 Okay, so with no further delay, I'm going to introduce Charles Swanton. Charles Swanton is a medical oncologist, but is also a leading researcher at Cancer Research UK, and he has been the one who triggered that heterogeneity issue, and we are really sampling the terms of kidney cancer because it puts RCC in the limelight of all cancer biology, and we are pleased to listen to you. To follow that up, kidney cancer is actually quite an extraordinary disease, and to study it has proven, I think, quite fruitful, from a simple perspective of this, is that studying kidney cancer revolution is a lot more simple than studying lung cancer revolution, because the disease is much more genomically stable, so that gives us an opportunity to really understand cancer biology in general, I think. So I'll try and illustrate some of the work that James Larkin, my colleagues in London, and various very supportive surgical colleagues at the Marston and Guy's Kingston Thomases have been helping with over the course of the last year or two, specifically. So the implications of the therapy and outcome, I think, are the following when we think about heterogeneity. So we see heterogeneity both between patients with the same histological subtype of disease, be that breast cancer, ER negative breast cancer, or clasile carcinoma of the kidney. We see heterogeneity within tumors, so-called intratumor heterogeneity that can be regionally separated. And we see heterogeneity at down at the single cell level, where we're beginning to appreciate that even within a tumor, every tumor cell may be subtly different and differed by distinct genetic events. And I think answering one of Lawrence's questions earlier, we have known about genetic heterogeneity for 30 or 40 years. It's just current technologies enabling us to explore and decipher it at much higher resolution than ever before. So our venture into this field started about four years ago, when we set up, set about trying to address or identify in the context of a Framework 7 consortium with the Institute of Gustaf Roussi and the Royal Marsden and ourselves at Council Church UK, biomarkers of response to antigenic therapies, as well as mentor inhibitors. And to do this, we had to sort of take the fundamental premise that if you take one biopsy of a tumor, that biopsy would be representative of the entire genomic landscape of the patient's tumor. So we did the simple experiment where you take 10 biopsies from this patient that's refracting me as well as this perinephric metastasis and chest wall metastasis and sequence every gene in the human genome, all 22,000 genes, at about 100 X coverage, which means we try to decipher somewhere the heterogeneity within an individual biopsy. And what we found is that only about a third of mutations shown in this heat map in red are present in every region of the tumor. And about two-thirds of the mutations present, these are coding mutations across the genome, are present in one region but not another, suggesting that majority of mutations in a single biopsy are not going to be present across every region of the tumor that we analyze in this disease. Now the question we're really keen on answering now is, is this a trend or is this the norm in kidney cancer, and more specifically, are we likely to see this effect in other tumor types as well? I'm afraid I don't have answers to those questions, but we've got some early insights into clear cell carcinoma biology, both sporadic and more recently, germ9-VHL associated, which I'll show towards the end. So the first observation is in the first 10 patients we've looked at is that we see branched evolution in every case. So what do I mean by branched evolution? What I mean is that these tumors essentially are heterogeneous. That is that we see regional separation of mutations that are present in one region but not another. So in this case here, we see clear separation of the metastatic sites from the primary regions, and they share a common clonal origin, in this case, like all sporadic tumors that we sequenced so far. A second hit in VHL, either the first hit being VHL mutation, the second hit being 3P loss of heterozygosity, which is where VHL is encoded. So we see branched evolution in the first 10 patients we've sequenced with clear cell renal carcinomas as sporadic disease. And the intriguing thing is that each patient has a different phylogenetic tree, a different tree shape. And I think that tree shape ultimately, it's going to be quite interesting to address what is the relationship between the shape of the tree and the patient's subsequent disease course. Because one of the hypotheses we have is that patients with more simple tumors like this with limited branched evolution and limited heterogeneity, limited diversity, may fare better and have a better disease course than, let's say, this patient with a bow-bowl tree-like tumor, which is branched almost from the moment it's seeded, as it were. And these very heterogeneous tumors, we wonder, may have a worse prognosis. And that's something we're planning to address in the context of a longitudinal study between guys, Tim O'Brien and the Marsden, David Nicoll and James Larkin. So I've shown you ever as for what we call, what Darwin called, microevolution, and we're just simply poaching his terms. And Darwin really argued that nature never makes major leaps in Latin, nature alone, facet sortum. And he said that profound change at a population level is the result of a very slow but continuous process. So in the case of genomic evolution in the cancer, we can think about that at the single gene level, where a single gene, single point mutation over time accumulates with many other genes to lead to profound change in clinical behavior, perhaps. And these gradual accumulation of the small mutations of the major driver of change, and that's been elaborated in neo-Darwinism theory over the last 40 or 50 years. But the challenge, I think, as we look ahead, is this one of cancer macroevolution? So what do I mean by macroevolution? Well, this chap, Goldschmidt, was a sort of pariah of evolutionary biology in the 60s and was ostracized from most of the major Ivy League universities, as well as Oxbridge in evolutionary theory. Because he proposed that speciation was driven by macroevolutionary changes, that is, changes in whole chromosomes or parts of chromosomes that lead to the origin of new species. And these rare events result in profound change. And he coined this phrase, which led him to be a sort of a laughing stock with the evolutionary community, called Hopeful Monsters. And he published this book called The Material Basis of Evolution. And unfortunately, because of this term, Hopeful Monsters, he was ostracized from the evolutionary community. And his very valuable thoughts in this book have been rather marginalized. So why do I think these thoughts are valuable? Well, the first thing is that in this book, in Figure 35, he draws this picture here of simple chromosomal rearrangements that he'd observed or was beginning to observe in nature. And we see these chromosomal arrangements time and time again, not just in kidney cancer biology, but in all tumor types we look at. And he postulated that this macroevolution must proceed by a different genetic method rather than single point mutations, but through the rearrangements of serial chemical constituents of the chromosome, essentially. So why does any of this matter? Well, it matters because we see chromosomal aberrations time and time again in pretty much all solid tumors. And these can be chromosomal instability, in terms of structural and numerical abnormalities, or rearrangements of those chromosomes by chromothripsis or chromoplexi, where you essentially get fragmentation of the chromosome and re-annealing either to the same chromosome or parts of different chromosomes, really creating sort of a mishmash, a jumble of chromosomal DNA that can essentially encode for new genes or perhaps amplified genes in ways which a normal eukaryotic cell is not designed to tolerate. And so I think this aspect of cancer macroevolution is something we need to think about in future studies of this disease, both kidney cancer and other tumor types. And I'll show you some brief evidence of that. So when we looked at our first patient, we looked at this phenomenon of chromosomal instability. Because what we saw at the metastatic sites were the mutations in the metastatic sites were essentially pretty much all the same. So in the chest wall metastasis and the perinephric metastasis, the mutations were almost identical. So this made us think that the patient was still alive at the time. We thought, well, perhaps we could identify a mutation that's present in all metastatic sites who might be able to find a new drug to offer this patient that might be effective and treat all of the metastatic disease. So we did a control experiment to look in a bit more detail whether macroevolution might now be fostering change in this tumor. And so what we did is we applied fax analysis, which essentially where we count the chromosomes in each biopsy to work out whether chromosomal instability was occurring in this tumor. And what we found is that region four of the tumor was most similar to the chest wall metastases and the perinephric metastasis. Most of the regions of the primary were diploid, like all of the cells in our body, but region four had doubled its genome and become tetraploid and had spawned the metastatic colonies that had now become chromosomally unstable. And so you can look at this at a more microscopic level or higher resolution level by applying SNP-CGH technologies to look at the ratio between the maternal and paternal allele across the genome. And what you see in the region four, this perfectly balanced tetraploid genome, double genome, when you look at the ratio of paternal to maternal chromosomes, chromosome 3P is showing a lealic imbalance. This is where VHL is encoded. So we've got loss of one allele on chromosome 3P and we have another small area of loss of heterozygosia in chromosome 16P. But other than that, the tumor genome is perfectly balanced. So the ratio of maternal to paternal alleles are identical. But now let's look at the chest wall metastases here, M2A and M2B. These are biopsies separated by only a centimeter that have exactly the same exome level mutations. Here you can see they have very different chromosomal copy number events across the genome. And what this indicated to us is that actually these tumor genomes are not the same at all. Despite having very similar exome mutations, they have very different gene dosages. And I think this diversity within a single metastasis is something that we need to worry about or at least think about as we move ahead with these types of studies. And we've known for many years that chromatominal instability, that's variation in whole chromosomes or parts of chromosomes, is a poor prognostic feature. Again, emphasizing Laurent's question earlier is heterogeneity a new phenomenon. It's not a new phenomenon. We're just simply being able to resolve it much better than ever before. So why is any of this relevant to clinical trial opportunities or drug development in general? Well, I think one aspect of this is if we think about tumors like trees, targeting the early events in the tumor, the so-called trunk events, that are present ubiquitously at every site of disease, might be a more tractable approach than targeting individual subclones with drivers present in one branch but not another. So in the context of clear cell carcinoma, the kidney, what are these truncal ubiquitous events? Well, they are, in fact, quite simply VHL. VHL mutation and 3P loss of heterozygosity are the core truncal events that are present in every sporadic tumor that we've looked at and that'll come as no surprise to this audience. So how can we identify these truncal events? This gets a little bit geeky but I think it is quite important because I think it really gets to grips with how we can use emerging technologies to better structure clinical trials and some of the caveats with new technologies when thinking about structuring clinical trials to identify those trunk events. So when we look at, we take a next generation sequencing approach what we're trying to do is we're sequencing each gene in turn and we do it multiple times. A number of times we capture a mutation tells us something about the prevalence of that mutation within that cancer that we're sequencing and that's called the variant allele frequency, the VAAF, and we can mirror that on a graph between naught and 100%. A heterozygos mutation in a perfect tumor with no stromal contamination will occur about 50% of reads. So we're about 50% of times we will identify that mutation and that's what we see with VHL and this tumor here across these multiple biopsies. It's occurring in about half of all the reads in keeping with a heterozygous event. Now if we then look at other mutations in this tumor, other known drivers of kidney cancer biology, P10, PBRM1, and ATM, these are also occurring between about 40% and 50% in this biopsy here shown in yellow. So we get the idea from this that perhaps all four of these genes are truncal because their presence uniformly at around 40% to 50% of reads that we're getting from the sequencing reaction and that's zoomed in a bit here. But then when we look at more biopsies, what we suddenly find is that these drivers that we thought were truncal that were present early on in tumor evolution are in fact not truncal at all because they're completely absent in these regions here, region 10, the venous thrombus, and region three. So these so-called clonally dominant lesions here are actually subclonal because they're not present here. So in fact, when you portray them on a tree, the only one that is in the trunk is VHL and these three here are all in separate branches of the tumor. So one biopsy in clear cell carcinoma, in this case, is misleading. We cannot infer clonal dominance from just looking at one biopsy. And the reason for that I think is that in clear cell carcinoma, we're dealing with such large tumors that have evolved potentially over quite a long time. There's been a lot of time for regional separation of these subclones, generating what we call or would adopt to the sort of evolutionary term allopatric separation of subclones. A little bit like allopatric speciation where you get new species developing on islands that are geographically isolated. Not to dissimilar to that is what we see in clear cell carcinoma, the kidney where you see independent subclones evolving within the same tumor separated by what we can anticipate might be conceived as clear tissue planes in this patient's tumor. So what next? Well, I think part of the problem for treatment failure and resistance to targeted therapies that we see so often in the clinical setting I think is likely to be occurring through this problem of subclonal driver events. So what do I mean by subclonal driver events? Well, simply put, these are driver events that are in the branches but not in the trunk. So these are driver events that one biopsy is not going to be able to identify. And the problem is at the moment we simply don't know how many of these driver events are actually operating in a typical patient's clear cell carcinoma. We really don't know for sure how many driver events are really implicated in the biology of a single patient. In this patient you can see here we've got VHL as the trunk event and then we've got three or four driver events that are present in some regions but not others. And so when we look at the first 10 patients, eight of those patients, eight of the 10 patients have more subclonal drivers that is drivers that are spatially separated in the patient's tumour than they have drivers that are present in every biopsy. I think this begins to shed some light on why resistance to therapy in advanced disease, it's targeted therapy at least, is so inevitable. So when we try to calculate the number of driver events, these are recurrent events that occur time and time again that we identify sequencing multiple clear cell carcinoma that can be well described by at least four groups internationally. When we try to define the number of driver events in a patient's tumour by a single biopsy, we think now we're underestimating the number of driver events quite considerably. So let's take P53 as an example, a common tumour suppressor that's commonly mutated. In kidney cancer it's estimated on a per biopsy basis, about one in 20 patients has a mutation of P53. When we look at this on a per patient basis, sequencing between five and 10 biopsies per patient, we can find evidence of a P53 mutation in up to 40% of patients. So this driver event is more common than a single biopsy would have us believe. So how on earth can we map the total genomic landscape of a tumour? Well, the simple fact of the matter is it's difficult. It's very difficult in clear cell carcinoma in the kidney, I think. And one way we're exploring at the moment is taking circadian-free DNA from patients' blood and then sequencing that to try to identify the number of drivers that might be operating. And we're having limited and early success in this area, identifying branch mutations that were only identifiable through spatial separation of biopsies. We can now identify many of these spatially separated mutations in a single blood draw by ultra-deep sequencing at 10,000 X coverage. Suggesting that in the future, we hope, this may be a way of resolving in a sort of super-tumour material, if you like, the diversity present in one tumour mass. So where can we go moving ahead? What new drug targets can we think about trying to exploit? What does a future hold in the next 10 years that Martin might talk about tomorrow? Well, I think one of the issues is this issue of convergence. It's quite phenomenal event in evolution that's been well-described before in ecology that we're beginning to see in cancer, in cancer medicine. Can we start thinking about tumours like a game of chess in three dimensions, in a way? So can we predict the next chess move of playing a game of chess against this grandmaster? Can we predict Kasparov's next move, 10 moves in advance, and do something about it to forestall Kasparov or Karpov for making that move? So what do I say this? Well, in the context of clear cell cast in the kidney, we're seeing a lot of convergence. So what do I mean by convergence? What I mean by convergence is, despite the heterogeneity, we see these tumours have to inactivate the same gene time and time again in different branches, different spatially separated branches of the tumour. So this is the first patient we looked at. This patient has three different set D2 mutations and three different KDM5C mutations in five different regions of this patient's tumour. So suggesting that wherever these subclones are evolving, the evolutionary pressures are similar, resulting in activation of the same genes. So if we knew that by understanding the trunk in a bit more detail and we had drugs that might forestall the tumour from making that move, could this be a tractable way of using evolution to the patient's advantage? And so we're seeing this time and time again. Here's another example. This is a chromatin remodelling complex called the SWY-SNF complex. What we see is a very complicated complex, protein complex formed by 10 or 11 proteins in a cell. And this tumour has inactivated the protein complex in three different ways by mutating one or more of the member of this particular complex, PBRM1, arid 1A or smarke 4. And we see this in the context of mTOR pathway activation, a common conversion to venting clear cell carcinoma. On a per biopsy basis, we see about one in five patients has an activating event of the mTOR pathway. But when we look at a per patient basis, we see that up to 60% of patients can have at least one or more aberrations in this pathway shown here, where we see distinct mutations in P10, a P10 mutation in one branch and an mTOR activating mutation in another, or two distinct PIC3CA mutations in different branches of the patient's tumour. And so overall, we see genetic convergence in six out of 10 of these patients, suggesting that perhaps there may be some future in predicting the next evolutionary move. So to elaborate on this, I think we need to think about targeting the trunk drivers that are present in every cell. So we need new drugs that can optimally target VHL loss of function. And we need to think about exploiting some of the convergence events in the tumour mass and predicting the next resistance move that tumour may take by selection of one branch over another. What new subclones may dominate the disease at resistance. But we're also not talking about somatic mutations. I mentioned macroevolution. We have this problem of copy number aberration, so-called SCNA heterogeneity. And the context of clear cell carcinoma, we see that the only copy number event that's known in clear cell carcinoma, which there are about eight to 10, that recurrently occur as either recurrently gained or recurrently lost, these are parts of the chromosome or chromosome arms that are currently gained or lost. The only one that's always present in every biopsy is 3P loss of heterozygosity, shown here where my green arrow is. All the others are present in some biopsies, but not others. And so when you then map the copy number aberrations with the somatic mutations onto these phylogenetic trees, you can see they get a lot more complex. And I must say that this is probably only the tip of the iceberg for each patient, because we still are only sequencing 10 biopsies per patient in these very large 250 centimeter cubed tumours. And that's not including most of the metastatic sites. I think if we did that too, these trees would get a lot more ornate and a lot more complex. But they would still share VHL as the core truncal driver. So some of this I think has some impact on our understanding of tumor biomarker developments and thinking about the next generation of biomarkers. To have a biomarker that's gonna tell us something useful, it's got to evade sampling bias if we're relying on a single biopsy in the context of clear cell carcinoma and the kidney. So we've applied more recently a poor prognostic and a good prognostic mRNA expression signature set. It's a so-called CCACCB signature to clear cell carcinoma that kidney these 10 tumours that I've shown you today. This is an expression profile that segregates out good from poor prognostic patients. And what we find is that in eight out of 10 of the tumours that different regions give us different outcomes. So we showed this initially in patient one back in 2012. And we've more recently applied this to the 10 patients. And in eight out of the 10 patients, we see divergence with some regions harboring a poor prognostic signature, some regions harboring a good prognostic signature. The fact is though that in our meta analysis that we're about to submit for review, this is still the best prognostic signature we've come across. So we've tested multiple prognostic signatures and this appears to be the best. So it does work in some cases, but potentially not quite all simply because of this issue of sampling bias that some subclones still remain in the tumour and harbour a good prognostic subclone compared to the majority of this tumour here which might be a poor prognostic subclone. So this is a complicated problem. Sampling bias is likely to impact and confound biomarker development. And there's something I think we need to think about when coming up with new biomarkers that better forecast outcome for patients. So one question we're really very interested in really getting to grips with is how many drivers are present in these tumours? How many drivers are present in a typical large clear cell carcinoma kidney with multiple metasatic sites? That's a coin of phrase from a collaborator of ours Andy Futrell who's now at MD Anderson who asked this question after our first patient was sequenced. He said, Charlie, the problem is we don't know how deep the rabbit hole is. So we're trying to now address exactly how deep the rabbit hole is by taking hundreds of barbs across these tumours to count the number of drivers that present in the tumour. And I don't think that's a stamp collecting exercise. So the simple reason is that traditionally we feel or we know from mouse models of cancer, et cetera, that four or five drivers are required to initiate tumour agenesis. But we don't know yet how many drivers commonly occur in a patient with advanced metasatic disease. And I think clear cell carcinoma in the kidney will lead the way in being able to tell us and answer that quite quickly. And we're doing this because when we plot the number of biopsies against the number of driver events, you can see in some of these tumours that we're seeing, we're seeing no tailing off of the curve. So we're just seeing and more biopsies we look at, the more driver events we find. And we're getting up to 20 driver events in this patient with a slight tailing off of the curve here. So we hope by looking at more biopsies we'll be able to understand how many drivers really are implicated, which might have some impact on targeted therapies and thinking about future drug development strategies. So to summarise two thirds of my talk, I'd just like to say about two thirds of the drivers are heterogeneous and spatially separated. We can't see these branch drivers from a single biopsy. So we need multiple biopsies, at least using this type of strategy to resolve the number of subclonal drivers that are readily distinguished in clear cell carcinoma kidney. Current sampling techniques underestimate the number of driver events in clear cell carcinoma of the kidney. And I think if you think about this sort of philosophically, if each tumour harbours a similar trunk event, so they've all got VHL loss and they've all got VHL mutations, but every patient's outcome is different, then perhaps that's telling us something about the branch evolutionary events that are occurring in the tumour and the potential impact of those branched events on patient outcome. So we probably do need to understand this diversity to really be able to forecast drug resistance potentially and patient outcome. So I'll just finish off the last two or three minutes of the talk by showing some evidence of recent data that's really surprising us and shining a bit of a light on some of this problem of heterogeneity. And dare I say it, flip the coin to show you some evidence of homogeneity in one isolated tumour that is surprising us that might shed some light on the differences between germline VHL syndrome patients with clear cell carcinoma in the kidney and sporadic clear cell carcinoma in the kidney that also shines a light on some interesting questions in evolutionary biology. So we started off on this experiment to ask a very simple question. If you have one patient with a germline mutation in VHL who has multiple clear cell carcinomas in the kidney, how similar or how different are they? Given that the patient is suffering from clear cell carcinoma in the kidney so that the microenvironment is the same, the genetic background that tumour is growing in is essentially the same. So how similar are different tumours from the same kidney or between kidneys from a single patient? And these arguments go back 30 or 40 years in evolutionary biology to Stephen Jay Gould who argued for the radical contingency of the human species. So he argued if you wind back the tape of life the early days of the Burgess Shale and let it play again from an identical starting point, the chances of us all ending up in this room as human beings with brains capable of understanding genetics, et cetera, and everything else would be vanishingly small and that there are many routes through evolution in general. But then Conway Morrison others argued that actually there are many constraints to evolution that would actually mean that if you did the same experiment that Stephen Jay Gould argued, in fact he would argue get the converse and it's quite likely that you might end up with the human species all over again. Now of course we can't address that within the context of cancer biology but we can address some questions about evolution in general by looking at clear cell carcinomas that are growing within the same patients at the same time within the context of a patient with VHL syndrome. So this patient who's from originally seen at Sharon Cross and then by David Nickel at the master 26 year old male with a germline VHL mutation who had a right radical nephrectomy turned out to have two tumours. Although there were one consistent mass by genetics we could decipher these were two separate tumours. I'll show you the data for that in a minute. And then about a year later the patient had a left partial nephrectomy for a further two tumours on the left side. And the results are shown here. So here we've got the four tumours shown along here, tumour one, two, three and four and where you see blue that's a point mutation and where you see gray there's no mutation. And so the first thing you can see from here is that these tumours are relatively simple and that's not because of tumour stage. We've got mixed stage here, we've got two stage three tumours and two stage one tumours. And bear in mind many of the clear cell sporadic tumours we were looking at were stage three and yet we're seeing very limited evidence of diversity here and quite simple tumours in this young patient. With only 13 or 14 mutations about a tenth of the number that I showed you for patient one earlier. And here none of these tumours are showing branch evolution. They're all showing linear evolution as far as we can tell. The second hit here is 3P loss of heterozygosity. So every tumour of course has the same germline event same VHL mutation. But the second hit here is occurring at three different sites in 3P chromosome 3P. So this is chromosome three along here and where I've got the red circles in these four different tumours you can see the four different sites where the break points occurring resulting in loss of the 3P chromosome that is loss of the other VHL allele. So we think the second hit is 3P loss of heterozygosity that's distinct in each one of these tumours. And then when we look at the drivers what we see is that each tumour harbours a distinct subset of drivers. So the right kidney tumour two has an IDH2 mutation. The right kidney tumour one has an ARID1A mutation. The left kidney has four different mutations that are distinct from tumour four. Although tumour four also has an M-tour mutation that is distinct but likely to be an activating event similar to the left kidney M-tour mutation which is distinct showing convergence. So here I think we've got some evidence for contingency and convergence. Convergence in that you're seeing two distinct M-tour mutations in the same kidney but also contingency in that every tumour is different despite the fact it starts off from the same germline mutation and the same microenvironment. So in other words, I think Gould and Conway Morris are both right in this context. So in the context of the M-tour mutation what we see is that there are two different M-tour mutations occurring, 3N1652 and Lucene2427 that predicted both to be activating events. And when you look at when we did phosphoproteomic studies with my colleague Fabrice Andre at Institute Gustave Roussi what Fabrice managed to show is that this M-tour activating event is indeed likely to be an activating event in that we see hyperactivation of phospho-AKT, a target of M-tour in the tumours on the right side indicating that this M-tour mutation that's distinct between tumours three and four has the same biological endpoints targeting the phospho-AKT 3N308 residue. And this leads me on to the last 30 seconds of the talk to say where we're heading next. So James Larkin, myself, Tim O'Brien and David Nicol LeMarsen are setting up a longitudinal study in clear cell castenome of the kidney to address some fundamental principles of tumour evolution. We hope to ultimately be able to optimise some of the clinical trial aspects, identify the branch drivers on a patient-by-patient basis, exploit some of the constraints to cancer evolution over the next 10 years we hope and begin to define some of the relationships between diversity, clinical stage and cancer outcome. And in time we hope to match into these longitudinal studies autopsy programmes so we can begin to define the origins of the lethal subclone of multiple metastatic sites of disease and identify some of the impacts of therapy on cancer evolution as well as develop non-invasive techniques to monitor cancer evolution non-invasively through blood tests. So in summary, I think we need to address the two fundamental principles of Darwinian evolution. We need to identify the mechanisms driving branch evolution. We need to understand why some tumours are branched and some tumours aren't, why the VHL syndrome tumours show linear evolution. We don't understand this. I think it's the patients that are a lot younger. They're diagnosed earlier through screening programmes perhaps. And we need to improve our cancer selection pressures. We need to optimise drug development against VHL loss of function and resolve a subclonal dynamics as well as understand the evolutionary pressures through these longitudinal studies. So thank you very much for listening. If there are any questions, I'd try to ask them. I should just thank very quickly the people who've made this work possible. James, my colleague and friend in the Marsden, as well as the surgeons who've made this work possible, as well as my colleagues in Sikusa for UC and all the people in the lab, both here, both at CRUK and University College London, who've helped with this work. Thank you very much. Thank you, Tim. Hello, Charlie, thank you. That was a fantastic talk. Really enjoyed it. I've got two brief questions. The first is, on the mechanism of heterogeneity, why it's so very impressive, impressively sort of tree-like, the data that you told me about before in colorectal cancer, I think, with replication for cannibalism, is that relevant to renal cancer? That's the first question. And the second question is, I'm not sure you're modelling is the right term, but you're considering these subclones as independent of each other. Do you have any data to say whether they interact, perhaps to compete for resources, for example? So that targeting one clone may actually release a second clone? Yeah, two great questions, Tim. Answer your first question. We don't have an answer to either of the questions, but I'll speculate wildly, if you don't mind. The first question, we strongly suspect that some of these early branch drivers are actually initiators of heterogeneity themselves. So we're spending a lot of time exploiting the function behind these inactivating events of these common drivers to see if they themselves can drive branch evolution. The second question is there clonal competition or clonal synergy? I think inevitably there must be. And in any population there is synergy and antagonism and competitive release. I think it's something that we probably see quite frequently in clinical trials where we target, say, drug-sensitive subclone and the resistant one springs up to compete within the same environment. So this is something we're trying to model, but to model it, we need better animal models actually. So the sort of starting point for this is to develop an animal model of VHL, clear cell carcinoma, the kidney number one, and then to try to model some of these branch driver events on top of that to try to initiate diversity within these tumor models to start to study those critical questions that have quite, I think, major implications, I suspect, for the way in which we think about targeted therapies, targeting drug-sensitive subclones to lead to the release of resistant ones that we can't do anything about. Yeah, thank you very much. I have a question concerning the source of your biopsies. We have learned during the last years that the sources of a liquid biopsy may be a much better one to identify driver mutations from passenger mutations. Do you think also, too, is this a liquid biopsy as a source for mutations? Yeah, absolutely. I think, you know, circling 3D and A and potentially CTC analysis is gonna be a very good source to identify driver mutations. Whether it's the best, we don't know. And one of the purposes of Tracer-X is to address that question so we can compare the heterogeneity of an nephrectomy specimen with what we're seeing in CFDNA. Does it really adequately reflect the diverse landscape that we see in a primary metastatic tumor? Nobody's really done that experiment methodically enough to be able to come up with a solution. The emerging evidence is that it is pretty good. How good, we don't know. Well, thank you.