Despite remarkable progress in recent decades, cancer remains the leading cause of death by disease in children and young adults.

A major reason is the intrinsic ability of some cancers to become resistant and relapse, even after an initially successful response. Unfortunately, once relapse occurs, therapeutic options are scarce and survival rates remain dismal.

In 1976, Peter Nowell proposed that cancer evolves like any other ecosystem, through natural selection of the fittest clone within a diverse population of cells adapting to environmental change. The advent of powerful DNA sequencing technologies confirmed this in adult cancers, revealing rich genetic diversity and clonal dynamics that evolve through complex interactions with the tumour environment. These insights defined cancer as a Darwinian process unfolding in real time.

A different evolutionary path in children

Yet, unlike adult tumours, childhood cancers harbour remarkably few genetic mutations. Very few recurrent DNA alterations are enriched in relapsed or resistant disease, suggesting that paediatric cancers may follow distinct evolutionary trajectories from those of their adult counterparts.

Paediatric tumours arise during development, from cells that are inherently plastic. During embryogenesis, cells with identical DNA can adopt different phenotypes to form all the cell types of the organism. This developmental origin endows childhood cancers with an intrinsic capacity to change their phenotype without altering their genetic code.

Their early onset also means there has not been enough time for Darwinian selection to give rise to a fully resistant phenotype distinct from its original sensitive one. It’s therefore more plausible that a faster and more efficient mechanism drives their evolution. We hypothesise that cellular plasticity is this driving force for the early stages of adaptive evolution in response to treatment in childhood cancers – this allows rapid and reversible phenotypic switching to resistant cell states across many cells simultaneously.

Emerging clinical and preclinical evidence supports this view. Plasticity gives cancer cells the flexibility to enter drug-tolerant, dormant states during treatment. Once treatment is removed, these plastic cells reawaken and regain proliferative capabilities to drive relapse. Cells can effectively reprogramme themselves, turning off drug-sensitive features and adopting transient, therapy-resistant ones.

This adaptability effectively allows them to “play dead” during treatment and later “wake up” and regain – or even enhance – their aggressive form. It’s a striking example of cancer’s evolutionary ingenuity. And because this process occurs quickly – due to its independence from changes in DNA sequence – it represents the first adaptive response to therapy. Yet, standard genomic tests fail to detect it, because they search for mutations, not shifts in cell identity.

Our challenge, therefore, is to capture and understand this non-genetic evolution so that we can intercept the earliest stages of tumour adaptation.

A centuries-old idea revisited

The concept that phenotypes can influence evolution is not new. It has been debated for over a century, most notably through the work of Conrad Waddington and James Mark Baldwin. Both proposed that an initial plastic or behavioural response could confer a survival advantage, allowing a population to persist in a new environment without changes in its DNA. Over generations, natural selection could then favour genetic mutations that stabilise this adaptive trait.

Waddington suggested that such traits could become canalised – that is to say, robustly remain in place in the face of any kind of challenge – while Baldwin proposed that plasticity itself could evolve. These ideas, collectively known as plasticity-first evolution, remained largely theoretical for decades.

Recently, scientists studying side-blotched lizards on the Pisgah Lava Flow have documented this process in remarkable detail. They showed that individual lizards can adapt to a change in the landscape through changing their colour to match the dark volcanic rock, identified the genes regulating pigmentation, and demonstrated that the adapted population had genetically encoded this darker trait. Their findings provide one of the clearest examples of the Baldwin effect in nature.

Plasticity in cancer – evolution at warp speed

In cancer, similar principles may apply but on dramatically faster timescales. Studying plasticity in tumours, however, has been limited by technology. Most analyses rely on single-cell RNA sequencing, which offers only static snapshots of cell states and lacks temporal resolution to capture transitions. Computational models can simulate these changes in two- or three-dimensional landscapes, but direct molecular tools to trace real-time phenotypic transitions have been missing.

To overcome these limitations, my laboratory combines single-cell multiomics, bioengineering, preclinical models of dynamic treatment responses, and experimental evolution approaches.

We have developed a novel molecular tool, engineered to record on a synthetic DNA cassette the evolutionary lineage history of each cell alongside its phenotypic transitions. It uses Cas9 approaches to write and save evolutionary information on the cassette that can later be read and decoded. This creates a temporal multidimensional map of lineage relationships, phenotypic changes, and clonal dynamics. This framework allows us to study plasticity with unprecedented molecular and temporal resolution.

Using this system, we can see that not all cells within a childhood cancer population are equally plastic. Some can transition between differentiated, proliferative states and quiescent, less differentiated ones, whereas others transition only once; or not at all. The timing and frequency of these transitions are variable – very variable – even within an apparently homogeneous population at the DNA level.

By integrating this information with chromatin accessibility data, it’s clear that the epigenetic landscape regulates transcriptional variability at genes controlling cell-state identity. And when it comes to insults these cells face – such as cancer treatment – it seems this induces chromatin rewiring which increases this variability thus expanding the range of possible cell states.

We have also seen that histone modifiers – such as KDM5 – appear to regulate this transcriptional flexibility and create a molecular “memory” that primes the population to adapt more efficiently to future stress.

Towards anti-plasticity therapies

To translate these discoveries into the clinic, we are integrating our molecular recorder with ex vivo drug testing of patient-derived samples. The aim is that this will allow us to develop early warning systems for adaptation, identifying when and how cancer cells begin to reprogramme under therapeutic pressure.

Our goal is not only to predict when relapse is likely but also to identify therapies that block these transitions – to develop an anti-plasticity strategy. This may involve targeting chromatin regulators that enable state switching or exploiting metabolic vulnerabilities specific to dormant, drug-tolerant cells.

By combining live-cell lineage tracing with single-cell multiomics, we are beginning to map the molecular logic of adaptation and, ultimately, to learn how to lock cancer cells into non-adaptive fates. This is a shift from treating cancer as a static disease to managing it as an evolving ecosystem. If we can control its ability to change, we can change its fate and the outlook for children facing these devastating diseases.

Oncology research has long focused on what changes in the genome. We are now realising that the answers to relapse may lie in what does not. Plasticity is the hidden language of adaptation and learning to read and rewrite it may be one of our best chances to prevent relapse in childhood cancers.