Introduction: Why It Is Hard to Model Aging and Cancer Together
In earlier parts of this Expert Series, we have discussed:
- Tools to visualize aging (epigenetic clocks, single-cell profiling, image-based metrics)
- Tissue-specific aging profiles and genetic background
- Cancer and cancer therapy as drivers of systemic aging
- Reproductive aging as a hub linking women’s health and cancer
Most of these insights are grounded in human cohort and patient data. Yet to truly understand mechanisms and identify actionable intervention points, we also need models that allow us to manipulate variables and test causality.
This is where we encounter a fundamental challenge: both aging and cancer are long-term, dynamic processes that unfold over years and involve multiple levels of organization:
- Aging is slow, cumulative, and heterogeneous across organs
- Cancer involves evolution, selection, and complex interactions with the microenvironment
Recreating both processes “faithfully enough” in an experimental system is non-trivial.
In this article, drawing on discussions such as those in “Challenges and opportunities for modeling aging and cancer” and related work, we will examine:
- What we actually need to capture when we model aging × cancer
- The strengths and limitations of cell-based, organoid, animal, and human-data-based models
- Emerging directions such as physiological 3D culture and image-based aging metrics
The aim is to provide a structured framework for thinking about models, rather than a catalog of systems.
What Should Models of Aging × Cancer Capture?
1) Time: Chronological vs Biological Age
Any model of aging must confront the gap between chronological and biological age. In animal studies, we often approximate:
- “X-month-old mouse ≈ Y-year-old human”
But in reality:
- Biological age varies between individuals of the same chronological age
- Different tissues in the same organism age at different rates
Therefore, it is increasingly important to supplement “old” vs “young” labels with aging markers—epigenetic clocks, functional readouts, or composite indices—so that we know what kind of aging state we are actually modeling.
2) Scale: From Molecules to Whole-Body Physiology
Aging and cancer span multiple scales:
- Molecular (DNA damage, epigenetic drift, metabolic shifts)
- Cellular (senescence, proliferation, stress responses)
- Tissue (architecture, extracellular matrix, microenvironment)
- Whole-body (immune system, endocrine and metabolic networks, behavior)
Different questions require different levels of fidelity. A model suited for dissecting DNA repair pathways may not be adequate for studying how frailty affects immunotherapy tolerance. The key is to ask:
- “For this specific question, which scales must be represented, and which can we safely abstract away?”
3) Distance from Humans: Translational Relevance
Results from mice or cell culture do not automatically generalize to humans. Species differences in:
- Tumor spectra and mutational landscapes
- Immune system architecture and metabolism
are especially important in aging and cancer. This makes it critical to plan, from the beginning, how insights from a given model can be validated and calibrated using human tissues, organoids, or clinical data.
Cell-Based Models: 2D Culture, Senescent Cells, and Cancer Cell Lines
1) Replicative and Stress-Induced Senescence
Classical aging models in vitro include:
- Replicative senescence, where normal human fibroblasts are passaged until they reach a division limit characterized by telomere shortening, DNA damage responses, and upregulation of p16/p21
- Stress-induced senescence, where radiation, drugs, or oxidative stress drive cells into a senescent-like state
These systems are powerful for studying molecular signatures of senescence and SASP profiles, but they are inherently limited:
- Single cell type, 2D culture, lacking tissue architecture and multi-cellular interactions
2) Senescence-Like States in Cancer Cell Lines
Cancer cell lines can also be driven into senescence-like states by therapies or genetic manipulations. These models are useful for:
- Exploring “pro-senescence” therapies that aim to arrest, rather than kill, cancer cells
However:
- Cancer cell lines are already highly selected and adapted to culture
- They lack the heterogeneity and microenvironmental context of patient tumors
As such, they are best viewed as tools for mechanistic dissection rather than direct surrogates for human tumors.
3D Culture and Organoids: Bringing in Architecture and Microenvironment
1) Advantages of Organoids and Spheroids
Three-dimensional models such as spheroids and organoids introduce:
- Tissue-like architecture and polarity
- Cell–cell and cell–matrix interactions more reminiscent of in vivo conditions
With normal-tissue-derived and patient-derived organoids (PDOs), we can approximate:
- Individual patients’ tumor and normal-tissue biology
Drug responses, differentiation states, and signaling network behaviors often align more closely with clinical behavior than in 2D cell lines.
2) Physiological Aging in Three Dimensions
Yet most organoid models are derived from relatively young donors or from tumors whose microenvironment has already diverged from normal aging tissue. This raises the question: how do we model “physiologically aged” tissues in 3D?
Emerging approaches include:
- Deriving organoids from aged donors
- Long-term culture paradigms that induce aging-like signatures
- Engineering 3D microenvironments that mimic aged tissue properties—matrix stiffness, oxygen and nutrient gradients, inflammatory milieu
These “physiological aging in 3D” strategies aim to embed not just tissue architecture but also key aspects of aged biology into organoid models, enabling more realistic studies of aging–cancer interactions.
Animal Models: Short-Lived Organisms for Long-Term Processes
1) Naturally Aged vs Progeroid Mice
In mice, aging can be modeled using:
- Naturally aged animals, which more closely resemble typical aging but are time- and resource-intensive
- Progeroid models (e.g., DNA repair defects) that show rapid onset of aging phenotypes but may not mirror normal aging
Both approaches are useful, but each has caveats. Progeroid mice highlight mechanisms that are amplified or distorted relative to human aging, whereas naturally aged mice require long-term, carefully controlled studies.
2) Combining Aging with Cancer Models
Cancer models in mice include:
- Chemical carcinogenesis
- Genetically engineered models (e.g., KRAS-driven lung cancer)
- Patient-derived xenografts (PDX)
Overlaying these with aging allows comparisons such as:
- Young vs old hosts with the same oncogenic driver
- Differences in tumor initiation, progression, histology, and immune responses
Some studies have reported counterintuitive findings—for example, oncogenic KRAS may drive more robust lung tumor formation in younger mice, with aging altering tumor suppression pathways and microenvironmental responses. These results remind us that:
- Aging does not uniformly promote cancer; in some contexts it may constrain certain tumorigenic processes while reshaping the overall landscape of risk
3) Species Differences and Tumor Spectra
Critical limitations include:
- Differences in the types of tumors that arise spontaneously in mice vs humans
- Species-specific features of immune, endocrine, and metabolic systems
Therefore, a central task in modeling is to map:
- “This mouse phenomenon most closely corresponds to that clinical scenario in humans”
rather than assuming one-to-one equivalence.
Bridging to Human Data: Single-Cell Omics and Digital Pathology
1) Single-Cell Omics in Tumors and Microenvironments
Single-cell transcriptomics and related methods applied to human tumors and adjacent tissues reveal:
- Aging-like signatures in cancer cells and stromal populations
- Distinct immune-aging profiles in different patient subgroups
These data sets allow us to test whether mechanisms proposed in mouse or organoid models are actually observed in human disease, and to discover human-specific patterns that models must eventually accommodate.
2) Image-Based Aging Metrics Such as ImAge
Methods like “ImAge,” which infer aspects of epigenetic aging from microscopic images at the single-cell level, offer a complementary strategy:
- They can be applied retrospectively to archived histology slides
- They provide a way to quantify aging-like features directly in clinical specimens at scale
As such tools mature, they may enable modeling pipelines in which:
- Mouse and organoid models are benchmarked against human tissues using shared, image-derived aging metrics
This type of “common language” for aging across models and humans could substantially improve the translational value of preclinical work.
Practical Principles for Building Better Models
1) Start from the Question, Not the Model
Because no single model captures everything, it is crucial to begin with a focused question:
- “I want to understand how immune aging alters checkpoint inhibitor responses.”
- “I want to know how an aged stroma affects KRAS-driven tumor growth.”
Only then can we choose the simplest model that still contains the necessary elements. Overly complex models may be difficult to control and interpret; overly simple models may miss essential biology.
2) Assume You Will Need Multiple Models and Iteration
A realistic strategy is to assume from the outset that:
- Mechanisms identified in cell or organoid models will need confirmation in animals and human data
- Patterns observed in human cohorts will require mechanistic testing in experimental systems
This “back-and-forth” approach between models and human data is more robust than relying on any single system. It is also important to agree, within a research team, on:
- What level of evidence (which models, which endpoints) is sufficient to justify moving toward clinical translation
3) Measure Aging Markers in Models, Not Just in Humans
As epigenetic clocks and other aging indices become widely used in human studies, there is a strong case for measuring analogous markers in preclinical models:
- Using aging clocks to characterize baseline and post-treatment states in mice or organoids
- Asking whether a given manipulation makes a model “older” or “younger” in biologic terms
This could help align preclinical experiments with the kinds of aging metrics that will eventually be used in human trials and clinical practice.
Conclusion: Modeling Aging and Cancer Is About Bridging Scales and Timescales
In this fifth Expert Series article, we have:
- Outlined key considerations when modeling aging × cancer (time, scale, and distance from humans)
- Reviewed strengths and limitations of cell-based, organoid, and animal models
- Highlighted new directions such as physiological 3D culture and image-based aging metrics
- Proposed practical principles for building and combining models
Modeling aging and cancer is fundamentally about compressing and reconnecting scales—molecular to organismal—and timescales—from days in culture to decades in human life. No model will ever be perfect, but thoughtfully chosen and combined models can provide “good enough” approximations to guide interventions in real patients.
In future parts of this series, we will examine concrete case studies, such as:
- Interactions between specific oncogenic pathways (e.g., KRAS-driven lung cancer) and host aging
- How aging models inform immunotherapy and targeted therapy development
My Thoughts
Working through the landscape of aging–cancer models, one is reminded that modeling is as much about deliberate omission as it is about inclusion. If we try to capture everything, our models risk becoming as opaque as reality itself. If we simplify too aggressively, we drift away from the clinical phenomena we hope to influence. Navigating this tension—deciding what to keep and what to leave out for a given question—is at the heart of rigorous work in this field.
Another encouraging development is the move toward quantifying “how old” a model actually is, rather than assuming that a certain number of months in a mouse or days in culture map neatly onto human aging. As clocks and image-based metrics migrate from human cohorts into model systems, we may gain a clearer sense of which preclinical results correspond to which segments of the human aging spectrum. That, in turn, could sharpen our instincts about when and in whom to test new interventions.
Ultimately, however, even the best models are tools for refining hypotheses, not substitutes for human evidence. In a domain where both biology and clinical trajectories unfold over years, the judicious allocation of time and resources across different models is itself a critical strategic decision. My hope is that a clearer conceptual framework for modeling—such as the one sketched here—can help researchers and clinicians make those decisions more deliberately, and patients benefit sooner from the knowledge we generate.
This article has been edited by the Morningglorysciences team.
Related Articles
Comments