ESMO 2018: Big Data, Genomics, and Stealthy Immune Avoiding Tumours

Prof Charles Swanton, MD, PhD.


October 23, 2018

Good afternoon. My name is Professor Charles Swanton. I'm the professor of personalised medicine at University College Hospital and a senior group leader at the Francis Crick Institute. I'll be talking today about the session at ESMO on big data, where myself, Stefan Pfister and Carlos Caldas, talked about advances in lung cancer, CNS malignancies, and breast cancer, using genomic profiling.

In my talk, I covered a study called TRACERx, which stands for TRAcking Cancer Evolution through therapy (Rx) x, which is aiming to try to understand the impact of tumour heterogeneity on disease outcome in early stage, non-small cell lung cancer. And by tracking patients' disease from diagnosis through to death we hope to be able to understand the origins of the lethal metastatic subclone, and most importantly, mechanisms of branch evolution and immune evasion.

I presented data from the first 100 patients in this study, which shows some of the mechanisms that drive truncal evolution of early clonal mutations present in every tumour cell, and mechanisms that drive branched evolution - that is mutations that are present in some cells but not others. And what we find is that tobacco associated carcinogenesis is associated with the early clonal alterations present in every tumour cell, the so-called C to A mutations. In contrast, later on in tumour evolution, that is when a patient presents in the clinic with early stage non-small cell lung cancer, what we find is that the mutational processes now become dominated by a signature called APOBEC, which is a cytidine deaminase, an endogenous enzyme, that appears to be switched on as tumours evolve, creating more mutations that differ between cells.

So what that means is that at presentation, the dominant ongoing mutagenic process is actually an endogenous mutagenic process rather than exogenous carcinogen exposure from tobacco.

Tobacco is still contributing to mutations, but APOBEC mutagenesis is contributing more to ongoing mutagenesis. So, then we asked what are the determinants of clinical outcome in the early stage setting in non-small cell lung cancer? We looked at a point mutation heterogeneity on the one hand, and we looked at chromosomal instability on the other.

And what we find is that it's actually chromosomal instability, that is variation in the degree of chromosomal copy number events that differ between subclones in the same primary, appears to be the greatest and poorest prognostic feature in early stage non-small cell lung cancer.

So then we went on to ask why is it the chromosomal instability, that is the rearrangements in whole chromosomes or parts of chromosomes between cells in the same tumour, why is it that that is associated with poor outcome?

And what we find, is in fact that chromosomal instability appears to be a very potent driver of diversification. Most importantly, it appears to drive diversification and selection through parallel evolution of copy number events. What that means is that for instance, if you got an oncogene like HER2, we find oncogenes like HER2 or CDK4, are amplified on multiple occasions in separate subclones in the same tumour.

So early chaos brought about by chromosomal instability results in later selection of chromosomal copy number events in individual subclones, that presumably endow evolving cells with ongoing fitness that allows them to adapt to environmental pressures.

The next aspect of tumour evolution we looked at was how the cells adapt to immune pressure, that is, to cytotoxic T-cells coming into the tumour. And what happens in terms of selective process that it permits ongoing immune evasion and evolution in the face of that immune pressure. And what we found is that HLA loss of heterozygosity, that is class 1 human leukocyte antigen, or the M.A.C complex as it's known, is recurrently lost in subclones later on in tumour evolution in about 40% of cases.

And it's by far and away, we think, the most common cause of immune evasion, much more common than mutations in HLA or mutations in beta 2 microglobulin. And intriguingly, this allows tumours to now evade the immune system by stealth. In other words, they can develop more mutations that are no longer seen by the immune system.

And so what happens is that subclones that have lost HLA now acquire many more mutations, than subclones adjacent to them that haven't lost HLA.

So in a sense, we think, HLA loss of heterozygosity is actually a mechanism of branched evolution because it permits ongoing evolution by stealth. It enables the prey, if you like, to evade the predatory immune system, simply because the tumour cells are no longer presenting the mutations to the immune system. In other words, they can evade the immune system by stealth and acquire new mutations without being recognised by cytotoxic T-cells.

So, I ended my talk by suggesting what we're going to do about this problem of ongoing evolution and branched heterogeneous mutations and copy number events that occur in some cells, but not others. Essentially, this rapid diversification of tumours that presumably confers selected fitness upon an evolving population of tumour cells.

And the hypothesis that we're working towards now, is that we've noticed in smoking induced lung cancer, the fact that patients with smoking associated lung cancers have very long trunks, what that means is they have many mutations, coding mutations, present in every tumour cell. With those mutations comes the propensity of those mutations to be recognised by the immune system.

And what we've now found are T-cells present in individual tumours, that specifically recognise these clonal or truncal alterations present in every tumour cell. And of course, it's those truncal mutations we want to be targeting, because when we hit those truncal mutations with therapies, we hope to see responses at all sites of disease.

Now what we now are working towards is a first in man study early next year where we hope to be able to treat patients with these T-cells that recognise multiple truncal alterations.

So I finished by saying that chromosomal instability is a force to be reckoned with, we need to understand the mechanistic basis of adaptations over time, we need to understand what drives chromosomal instability. And most importantly, we need to think about how to target it. How to target those mutations present in every tumour cell. And we hope that T-cells recognising those truncal antigens are going to be a key to help us improve outcomes in lung cancer.

Profiling CNS Tumours

The next talk after me was from Prof Stefan Pfister from DKFZ who talked about advances in profiling of central nervous system tumours, and how those advances can ultimately lead to better disease stratification. It can ultimately lead to a better understanding of why some patients do and don't respond to trial interventions.

And he showed beautiful data that convincingly demonstrated that epigenetic profiling of tumours is a very accurate way of enabling the molecular pathologists to segregate central nervous system tumours into better disease categories that will enable us to optimally stratify patients for therapies and actually most importantly, shed light on the molecular basis of the disease that will ultimately help us develop new therapies in stratified patients in the hope of better survival outcomes.

Breast Cancer Integrative Clusters

Last but not least, Carlos Caldas talked about his seminal work in the context of integrative clusters in breast cancer. Now these clusters he's derived through profiling of tumours through copy number aberrations. And he's shown very nicely that these copy number aberrations can segregate breast cancers into distinct integrative clusters that can ultimately give much more detailed understanding of the way in which the disease plays out phenotypically, and most importantly, how that disease behaves.

He gave one very clear example of this amongst many others, but one that sticks in my mind more than others I think was in the ability to be able to segregate hormone receptor positive breast cancer into multiple distinct integrative clusters that have very distinct survival outcomes. And he showed one particular area of unmet need in this group of patients characterised by the 11q amplicon.

That 11q amplicon, on the long arm of chromosome 11, encodes amongst other oncogenes, PAC1 and cyclin D1. And he made a very robust defence suggesting that targeting these tumours with palbociclib and CDK4 inhibitors etc., may not be the answer we're looking for, simply because there are other oncogenes on this amplicon, including PAC1, and combination approaches might need to be considered when thinking about targeting these integrative clusters.

In summary, we're learning a huge amount through deep molecular profiling of tumours. The future is going to be understanding the way in which the microenvironment constrains tumour evolution and understanding how we can impact therapeutically to improve outcomes in these segregated patient populations.

We've seen at ESMO, in the last 4 days, results from palbociclib patients, and overall survival outcomes in patients treated with hormone receptive positive breast cancer with palbociclib.

We've also seen the way in which disease can be stratified in the context of advanced prostate cancer into all oligometastatic disease and potential benefits from local radiotherapy.

And last but not least, we've seen results from the alpha specific PI3-kinase inhibitor in advanced breast cancer.

The future I think is very bright. Our drugs are getting better all the time and our understanding of the molecular basis of the disease is improving all the time - that will enable us to use these drugs more effectively. Thank you.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: