Session Z44: Quantitative Insights into Cancer Evolution

Quantifying metastasis heterogeneity in colorectal cancer

Metastasis is a process that in humans takes place completely unobserved, over long time periods, and we know very little about it. Animal models are valuable tools for studying metastasis mechanistically, but only if we have an accurate picture of the metastatic process in humans, so that we can select the most appropriate models and ask the right questions. One our lab’s main interests are host organ-specific metastasis patterns. It is currently unknown whether metastases in different organs form through distinct evolutionary mechanisms, although different metastasis types display divergent clinical behavior. We have evidence to suggest that the characteristics of colorectal cancer metastases - in particular their genetic diversity and their phylogenetic relationships to the primary tumor – differ by host organ. We are taking advantage of a unique collection of multi-region sampled metastatic colorectal cancer cases that we have been building over years, encompassing more than 1000 distinct tumor samples from more than 100 patients. We have reconstructed phylogenetic trees that visualize the evolutionary history of these cancers and find recurrent, quantifiable differences in the position of different metastasis types on the phylogenetic trees. For example, liver metastases are homogeneous and diverge late in tumor evolution, indicating late metastatic seeding by single cells or small groups of cells. Their inter-lesion uniformity suggests selection for a liver-competent growth phenotype. Locoregional lymph node metastases, in contrast, are highly diverse and associate closely with the primary tumor, indicating continuous seeding and comparatively mild selection pressure. The unique phylogenetic patterns associated with different metastasis type encode important information about formation mechanisms and have implications for surgical and medical treatment.   

Quantifying the somatic evolution occurring in healthy and pre-malignant human tissues

Many human tissues are maintained by large numbers of long-lived stem cells. Positive and negative selection on somatic mutations that occur in these stem cells can drive rapid evolution in human tissues over timescales of years to decades. Blood is an ideal system for gaining a quantitative understanding of these dynamics because it is easily sampled, less spatially structured relative to other tissues and genomically well characterised. In this talk I will describe how large-scale sequencing data from hundreds of thousands of individuals can be used to provide quantitative estimates for the key parameters governing clonal evolution including the “fitness effects” conferred by specific variants, the mutation rates of alterations and the population sizes of cycling stem cells. These quantitative analyses reveal the majority of events driving clonal expansions in blood (“clonal haematopoiesis”) occur outside of canonical cancer genes suggesting a large number of “missing” driver mutations yet to be discovered. I will show recent work that exploits unique collections of serial blood samples to reveal how expanding somatic clones arise and compete with each other in healthy individuals and in people destined to develop a future leukaemia. These data shed light on the evolutionary dynamics occurring in pre-cancerous stem cells and suggest that many aspects of the observed dynamics can be understood within a surprisingly simple model of the evolutionary dynamics.   

Predicting cancer evolution: quantifying immune selection and other selective pressures

Tumors are heterogeneous populations of cancer cells that evolve and adapt, with the fittest clones surviving and, possibly, leading to metastasis. Certain mutations can confer fitness advantage to cancer cells, for instance, by altering function of crucial proteins. Mutations can also lead to fitness disadvantage, notably by giving rise to neoantigens – mutated protein peptides that may become foreign enough to be recognized by patient’s own immune system. Here we develop a biophysically grounded model to quantify immunogenicity of tumor neoantigens. We use the model to define the fitness of tumor clones as a balance between negative selection due to immune recognition and positive selection resulting from oncogenic mutations. We apply the model on various cohorts to determine the selective pressures that govern the evolution under different therapy and endogenous immune activity scenarios. In particular, we investigate how pancreatic cancers – in general a lowly mutated, poorly immunogenic cancer, largely presumed to not be subject to immunoediting – evolve over years. With the fitness model, we show that the tumors of rare long-term survivors show signatures of evolution under strong negative selection due to immune recognition. Their tumors have relatively less clones and are deprived of highly immunogenic neoantigens. Importantly, the fitness model predicts the clonal composition of recurrent tumors of the patients. Thus, we submit longitudinal evidence that the human immune system naturally edits neoantigens. Furthermore, we present a model that describes how tumor cell populations evolve under immune pressure over time, with implications for design of cancer treatments.   

Characterizing the T cell response to a personalized cancer vaccine

Recent technological advances in sequencing and mRNA vaccines have allowed for the potential new immunotherapy strategy of personalized cancer vaccines. A patient’s primary tumor can

be resected and sequenced to computationally identify mutated, tumor-specific peptide fragments that are likely to generate a T cell immune response. These putative targets, called neoantigens, can then be used to design a custom mRNA vaccine against the patient’s specific cancer. In collaboration with Genentech and BioNtech, we recently completed and published a phase 1 trial for a personalized cancer vaccine in pancreatic cancer (Rojas et al., Nature 2023).

In this talk I will discuss how to leverage T cell receptor (TCR) repertoire sequencing to characterize and monitor responses to the vaccine. Serial bulk TCR sequencing facilitates the identification and frequency tracking of vaccine responsive T cell clones while paired single cell RNA-TCR sequencing enables phenotyping of the same clones. Together, these data show that patients can mount a durable, long-lived response to the vaccine.   

Finding Mechanisms of Resistance In Cancer

Cancer evolves by sequentially acquiring driver events that increase the fitness of the cells, going from normal, to pre-cancer clonal expansion, to cancer and finally to becoming resistant to therapy. The cancer cells can leverage different mechanisms to evade the therapeutic pressure and develop drug resistance. In order to delay or prevent resistance, one needs to map and understand these mechanisms which can then lead to targets for developing new drugs as well as guide the selection of combinations of therapies that have non-overlapping mechanisms of

resistance. In this presentation, I will discuss different approaches for finding mechanisms of resistance, including studying pre-treatment and post-resistance samples; estimating clonal fitness; and identifying convergent evolution across subclones. Rich phylogenetic trees can be used to identify convergent evolution, a strong statistical signal of positive selection. Such trees can be built based on analyzing multiple tumor samples from the same patient, obtained from liquid biopsies and rapid autopsies. I will give examples from studying resistance in breast cancer and describe novel analytical approaches for detecting convergent evolution. Finally, I will demonstrate the benefit of using whole-genome sequencing data to further enrich the detection of resistance mechanisms.