging.

At early phases of aging, this feedback circuit functions to remove age-induced damaged proteins and maintain the proteasome homeostasis. However, during aging, reduced transcriptional capacity of the proteasome genes compromises this GRN[77,78]. Single-cell RNA-sequencing over the lifespan of C. elegans revealed that the most commonly down-regulated genes across cell types were those involved in proteostasis within the endoplasmic reticulum (ER), including the chaperone hsp4/BiP[78]. An age-dependent decline in the activity of heat shock factor 1 (HSF1), a conserved transcriptional activator of the proteasome genes[79], has also been observed in various tissues and organisms[80–82]. These age-induced changes take effect together and push the feedback system out of its homeostatic regime, leading to further reduced proteolytic activity and uncontrolled protein damage accumulation and eventually cell death, as observed in various types of organisms[83–88][73,89,90] In support of this scenario, deletion of RPN4, leading to a reduced proteasome pool, shortens the yeast lifespan, whereas deletion of UBR2, a ubiquitin ligase that mediates Rpn4 degradation by the proteasomes, leads to elevated proteasome capacity[91] and extends the lifespan[92].

A brute force strategy to overcome the effects of aging is to simply increase the proteasome pool by gene deletion or overexpression, which can indeed lead to lifespan extension to some extent[92]. However, constitutive overexpression or overactivation of the proteasomes has detrimental side effects and impairs cell growth and viability[93,94]. Therefore, a more delicate and effective strategy to enhance the resiliency of the proteasome feedback circuit can be achieved by engineering the promoters of key factors in the circuit. For example, increasing the transcriptional capacity and dynamic range of Rpn4 and proteasome genes may endow a more robust response against age-induced alterations (Figure 3, blue dashed lines). Alternatively, an orthogonal negative feedback loop that is less sensitive to the effects of aging can be introduced to drive the expression of proteasome genes. Such efforts can be built on previous work that created synthetic gene oscillators using standardized transcriptional control elements[95–97].

Energy homeostasis circuit

Metabolic reprogramming is associated with aging, and modulating energy intake or metabolic rate can dramatically influence longevity[98–100]. For example, mTORC1 is well known to be negatively associated with lifespan[101] . Deletion of mTOR in yeast extends lifespan[102] . Depleting mTORC1 activity either through deletion of its downstream effector S6 kinase or by introducing a hypomorph[103,104] leads to increased survival in mice. Pharmacological inhibition of mTORC1 by rapamycin has shown increase in lifespan in yeast, worms, flies, and mouse[105].

It was recently found that the divergent aging trajectories of single yeast cells are associated with distinct metabolic changes – mode 1 aging features a transition from fermentation to respiration, whereas mode 2 aging features enhanced glycolysis and suppressed respiration, which may result in changes to intracellular ATP[50]. Importantly, many studies showed that ATP levels decline with age in various organisms, indicating that loss of energy homeostasis is a major hallmark of aging[50,106,107].

Maintaining energy homeostasis requires balancing ATP production and consumption. The AMP-activated protein kinase (AMPK) is a primary energy sensor, conserved throughout eukaryotes[108–110]. AMPK senses increases in the intracellular ratio of AMP/ATP and promotes catabolic pathways to generate ATP. Under normal physiological conditions when ATP is sufficient, AMPK is inactive. However, when the intracellular ATP level goes down, AMPK becomes active and phosphorylates metabolic enzymes and transcription factors to promote ATP-producing catabolic pathways and inhibit ATP-consuming biosynthetic pathways[111] (Figure 4). AMPK is switched off when the ATP level is restored to its normal state.

Figure 4. Regulation of AMPK and its role in maintaining energy homeostasis during aging.

When intracellular ATP levels drop, AMPK becomes active and promotes ATP-producing catabolic pathways while inhibiting ATP-consuming biosynthetic pathways. During early aging, this system remains effective. However, as aging proceeds, mitochondrial dysfunction reduces ATP production, and AMPK’s responsiveness declines, leading to energy imbalances and cell death. Aging drives the dysfunction of energy production, whereas AMPK promotes catabolic pathways to restore ATP levels. Decreased AMPK sensitivity disrupts energy balance, pushing the cell towards death. A proposed intervention (blue dashed lines) includes introducing an additional AMPK gene copy under a negative feedback loop to dynamically regulate AMPK expression and counteract its age-related decline in activity. This intervention aims to restore energy balance, while preventing the determinantal effects of AMPK overexpression on longevity.

At early phases of aging, the AMPK pathway is sufficient to counteract age-induced ATP changes and maintain energy homeostasis of the cell. During aging, mitochondrial dysfunction causes a substantial decrease in ATP production. However, at the same time, the responsiveness of AMPK declines with age[112–116]. Whereas the mechanisms underlying this reduction remain unclear, it may be related to age-dependent elevation in the expression or activity of protein phosphatases that function to turn off AMPK[116]. The decreased sensitivity of AMPK breaks the balance between ATP sensing and production, resulting in uncontrolled energy deficiency in aged cells. In agreement with this, boosting AMPK activity slows aging and extends lifespan[117–119], whereas deletion or inhibition of AMPK shortens lifespan[120].

Similar to Sir2, HAP, and the proteasomes, although AMPK is generally considered a pro-longevity factor, constitutive overexpression or over-activation of AMPK can elicit negative effects on various physiological processes[108,121] and lifespan[120]. A potential strategy against the effects of aging could be to engineer a robust negative feedback loop to drive the expression of AMPK. This could be achieved by introducing an additional copy of the AMPK gene under an AMPK-repressible promoter (Figure 4). Governed by this circuit, the expression of AMPK can be dynamically adjusted based on its activity and can counteract its reduced responsiveness in aged cells.

p53 feedback circuit

p53 is a central transcription factor within the GRN that responds to DNA damage, thusly dubbed the guardian of the genome. The major regulator of p53 dynamics is its transcriptional target and repressor MDM2, yielding a negative feedback loop (Figure 5). p53 was predicted and later experimentally verified to oscillate in response to DNA damage[122–126]. Upon DNA double-stranded break (DSB), the MRN complex (including Mre11, Rad50, and Nbs1) recognizes these breaks and recruits ATM to the site[127–129]. Upon recruitment, ATM becomes autophosphorylated and activated as a serine/threonine kinase[130,131]. Phosphorylated ATM then upregulates p53 levels by inhibiting MDM2, preventing its ubiquitination and degradation of p53, and by directly phosphorylating p53, enhancing p53’s stability and activity[132,133]. p53 subsequently promotes transcription of hundreds of genes in a cell-type dependent manner[134], including its negative regulator MDM2. The promoter of the gene MDM2 has a relatively low activation threshold making it highly sensitive to p53 pulse duration[135], enforcing p53 homeostasis. If there is continual stabilization of p53 by ATM, the cycle occurs again, yielding the observed p53 oscillatory behavior. These oscillations enable repetitive surveillance and maintenance of DNA integrity in the form of G1 and G2 cell cycle arrest without surpassing thresholds that would trigger apoptosis or senescence[136] (Figure 5). In mathematical models, parameters influencing the rates of protein production and degradation, particularly that of MDM2, can explain the differences in pulse amplitudes[123,124]. Experimentally, p53 pulses can be modulated by addition of Nutlin-3, an inhibitor of MDM2, demonstrating that MDM2 is the major influencer of p53 oscillatory dynamics[135].

Figure 5. The p53-MDM2 negative feedback loop displays an aging-dependent bias towards senescence.

The central role of the p53-MDM2 negative feedback loop is maintaining genomic stability and regulating cell fate decisions in response to DNA damage. p53 activates the transcription of MDM2, forming a feedback loop that limits p53 activity and prevents its overactivation. Oscillatory p53 dynamics enable repeated surveillance of damaged DNA, promoting cell cycle arrest to allow repair, or initiating apoptosis if the damage is irreparable. Prolonged or excessive p53 activation due to persistent or excessive DNA damage can trigger apoptosis or senescence. However, aging appears to bias cell fate towards senescence, contributing to the SASP and inflamm-aging. Targeting the p53-MDM2 interaction to modulate senescence and apoptosis presents a potential strategy to mitigate the increase of senescent cells observed throughout aging.

The p53 network includes positive and negative feedback mechanisms to precisely regulate p53 activity; negative feedback prevents p53 hyperactivity while positive feedback supports p53’s crucial role in activating DNA damage response pathways. Therefore, disrupting members of the p53 network that regulate its activity has significant consequences on health and lifespan. Deletion or inactivation of p53 is observed in more than half of all cancer types, and mice and zebrafish with genetic depletion of p53 often develop various types of tumors[137,138]. In contrast, mutations that result in hyperactivity of p53, such as truncations at the C’ terminus, can lead to a significant reduction in lifespan and are associated with aging-related characteristics [139–141]. p53, therefore, exerts a Goldilocks effect, where its activity must be precisely regulated—not too little, to prevent uncontrolled cell proliferation and cancer, and not too much, to avoid excessive cell death and accelerated aging. This is in part exemplified in mouse models with mutations that weaken p53, as mice that do not develop tumors tend to have longer lifespans[142].

Germline mutations in MDM2 can result in premature aging due to increased p53 activity[143]. Specifically, targeting MDM2 deletion in the epidermis has been shown to cause premature skin aging[144]. Pharmacological inhibition of MDM2 decreases the activity of senescent cells in primary human fibroblasts, reducing senescence-associated secretory phenotype (SASP) factors[145]. Mice with MDM2 mutations with a reduced ability to inhibit p53 exhibit premature aging phenotypes[146]. Disrupting p53’s ability to bind to MDM2 reduced the number of senescent cells in skeletal muscle and enhanced its regeneration and repair[147]. Protein produced from a common splice variant of MDM2, MDM2-a, identified in both tumors and normal tissues, can bind to full-length MDM2 and inhibit the p53-MDM2 interaction, leading to a reduced lifespan of approximately 20%[148–150].

In the event of weakened negative feedback or stronger positive feedback, p53 displays dynamics that may induce senescence. Although senescence is an important safeguard to prevent tumor formation, the SASP has detrimental consequences[151]. Senescent cells are themselves one of the twelve hallmarks of aging, as the number of senescent cells increases throughout lifespan[4,152–154]. This increase is accompanied by chronic inflammation due to the SASP and can lead to various chronic disorders and tissue dysfunction.

RNA-sequencing of cells that enter senescence exposed to ionizing radiation revealed that up-regulation of FOXO4 sustains survival of senescent cells despite presence of pro-apoptotic genes like PUMA and BIM[155]. Interestingly, FOXO4 has been shown to colocalize with promyelocytic leukemia (PML) bodies, nuclear structures that are a hallmark of senescence and are known to restrain p53 activity[155–157]. Inhibiting FOXO4 allows p53 to be released from the nucleus, enabling its translocation to the mitochondria, where it can interact with apoptotic machinery and trigger cell death. Therefore, down-regulation of factors like FOXO4 in senescent cells, which are necessary for inducing and maintaining senescence, is a strategy which may enable apoptosis (Figure 5). FOXOs are a family of transcription factors that activate GRNs associated with stress response[158]. In yeast, over-expression of FOXO transcription factors extend lifespan while deletion shortens lifespan[159]. Over-expression of FOXO1 in mice does not significantly alter lifespan, however, because FOXO activity is suppressed by insulin/IGF-1 signaling, many lifespan-extending interventions in mice, including a brain-specific mutation and nutrient limitation, target this pathway[160].

Another approach could be to modulate cell-cell communication between senescent cells and neighboring cells, an approach known as senomorphics[161]. In this approach, rather than promoting apoptosis, the goal is to prevent the secretion of soluble factors, including pro-inflammatory cytokines (e.g., IL-6, IL-8, IL-11)[162–164]. IL-11, for example, can activate the ERK-mTOR pathway in cells, inducing senescence. Repressing expression of these cytokines specifically in senescent cells could be programmed by utilizing promoters of genes that specifically respond to sustained p53 signaling genes such as PML[136] or CDKN1A[135] to drive expression of IL-10, which may aid in attenuating the inflammatory response of senescent cells[165].

Concluding Remarks

GRNs orchestrate highly tuned, evolutionarily conserved responses to a variety of stimuli. These networks are dynamic, responding not only to binary signals but also to more nuanced factors such as the amplitude, duration, and gradients of stimuli or transcription factors. Aging, often characterized by a breakdown of these regulatory systems, provides a unique opportunity to understand and potentially optimize these dynamics for improving health.

Throughout the lifespan, gene circuits, like the Sir2-HAP circuit in yeast, govern critical cellular processes, maintaining a balance between cellular states such as rDNA silencing and mitochondrial function. However, as cells age, this balance deteriorates, pushing cells into detrimental modes of regulation, ultimately leading to death. By studying these breakdowns in gene circuits, it is possible to identify points of intervention, enabling synthetic engineering of prolonged lifespan through targeted regulation of these circuits.

Aging-related negative feedback circuits maintain cellular homeostasis through push-and-pull mechanisms and failure to remain resilient can accelerate aging (see Outstanding Questions). Approaches to enhance the robustness of these circuits, either through synthetic biology or nutrient limitations, can extend the resiliency of cells against the accumulation of age-related damage. Notably, these circuits do not act in isolation; they are interconnected, as seen in the Sir2-HAP system, where shifts in one pathway influence others. Additionally, pathways do not just affect one another in a cell intrinsic manner but also neighboring cells, including stem cells and cell types from other tissues. These cell intrinsic pathways can collectively contribute to aging in a cell extrinsic manner.

For example, hematopoietic stem cells (HSCs) are a well-studied population that undergo cell-intrinsic aging while also exerting systemic effects on overall organismal aging, including reduced immune responses and changes in HSC fitness[166] . HSCs can be genetically modified, enabling researchers to perturb these systems and examine their emergent effects on lifespan[167]. By focusing on these stem cell populations, researchers can model aging trajectories and test interventions without introducing excessive complexity from the multitude of cell types, interactions, and tissues in multicellular organisms. These approaches provide a scalable way to study aging while linking cellular and systemic perspectives.