Abstract
This article is based on the Ellison Cliffe Lecture delivered at the Royal Society of Medicine on 17th October, 2017.
At the Francis Crick Institute, outside St Pancras Station in London, we are trying to understand how cancer evolution impacts drug response and treatment failure. The cost of cancer drugs since 1965 has increased from US$100 per month to US$10,000 per month. Unfortunately, many of these novel anti-cancer drugs are not working any better than they were 40 or 50 years ago. If we take the last 12 years of Food and Drug Administration approvals between 2002 and 2014, of the 71 anti-cancer drugs that were approved, 23 of these drugs had no overall survival benefit.
Overall, the median overall survival improvement was 2.1 months. Similarly, of European approvals, 48 drugs were approved across 68 indications in a four-year period from 2009 to 2013. The median improvement in overall survival was only 2.7 months. In 41 of those 68 indications, there was no improvement in survival at all. I would argue that the cost is not matched by benefit, a point that needs to be addressed urgently. So the question is why is cancer the ‘Emperor of All Maladies’ and why are costs so excessive? To understand cancer, we really need to understand the human genome and its dyanmic complexity through space and time in cancer.
The human genome is made of the Watson and Crick base pairs of which there are three billion in every cell in the body. Imagine every time a cell divides, it replicates DNA extremely faithfully, each one of those three billion base pairs is replicated almost identically. Overall, only 1% of the genome is coding and the rest of it is non-coding DNA. There are 22,000 genes in the genome.
What happens when this process goes awry? Mutations in the DNA happen in cancer. Most of these mutations will be passenger mutations of no consequence. However, occasionally and very rarely a mutation may occur in a gene that imparts added fitness and alters the biology of the emerging cancer cell. It may make the cancer cell proliferate faster. It may make it die less. These are called driver mutations, a concept I will refer to many times during my talk.
So then the question is how does one identify a driving mutation in cancer when you’ve got three billion base pairs to look at. Well the answer is it’s down to statistics. We apply a significance threshold to the genes across the genome where we see mutations. So if we take, for example, two well-known genes: epidermal growth factor receptor in lung cancer and BRAF in melanoma. These genes have recurrent mutations, often at the same position, which render the protein hyperactive, driving proliferation and cell survival in the context of melanoma and lung cancer.
We can see that because these mutations at these positions within DNA and each gene pop up above a certain significant threshold, more frequently than we would expect by chance. We refer to these as driving mutations, and will influence the behaviour of the emerging cell.
We have known for many years that cancers evolve from a single cell. Now in terms of how cancers evolve from that single cell, this commonly viewed picture of linear cancer evolution is seen in the literature depicting a sequential series of mutations that have to occur in a relatively ordered manner.
This linear model from a single cell depicts sequential mutations and their orderly transition of one cell population to the next such that every cell in this system is essentially the same. Each mutation gives accelerated growth potential that now dominates the tumour following a selective sweep. In essence, the cells in this tumour mass should be largely similar at the genetic level. So the question in my mind has been, why is curing cancer so difficult? Why are cures so evasive if linear evolution is the norm rather than the exception. So we decided to investigate this process in more detail. Back in 2010, we were lucky enough to get a grant from the European Union (PREDICT) to come up with ways in which we could predict response to drugs like Avastin. These are drugs that block blood vessel development and prevent tumour oxygenation. And the idea was that if we could identify resistant tumours, we might be able to sequence the genes in the genome and find genes that promote resistance to the drug, ultimately with the aim of identifying patient groups who would specifically benefit from this class of drug.
We wondered if the linear model of evolution was an oversimplification of the complexities of tumour evolution and perhaps there was more diversity within a tumour than we had previously thought. We decided to analyse multiple spatially separated biopsies from large tumours and sequence the tumour DNA at the genetic level to attempt to understand how these tumours have evolved from a single tumour cell. We first sequenced about 15 different tumour regions, analysing every gene in the human genome. In 2010, sequencing 22,000 genes was quite an expensive task, only ten years after the human genome had been sequenced. With modern-day technologies, we can now sequence genomes very quickly. It took years to sequence the first human genome in the late 1990s; now modern technology allows us to sequence a human genome in a matter of hours at a fraction of the cost.
We found that about two-thirds of mutations were not found in every tumour region and this tumour was not evolving in a linear way, but evolving in a branched manner analogous to Mel Greaves’ recent findings in leukemia and Peter Nowell’s seminal hypotheses in the 1970s. What that means in the case of renal cancer is you have early mutations such as VHL mutations in the trunk of the tumour’s evolutionary tree present in every tumour cell and then the majority of the mutations are found in the branches; that is, they are found in some cells but not others. Such diversity is the real challenge that we're going to have to tackle to really combat cancer effectively. If we rewind another 150 years and go back to what Charles Darwin would have said about this it is all entirely predictable from 20th and 19th century evolutionary biology. Indeed, Charles Darwin on the origin of species in 1859 wrote natural selection is daily and hourly scrutinising, throughout the world, every variation, even the slightest; rejecting that which is bad, preserving and adding up all that is good; silently and insensibly working, wherever and whenever opportunity offers, at the improvement of each organic being in relation to its organic and inorganic conditions of life.
We have spent much of the last seven years trying to understand the causes and consequences of branching evolution and whether it is really the norm in tumours rather than the exception. We see branched evolution to varying extents in the majority of patients. We rarely see linear evolution in renal or lung cancers. When we do see linear evolution, it is probably because we haven’t sequenced deeply enough and cannot go back in time to decipher how the tumour evolved in the past. What we are seeing is those driver events I mentioned earlier can be missed by a single biopsy. This means if you take a sample from a very large tumour, we may not always be able to portray the genomic landscape of the tumour as a whole. A single biopsy may miss important genetic events that impact upon fitness of an emerging cancer cell population within a patient, raising concerns about efficient implementation of a personalised or precision medicine strategy.
When one considers an end-stage cancer with up to a trillion cancer cells and the fact that each cancer cell may be subtly or grossly distinct from the one next to it, the presence of a drug resistance mutation against a drug targeting a driver event is almost a fait accompli, which begins to shed light on the inevitability of drug resistance in the clinical setting.
Therefore, this sinister ‘magic’ witnessed in the clinic of the inevitability of drug resistance can be explained mathematically due to the combined forces of tumour diversity and natural selection. It is this diversity that we need to battle in the clinical setting. So now I just want to diverge very briefly and tell you all why surgeons are so vitally important.
My two children have these two surgeons to thank for having two grandfathers, Meirion Thomas and Timothy Christmas, who operated on my father and my wife's father respectively. Tim Christmas tragically died shortly before our NEJM paper was published. Our paper in the New England Journal was dedicated to him. He was a huge advocate for the study and really supported us to achieve more. We really have these two surgeons to thank for so much. On the basis of what we understand about tumour heterogeneity, it is plainly obvious why surgery is so essential because it really is the only way you can remove that cellular diversity from the patient immediately. It is very unusual for radiotherapy or drug therapy to have such an immediate and long-lasting effect.
Julian Huxley in 1958 suggested it would be very interesting to study tumour “genetical inhomogeneity”. He contended it would be very interesting to discover the extent of such new variation in cancers and the rate at which it occurs. Tackling this almost insurmountable problem may not be entirely hopeless. In some ways, this may be like a very complex game of chess against a grandmaster in three dimensions, the most complex game of chess one can possibly conceive. But, like every game, there’s a rule book. We just need to understand it more deeply. Let me illustrate that rule book. This is the first patient with renal cancer I’ve previously shown you. This patient’s tumour had three different mutations in the same gene, SETD2, in three different regions of the patient’s tumour. We thought it was so unusual to see three distinct mutations in the same gene in three different cell populations.
What this tells you is this tumour had to inactivate this gene SETD2. There must be a huge selective pressure to inactivate this gene multiple times in evolution. So the question we have been asking ourselves is if we understood more about evolution early on in the disease, could we predict the tumour’s next evolutionary move and offer a patient a drug in advance of that mutation occurring? A prophylactic therapy perhaps to block branched evolution in its tracks?
The extent of natural selection in cancer is just uncharted. Our follow-up work has shown parallel evolution upon the same gene or same pathway or the same protein complexes that are currently in activated time and time and time again during the evolution of a patient’s tumour. And that provides a real vulnerability to treat that patient more effectively if we just had drugs available to do so. Charles Darwin noted this very phenomenon.
He said ‘the common rule throughout nature is infinite diversity of structure for gaining the same end’.
So this has led us to think we need to be much much more ambitious in our studies of nature’s infinite diversity instead of looking at one or two tumours. We need to look at hundreds to be able to really understand the evolution rule books by which tumours operate. And so Mariam Jamal-Hanjani and Samra Turajlic in my laboratory have set up two studies, a lung cancer study and a kidney cancer study called TRACERx lung and renal.
The idea is we analyse tumours from diagnosis through the disease course to understand the spatial and temporal evolution of the disease longitudinally and spatially instead of thinking about tumours as a static single entity. These are some of the questions we’re trying to answer. What are the relationships between diversity and the cancer and outcome that is are more diverse tumours with more complex phylogenetic tree shapes associated with worse outcome? Are there rule books and can evolution be predicted? Can we time events during evolution that will help to treat patients better to predict the next evolutionary move? Most importantly, can we identify where the lethal cancer subclone comes from that seeds metastatic disease? Can we analyse cancer evolution from blood? Now, we can actually monitor cancer revolution and its branching Darwinian evolution in real time, which I think is going to change the landscape of drug development and how we manage cancers. And I want to allude to some of the mechanisms that stimulate cell-to-cell variation in tumours.
The first question we were interested in asking, we took the first 100 patients from our lung cancer TRACERx study and we want to now ask what’s generating branched evolution. But to do that we need to separate out the mutations that are occurring early on in evolution in the trunk from mutations in the branches. That is, mutations that are present in some tumour cells but not others. Rachel Rosenthal, Nicolai Birkbak, Mariam Jamal Hanjani, Tom Watkins, Gareth Wilson and Nicky Mcgranahan, very talented scientists, took on this work. They took all the mutations and found that the trunk mutations were almost all smoking mutations and it turns out that there’s a direct correlation between the number of cigarettes smoked and the number of C>A mutations present in that patients lung cancer genome.
However, the ageing signature and an APOBEC mutagenic signature are stimulating cell-to-cell variation in the branches. APOBEC is a cytidine deaminase expressed in our cells, normally to combat viral infection. However, sometimes it is induced in cancer evolution. So even in a current smoker of 40 cigarettes a day, APOBEC and the mitotic clock signatures would be the dominant mutational process in the cancer genome.
Now I want to stop talking about point mutations. I want to now talk about larger scale changes in the genome brought about by chromosomal rearrangements. Richard Goldschmidt was a spectacularly interesting individual who published a book called The Material Basis of Evolution in 1962. He inferred that the generation of new species was brought about by large-scale changes in the genome through chromosomal rearrangements. We know this today in cancer as numerical and structural chromosomal instability or CIN as you’ll hear me refer to.
I want to illustrate this by showing you a paper from Carbone et al. published in 2013 investigating this beautiful ape called the gibbon, which I’m sure you all very familiar with.
Chimps, gorillas and orangutans have 24 pairs of chromosomes. Gibbons diverged from them about four million years ago during a time of very profound sea level change, presumably leading to species isolation, fertile ground for speciation. Now the divergence in the gibbon from the other genera appeared to coincide with chromosome disorder. The gibbon has somewhere between 19 and 26 pairs of chromosomes. What is fascinating about the gibbon is that they’re like us. They walk upright and they don’t have tails. They are monogamous and operatic sopranos and they swing at speeds from trees at up to 55 kilometres an hour. And they can do this because they’ve got a ball and socket joint for a wrist. The differences between the gibbon and other genera has puzzled evolutionary biologists.
It turns out that what’s happened during evolution is that there have been these viral insertion elements that have inserted across the genome >30 times and have inserted randomly and ended up nestling close to genes involved in faithful chromosome segregation that may have been responsible for the chromosomal differences and the profound distinctions between the gibbon and its ancestors.
Similar processes of chromosomal instability occur in cancer. A normal cell has 22 pairs of normal chromosomes and then two sex chromosomes. Spectral karyotyping, where each chromosomes labelled in a different colour, illustrates this very well. A cancer cell on the other hand is markedly different and jumbled up. We call this structural chromosome instability where two pieces of different chromosomes are fused. Here you have more than two pairs of the same chromosome, a process called numerical chromosomal instability.
Now this is important because every time you lose or gain a chromosome you lose or gain 1000 genes. So imagine a cell dividing in a tumour suddenly gaining or losing a chromosome. This is something you can see under a microscope happen frequently in live cancer cells; pancreatic cells can lose multiple chromosomes in one cell division. Every time this occurs, the daughter cells now have 2000 or 3000 more genes that differ in number, and sometimes structure, from their ancestors, resulting in profound and diverse behavioural changes of these cells. This process Goldschmidt termed ‘hopeful monsters’. Once in a blue moon, the diversity of chromosomal alterations manifest in a daughter cell might result in the generation of a cell endowed with profound new features and behaviours that might allow its future progeny to colonise new niches faster and more efficiently.
It turns out this is not the only way in which chromosomes are jumbled up in cancer. We can see whole chromosome doublings, or chromothripsis, where chromosomes fragment into very small pieces and realign in a disordered fashion putting genes next to each other that have never been next to each other before in evolution creating new functions.
Genome doubling is very common in cancer. Sixty percent of lung adenocarcinomas have doubled their genome. Us and others have shown that the process of genome doubling results in much more rapid evolution and more aggressive tumours in the resulting genome doubled populations. We think that genome doublings open up the floodgates to allow further diversity to occur and allow the tumour to become much more aggressive.
Again, this is not new. Julian Huxley noticed in 1942 that in ecology ‘genome doubled forms are essentially more vigorous. A fact reflected in their distribution which is frequently wider than that of the diploid variety.’ So in context of speciation in nature, genome doubling provides added fitness and the ability to colonise drier more arid ecological niches for example.
So now we can apply that genome doubling knowledge to lung cancer revolution and ask when does genome doubling occur in the context of lung cancer cell evolution from diagnosis through to death. It turns out that genome doubling occurs in about three-quarters of cases of early lung cancer and it occurs in the trunk very early on in cancer evolution. And we think this primes the cell for the acquisition of accelerated diversity upon which natural selection can act.
Genome doubling allows us to time mutations. If you have a mutation that occurs before genome doubling, it appears in our sequencing experiments at twice the frequency. This now allows us to build that evolutionary rule book and start to time mutations in the context of branched evolution. We can separate relevant mutations early in the trunk before genome doubling from those that occur later in the trunk after genome doubling, from those that occur very late in the branches after genome doubling in some cells but not others.
So going back to Goldschmidt who talked about chromosome order and chaos and hopeful monsters, we are interested in trying to understand what difference chromosomal order in chaos might make to human cancer evolution. Earlier, I talked about parallel evolution in the context of convergent mutations in SETD2. We have a talented medical student on the MBPhD program at UCL, Tom Watkins, who is responsible for this work. He asked a very simple question. Cancer genes can amplify themselves, so they can increase their copies multiple times. However, we have never really known if that occurs once in evolution or multiple times in evolution in separate subclones within the same tumour. Tom has devised a unique way of being able to map the genes that you inherit from your mother and genes that you inherit from your father and track them through evolution in real time to be able to understand when and where you see an amplification of a cancer gene. He can now use this tool to investigate whether cancer gene amplification derive from your mother or your father’s chromosome and whether amplifications occur on multiple distinct occasions during the evolution of a tumour.
In other words, can we now start to see an oncogene that causes cancer like HER2 being amplified from your mother’s chromosome in one region and your father’s chromosome in another region of the same tumour, indicative of parallel evolution. And what we found is on multiple occasions in TRACERx patients, we see multiple examples where genes are amplified multiple times from the maternal and paternal chromosomes in different subclones of the tumour. ‘Nature will find a way’, to quote Jurassic Park. That’s exactly what’s happening in tumours, oncogenic amplifications are occurring multiple times in human tumour evolution within different subclones of the tumour.
In terms of the primary endpoint of TRACERx: is there a relationship between the shape of the tree depicting cancer diversity and clinical outcome? The first thing we looked at was point mutational diversity and outcome: we found no relationship. So the next question is what about Goldschmidt’s chromosome instability ‘hopeful monster’ hypothesis. And here, there’s a very strong relationship the tumours that are more chromosomally unstable and have more diverse and more chaotic chromosomes recur much faster than the tumours which are more chromosomally stable. So in the context of prognosis and outcome Goldschmidt’s macroevolutionary theory trumps Darwinian evolution. It seems that, macroevolution (the idea of large-scale genomic changes) is having the most profound effect on what is governing outcome in the context of early-stage lung cancer.
So the question is what is so special about chromosomal instability? I have told you that gaining or losing a chromosome can lead to gaining or losing 1000 genes at any one time. We think about this in the lab and it is becoming increasingly clear that the reason this is such an important cancer feature is because it increases the chances of a tumour cell developing the optimal fitness level required to invade, spread and colonise new metastatic sites. There will be a lot of tumour cell death and wastage on the way as chromosomal instability generates many more hopeless monsters than hopeful monsters. However, tumours can afford wastage during evolution across a mass with hundreds of billions of tumour cells, such that one tumour cell endowed with metastatic potential, through chromosomal instability may be generated rarely, but with devastating outcome – the ‘hopeful monster’. Think of it like running a casino and following card shuffling rigging the game such that you always end up with four aces.
The hopeful monster contains literally all of the chromosomes required to form the most robust tumour cell, helping basement membrane invasion, immune escape and survival at distant metastatic sites. There are so many steps involved that a parsimonious way in which this could happen is through multiple rearrangements of ‘the serial chemical constituents of the chromosome’ as Goldschmidt would put it, generating that hopeful monster – coming up with the four aces with monotonous frequency.
The difficulty is processes through which chromosomal instability are generated are complex and the resulting cells very hard to drug through traditional drug discovery strategies. So what are the potential solutions? I think I’ve shown you evidence that targeting single point mutations in the trunk is not going to win the Nixon war on cancer. We need to do something cleverer. We need to target multiple events in the trunk of the cancer's evolutionary tree. Achieving this is not a trivial task since every tumour has a unique phylogenetic tree, a unique complement of trunk mutations that are never witnessed in any other patient who has ever lived or will ever live. It’s almost impossible to have two identical tumours. So to get to the stage of personalised medicine targeting more than one truncal mutation – ideally more than three to limit the possibility of acquired drug resistance – we need a bespoke therapy for every patient. And that’s going to be very difficult. And that’s not going to happen through traditional pharmaceutical methods of developing cancer drugs in a ‘one size fits all’ approach. So we have turned to TRACERx and cancer biology and asked is the answer inside every patient? We have some tantalising evidence that it might well be:
In a lung cancer genome from a smoker, the number of coding mutations in a smoking genome is about five to ten times more than a non-smoking genome. There are about 700 coding mutations present in every cell in the trunk of the phylogenetic tree in a smoking lung cancer genome. But in a non-smoker there are only about 50 to 100 coding mutations in the trunk of the tumours.
This is important because every time the tumour acquires a new coding mutation, that mutation has a risk of being seen by an immune T cell as being foreign and attacked. So there may be this ongoing tension between T-cell-mediated immune control of tumour growth.
Let us explore this concept in more detail. A coding mutation in a protein arises in a cancer and that mutation is processed in what’s called the proteosome where it is cut up into little pieces or peptides and presented on an immune scaffold on the cell surface called HLA or MHC.
The HLA-mutant peptide is presented to the T cell receptor as an ‘eat-me’ signal. So we wondered whether the truncal mutations in a tumour present in every tumour cell were important for immune-mediated control of tumour growth and whether there might be T cells in a tumour that recognise such trunk mutations that could one day be expanded and given back to patients.
First, we had to ask one fundamental question – are these trunk mutations actually important at all? So we turned to clinical trial data. There has been much excitement over the last five years surrounding a class of drug called immune checkpoint inhibitors such as anti-PD1/PDL1 and anti-CTLA4 therapies in melanoma and lung cancer. Many of these drugs appear to be particularly active in tumours with a large mutational load driven by environmental carcinogens such as ultraviolet (UV) light in melanoma and tobacco use in lung cancer. Indeed anti-CTLA4/anti-PD1 combinations may be curing up to 40% patients with advanced metastatic melanoma, something that would have been considered impossible a decade ago. How do they work?
The cancer displays mutations on HLA as I have previously reviewed. The T cell wants to eat that cancer cell. So the forces of natural selection and the predator prey relationship results in the selection of tumour cells that have evaded T-cell-mediated immune attack by upregulating PDL1 or PDL2 which bind to PD1 on the T cell and switch the T cells off. They anaesthetise the T cells such that the tumour cells are no longer subject to immune-mediated attack. Anti-PD1 drugs bind PD1 and reawaken the T cell to enable them to recognise the tumour cell and destroy it.
So knowing that Nicky Mcgranahan and Rachel Rosenthal in the lab were able to turn to clinical trial data and we found that in both lung and melanoma, tumours with a large truncal burden of mutations did better on these drugs than patients with these mutations dispersed in some tumour cells but not others. In order to drive an optimal immune response, T cells need to be re-awakened that recognise mutations in all tumour cells, rather than in a subset of tumour cells, or tumour cells at one site of disease but not another.
So then we wondered what is generating these branched mutations in melanoma? The trunk mutations were all UV-driven as expected, given the known mutagenic consequences of UV light initiating melanoma growth. However, we found that the majority of these branched mutations in three patients with a particularly large mutagenic burden were driven by the alkylating agents such as temozolomide. Therefore, these drugs are inducing subclonal branch mutations present in some cells but not others, driving further cell-to-cell diversity –which I have already highlighted as a key substrate for natural selection. These drugs are generating branch devolution. In short, they’re generating diversity. So what that means is although patients may have a transient response to cytotoxic treatment, this comes at the expense of more branched mutations in surviving cells as the tumour progresses.
Therefore, now we know that these trunk antigens may play an important role, we want to target them by leveraging the immune response. To do this, we turned to our collaborator at UCL Cancer Institute, Dr Sergio Quezada. Using our mathematical approaches to deduce truncal mutations in lung cancer, and Sergio’s skills in identifying T cells, we have been able to find patients with T cells within the tumour that recognise cancer mutations or neo-antigens present in every tumour cells. In one case, we found one patient for instance who has six different T cells that recognise six different truncal mutations, which we hope will be more than sufficient to drive T-cell-mediated cell death of the tumour while limiting the chances of resistance occurring – since a tumour cell would have to find six ways to evade the onslaught of such T cell therapy.
However, we must also second guess a tumour’s next evolutionary move and prepare for it. Therefore, lastly we asked, how else might a tumour evade the onslaught of the immune system? Nicky Mcgranahan and Rachel Rosenthal in the lab again turned to TRACERx for the answer to this question. Prior evidence had highlighted that mutations in HLA/MHC might drive immune escape, and many investigators have previously shown that HLA expression declines in some tumours as they progress. Therefore, tumours might alter HLA expression or select for mutations in HLA in order to prevent them presenting mutant peptides/neo-antigens to the immune system.
Nicky and Rachel have now shown that loss of one or more HLA alleles is very common in early-stage lung cancer, occurring in up to 40% of patients. These tumours delete one or more of six HLA class I alleles present on the tumour cell, derived from either your maternal or paternal alleles.
When this happens it often happens in the branches, later in tumour evolution, in some cells but not others. We also see parallel evolution where the same HLA allele is lost on multiple occasions in the same tumour, but in different cell populations or subclones, illustrating the immense immune pressures that tumour must be under to lose the identical HLA alleles.
Furthermore, we see that the cancer subclones that have lost one or more HLA allele(s) have more mutations and longer branches than sister subclones in the same tumour that have not lost an HLA allele. Therefore, HLA loss can permit branched evolution because that tumour cell now can emerge in the body by stealth, invisible to the immune system.
I would like to summarise by saying that cancer evolution is a real challenge to medical oncology. When you look on websites at the number of clinical trials, almost none of them take into account evolution and the changing dynamic nature of this disease over space and time. And I think to really succeed we have to battle evolution. We have to understand evolution in every patient. We may have to commit to precision therapies that are unique for every patient. These will be expensive and they will only be viable from a healthy economic perspective if they cure or dramatically prolong survival times. Absolute and meaningful changes to the survival time are really the only thing that matters. Surrogate endpoints, such as progression-free survival, are irrelevant. We really should be developing therapies now that cure patients with metastatic disease. And I think we need to think about adapting trials in the face of cancer evolution. We need to think about the ecosystem and environment in which the cancer evolves and developing therapeutics that can alter that ecosystem in favour of immune control and the changing nature of the disease over a space and time.
I’m hopeful we can target the processes in a tumour cell that results in tumour diversity and branched evolution that I have discussed today to limit tumour adaptation and drug resistance. I’m hoping we can predict and limit the constraints inherent to tumour evolution and exploit these cancer rule books for patient benefit. And ultimately I think the solution is going to be from within the patient leveraging that beautiful immune system that’s evolved over four billion years of evolution. My view is an adaptive immune system is going to be a very tractable approach to target a diversifying foe such as a cancer cell.
On that note I’d just like to end by saying thank you to patients and their families. We couldn’t have done any of this without them. I would like to thank Cancer Research UK for funding us and everybody in my lab, both past and present, and my great collaborators, colleagues, surgeons and friends for making this such an exciting and rewarding journey and, of course, to you for inviting me here today to the Royal Society of Medicine to give this lecture and accept the Ellison-Cliffe prize.