Tumour evolution: a tale of dispersal and interaction

Chance encounters at international meetings changed the course of my career and led to three recent papers.
Tumour evolution: a tale of dispersal and interaction

In late 2016 I flew to Germany to attend and speak at a conference on modelling tumour evolution at Bielefeld's Center for Interdisciplinary Research (ZiF). Among the conveners was ETH Zurich Professor Niko Beerenwinkel. Niko and I got chatting, he invited me to visit his group in Basel, and I was offered and accepted a postdoc position. During three and a half rewarding years in the Beerenwinkel group I pursued several projects that have recently come to be published, including those I'm going to discuss here.

In early 2018 I was travelling again, this time from Basel to Leiden in the Netherlands for a cancer evolution workshop at the Lorentz Center. There I came across Yannick Viossat, a mathematician based in Paris who'd recently become interested in evolutionarily-informed cancer therapy. During our few days together Yannick and I found we shared the goal of using mathematical models to determine when a treatment strategy aiming for tumour containment, rather than elimination, might achieve better clinical outcomes than the standard of care.

Yannick Viossat at the blackboard in Leiden
Yannick Viossat in Leiden, discussing cancer therapy models.

After the workshop Yannick emailed me: "I think it would be interesting, as a start, to write a short paper... Do you think that would be interesting?" After hundreds more emails, reciprocal visits to Paris and Basel, and numerous video calls, our paper was published in Nature Ecology & Evolution in April 2021. By combining mathematical analysis and numerical simulations, Yannick and I established when tumour containment is and isn't expected to substantially outperform more aggressive treatment, depending on how drug-sensitive and resistant cells interact. Our paper provides formulas for predicting clinical benefits in numerous scenarios, including for practical protocols. The article might not be as short as we'd envisaged, but it's at least as comprehensive.

At the same Leiden workshop, on the walk from the hotel to the venue, I happened to get chatting with Jakob Kather, who'd travelled from Germany. I explained to him that I was trying to determine how a tumour's spatial structure how cells interact with each other and how they disperse influences its mode of evolution (meaning the general pattern such as linear, punctuated, or neutral evolution). I'd developed a flexible computational model to investigate this question but I lacked the expertise to parameterise it. I needed a collaborator who could interpret pathology slides and other clinical data.

It turned out that Jakob was exactly the researcher I needed to meet. Not only was he a clinical oncologist with pathology training, but he also had experience developing computational models like mine, so he knew the prior work and appreciated what I wanted to achieve. I eagerly recruited Jakob to conduct image analyses of pathology slides and to help me tune my models to simulate different tumour types.

This second project was a real team effort. Yannick again contributed, this time using mathematical optimisation techniques to derive a relationship between evolutionary parameters, which helped us categorise tumour evolutionary modes. Talented ETH master's students Dominik Burri and Jeanne Lemant, and research intern Cécile Le Sueur, ran simulations, developed new evolutionary indices, and analysed data. Niko provided invaluable guidance and constructive critiques.

The fruit of this international collaboration is a second Nature Ecology & Evolution paper. Here we show that differences in the range of cell-cell interaction and the manner of cell dispersal are sufficient to generate a spectrum of tumour evolutionary modes. Our model predictions are moreover consistent with clinical data for cancer types with corresponding spatial structures. We conclude that, "By mechanistically connecting tumour architecture to the mode of tumour evolution, our work provides the blueprint for a new generation of patient-specific models for forecasting tumour progression and for optimising treatment."

These two projects have generated many further research questions that I expect will keep me and my collaborators busy for years to come. A reviewer's comment on the more recent NEE paper has already led to a further article in which Jeanne, Cécile, my PhD student Ves Manojlović and I develop new ways of quantifying phylogenetic tree shape.

I'm ambivalent about the future of academic conferences. As a parent of young children, and being keen to minimise my carbon footprint, I'm grateful for the option to take part online. But I'm also certain my recent career would have been quite different without those chance encounters that in-person events are so good at facilitating. Just as beneficial mutations spread faster in larger communities, so we can accelerate our research by making the most of opportunities to interact.

Read the new paper at https://www.nature.com/articles/s41559-021-01615-9.