Can We Spot Adaptation in Tiny Samples and Chance?

Jenn Hoskins
25th June, 2025

Can We Spot Adaptation in Tiny Samples and Chance?

The study’s mathematical models are validated by accurately capturing the average phenotypic distribution (e–f) and the full stochastic cell count dynamics (c–d) observed in the underlying individual-based simulations of continuous (a, c) and intermittent (b, d) drug treatment.

Image adapted from: Browning et al. / CC BY (Source)

Key Findings

  • The study from International Universities & Research Centers found that standard population-level cell count data cannot reliably show how much individual cancer cells differ in their drug resistance
  • Furthermore, these common cell count measurements also fail to distinguish if drug resistance emerges as distinct cell types or as a gradual shift across a spectrum of resistance
  • To truly understand this cellular adaptability, more detailed single-cell data, like observing individual cell growth and death events, are needed
A significant challenge in treating cancer is the emergence of drug resistance, where cancer cells find ways to survive therapies that were initially effective. This resistance often leads to treatment failure and patient relapse. One key mechanism driving this problem is what scientists call "phenotypic plasticity." This refers to the ability of cancer cells to change their characteristics or "phenotype" in response to stress, such as drug treatment, without necessarily acquiring new genetic mutations. Understanding and quantifying this cellular adaptability is crucial for developing more effective and lasting cancer treatments. Recent research from International Universities & Research Centers has shed new light on this complex issue[1]. Their work addresses a critical gap in our ability to characterize and distinguish different forms of drug resistance arising from this cellular flexibility. Phenotypic plasticity allows cancer cells to transform into a state where they are no longer dependent on the specific pathway targeted by a drug[2]. These "drug-refractory" cells can become slow-cycling, meaning they divide less frequently, and can either regain their sensitivity if the drug is stopped or, more critically, acquire permanent resistance, leading to the cancer's return[2]. This phenomenon has been observed across various cancers, from lung and prostate cancer to melanoma, highlighting its widespread impact on treatment outcomes. The underlying mechanisms involve complex changes, including the reprogramming of cell identity and the restructuring of their internal genetic material, known as chromatin remodelling[2]. Cancer cells are also known to be highly adaptable, co-opting normal stress-coping mechanisms to survive the harsh conditions of tumour growth, spread, and even immune system evasion[3]. Cancer therapies themselves act as a major stressor, prompting these rapid, non-genetic adaptive responses that help cells stay alive[3]. This adaptability contributes to what is known as "tumour heterogeneity," meaning that even within a single tumour, cells can be very different from one another[3]. For instance, studies have shown that human melanoma cells can exhibit significant variability at the single-cell level, with a very small percentage of cells already showing high levels of resistance markers even before drug exposure[4]. This "rare cell variability" involves a transient, non-genetic state that can be converted into stable resistance through epigenetic reprogramming once the drug is introduced[4]. This highlights that resistance isn't always about new mutations but can stem from pre-existing, non-genetic differences among cells. Despite the growing recognition of phenotypic plasticity and its role in resistance, a major hurdle has been the lack of quantitative tools to precisely measure and understand it. Specifically, it has been difficult to determine if resistance emerges as a "discrete phenotype"—meaning a distinct, separate type of cell that is resistant—or as a "continuous distribution of phenotypes," where cells simply exist along a spectrum of varying resistance levels. The research from International Universities & Research Centers aimed to tackle this by developing a sophisticated computational tool. They created a "stochastic individual-based model" to simulate how individual cancer cells adapt their phenotype. "Stochastic" means the model incorporates an element of randomness, reflecting the inherent variability in biological processes, while "individual-based" means it tracks the behavior of single cells rather than just averaging across a whole population. This model was designed to simulate cells in low-cell-count proliferation assays, which are common laboratory experiments used to study cell growth and division. Importantly, their model was formulated to align probabilistically with common "partial differential equation models," which are mathematical tools often used to describe how cell populations change over time and space. This allowed them to develop a way to quantify the "intrinsic noise"—the natural, random variations—that are always present in such experiments. Through their analysis, the researchers assessed how well key parameters of their model, such as the speed of adaptation and the extent of cell-to-cell heterogeneity, could be identified from different types of experimental data. Their findings revealed two significant insights. First, they discovered that "cell-to-cell heterogeneity"—the individual differences between cells—is practically impossible to identify reliably using standard "population-level data," such as simple cell counts or general markers of cell proliferation. This implies that even if individual cells within a tumour are vastly different in their resistance capabilities, as suggested by studies like[4] which describe rare resistant cells, our current methods of observing whole populations might make them appear uniform. This finding suggests that for understanding overall population behavior, simpler "homogeneous ordinary differential equation models" (which assume all cells are the same) might be sufficient, even if they mask underlying individual differences. Second, and equally critical, the study demonstrated that population-level data are insufficient to distinguish whether resistance is a discrete phenotype or a continuous distribution of phenotypes. This means that when we look at a large group of cells, it's very hard to tell if resistance arises because a few distinct, pre-programmed resistant cells (like those described in[4]) are present and expand, or if all cells gradually shift along a spectrum of resistance. This limitation has profound implications for how we interpret experimental results and design future studies. The findings from International Universities & Research Centers are crucial because they inform the design of future experiments and the quantitative analyses used to probe phenotypic plasticity in cancer. By highlighting the limitations of population-level data in capturing cell-to-cell heterogeneity and distinguishing between discrete versus continuous resistance, this research emphasizes the need for more sophisticated, potentially single-cell level approaches to truly unravel the complexities of drug resistance driven by phenotypic plasticity. This deeper understanding is essential for developing combination therapies that can overcome cancer's remarkable ability to adapt and evade treatment.

BiotechGeneticsEvolution

References

Main Study

1) Identifiability of phenotypic adaptation from low-cell-count experiments and a stochastic model

Published 24th June, 2025

https://doi.org/10.1371/journal.pcbi.1013202

Related Studies

2) The great escape: tumour cell plasticity in resistance to targeted therapy.

https://doi.org/10.1038/s41573-019-0044-1

3) Therapy resistance: opportunities created by adaptive responses to targeted therapies in cancer.

https://doi.org/10.1038/s41568-022-00454-5

4) Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance.

https://doi.org/10.1038/nature22794


Related Articles

An unhandled error has occurred. Reload 🗙