From killing cancer cells to changing what they can become
Cancer cells are remarkably adaptable. They shift between cellular states, regain stem-like properties, resist therapy and disseminate to distant organs. This plasticity is indeed one of the biggest hurdles in oncology. But what if, instead of trying only to eliminate cancer cells, we could make them less able to change?
This question is at the heart of our new study: transient modulation of cannabinoid receptor 2 (CB2R) can durably restrict breast cancer cell plasticity by stabilizing a luminal-like state. But this was not how we originally approached the problem.
When cannabinoids were not a differentiation story
I first became interested in cannabinoids during my PhD. Like many groups studying their antitumor activity, we worked with relatively high concentrations and prolonged exposures, analyzing experiments shortly after drug removal to see if we could damage or eliminate cancer cells.
Years later, during my postdoc, my perspective changed. Working with embryonic and induced pluripotent stem cells exposed me to a completely different way of thinking about cellular identity. The story began with an informal coffee at the Spanish National Cancer Research Centre with Manuel Serrano. He had just published his landmark 2013 Nature study showing that somatic cells could be reprogrammed in vivo, and that conversation sparked my decision to venture into stem cell biology.
The learning curve was enormous, but one lesson stayed with me: cellular identity is not fixed: Transient perturbations can leave lasting consequences. This led to our 2020 EMBO Journal study showing that transient exposure to miR-203 could enhance the differentiation capacity of pluripotent stem cells. When I established my own laboratory—the Cancer STEM Lab—I wanted to bring these ideas back to cancer and explore differentiation-based therapies.
A change in perspective
Working with stem cells gradually changed how I thought about cancer biology itself. I began to question a core assumption of molecular biology: that complex biological phenomena can always be understood by breaking them down into linear mechanisms.
This realization took me beyond biology. I became increasingly interested in the quantum physicists who helped shape the foundations of molecular biology. Schrödinger’s 1944 What Is Life? asked how the laws of physics could give rise to the remarkable order of living systems, inspiring Watson, Crick, and a generation of molecular biologists.
However, while molecular biology grew from this dialogue with physics, we biochemists reduced biological questions into tractable linear mechanisms as the field matured. The central dogma did provide a powerful framework, but biological systems are not linear machines: they are dynamic, noisy and context dependent. Signaling pathways feedback, epigenetic states influence transcription while transcription reshapes state and collective behaviors emerge. Uncertainty is indeed the law.
I became convinced that solving the hardest problems in cancer requires revisiting this systems-level perspective. Concepts from biophysics, statistical physics and complex systems provide a language for thinking about biological states as dynamic landscapes where cells move, fluctuate... and can sometimes be trapped in stable states.
This was the perspective I brought when returning to cannabinoids—looking at them through a different lens.
What happens if we change the dose—and the question?
Instead of asking whether cannabinoids kill cancer cells, we asked whether they could change cancer cell state.
Importantly, moving toward lower doses and brief exposures is strongly supported by therapeutic and translational rationale. High-dose continuous regimens often fail clinically due to psychoactive effects, toxicity and regulatory hurdles. Reducing exposure widens the therapeutic window, making clinical translation feasible.
This translational viability aligns with the growing global recognition of medicinal cannabis. In Spain, for example, Royal Decree 903/2025 established a legal framework allowing hospital specialists to prescribe standardized cannabis preparations (THC/CBD) controlled by the AEMPS for refractory conditions such as chronic pain or chemotherapy-induced nausea. Aligning our basic research with low clinically realistic exposures helps bridge experimental cancer biology with actual therapeutic scenarios. Thus, we moved away from high-dose cytotoxicity, using low nanomolar concentrations and brief exposures before removing the compounds. Surprisingly, the cells did not return to their original state: The effects persisted.
We tested two ways of modulating CB2R: THC and the selective ligand SR2. Both produced remarkably similar long-lasting effects. Treated organoids lost stem-like and invasive properties, became more responsive to tamoxifen, and showed reduced tumor-initiating capacity. Even when challenged with potent dedifferentiation signals (TGF-β, stromal, immune and mechanical cues), the cells remained resistant to returning to a plastic state. In vivo, the effects persisted for over 100 days after drug withdrawal. The drug was gone, but the phenotype remained.
Were the cells really becoming luminal?
Proving that cells had differentiated toward a luminal-like state was our next challenge.
Flow cytometry showed modest, reproducible shifts in CK8 expression, but no clean binary transition from basal to luminal populations. This forced us to think more carefully about "differentiation." Rather than complete binary lineage conversion, our data supported the durable stabilization of a luminal-like state.
Looking across biological layers revealed a coherent picture: temporal progression of luminal regulators (GATA2, FOXA1), followed by increased ER and GREB1 expression, enriched estrogen signaling and coordinated chromatin remodeling (via CUT&Tag). Functionally, cells became less self-renewing and invasive, while gaining endocrine sensitivity.
No single marker told the story—the phenotype emerged from the convergence of molecular, epigenetic and functional changes.
Another way of thinking about differentiation therapy
Traditionally, differentiation is imagined as a switch from one state to another, but tumors are heterogeneous populations occupying dynamic cellular states. The therapeutic goal does not have to be complete lineage conversion—it may be enough to stabilize the cell state. A cancer cell that can no longer return to a plastic, stem-like state is therapeutically very different, even if it retains intermediate features. Thus, the crucial property may not be what the cancer cell becomes, but what it can no longer become.
Where do we go from here?
Our findings raise key questions about how transient CB2R modulation initiates persistent chromatin changes and whether this approach can be combined with existing therapies to sensitize cells to treatment. What if, instead of only trying to eliminate cancer cells, we could make them less able to change? Our findings suggest this is a direction worth exploring.