“Cancer starts when a single cell in our body starts dividing out of control.”
We repeat this statement so often it would be banal, were it not for its implications.
But after a single, initial, malignant cell division, what happens to the two resulting ‘daughter’ cells? Are they identical to their parents? And what about their daughters? And their daughters’ daughters?
And what about the offspring that split off and travel around the body? After all, it’s usually cancer’s tendency to spread that makes it dangerous, rather than the initial tumour itself. But why is it that advanced cancer is so hard to treat?
Today, a team of some of the UK’s most exciting young researchers, funded by Cancer Research UK, University College London Cancer Institute, the Medical Research Council and the Wellcome Trust, has published results of a three-year analysis of kidney cancer samples.
Sequencing billions of ‘letters’ of DNA, the researchers looked in unprecedented detail at the relatedness of different regions of patients’ tumours, and between the patients’ primary tumour and the more distant secondaries, or ‘metastases’.
Their findings are stark: whichever way they looked at the data, no two samples from the same patient were genetically identical – not even samples next to each other in the original tumour. And the secondaries were significantly different from their parent tumour.
Their findings are the most compelling evidence yet that, like populations of animals in an ecosystem, tumours adapt as they grow, obeying the fundamental evolutionary laws laid down by Charles Darwin over a century ago. It seems this evolutionary aptitude may foster their ability to spread, and to become resistant to almost every treatment we can throw at them.
This heterogeneous nature of cancer has big implications for the way we think about the disease, and for how we continue to improve the way we treat it.
A patchwork quilt
Just as no two people are identical, we’ve long known that no two patients’ cancers are the same.
But in recent years, evidence has emerged that there are regions within the same tumour that have different genetic faults, and that behave differently. Scientists call this phenomenon ‘intratumour heterogeneity’.
To take just one jigsaw-piece of evidence, researchers have found that elevated levels of the Her2 protein – which makes breast cancers susceptible to the drug trastuzumab (better known as Herceptin) – aren’t always found in all regions of a tumour.
Likewise, a study last month showed that distant secondary tumours in children with brain cancer are genetically different from the original growth.
It’s a phenomenon that’s been puzzling scientists ever since the 1870s, when Theodor Boveri looked down a microscope and saw that cells from the same tumour sample had different-sized nuclei.
But these observations have raised several inconvenient questions, especially given recent progress in understanding the mutations that drive cancer, and the emergence of drugs that target these faults.
Does heterogeneity matter in practice? Does it occur in every type of cancer? Just how heterogeneous is the average tumour? And, most importantly of all, what effect does this patchwork nature have on treatment – especially targeted treatment?
Professor Charlie Swanton works at Cancer Research UK’s London Research Institute, and has been pondering the issue of intratumour heterogeneity for some time. He’s convinced it’s an “elephant in the room”, and it may help to explain three separate but related facts about cancer.
Firstly, cancer is very difficult to cure after it has spread. This is despite years of progress in chemotherapy and radiotherapy, two techniques that can offer respite to people with advanced cancer. And it’s also despite the emergence of targeted treatments like trastuzumab or vemurafenib.
Secondly, most advanced cancers eventually become resistant to every type of drug used to treat them – both ‘traditional’ chemo and these newer agents. This is quite extraordinary: tumours can work out how to cope with chemicals that they’ve never ‘seen’ before – a biological superpower far beyond that of infectious diseases. Just consider how it’s taken ‘multidrug resistant’ bacteria like MRSA decades to evolve. Yet cancers can do this in a matter of months or even weeks. How?
And finally, researchers haven’t yet managed to develop tests to predict how a patient’s disease will progress, nor monitor their progress (a field called ‘biomarker’ research) – this is despite years of research, and a lot of tantalising pilot studies. Sometimes researchers detect a promising ‘signal’ by looking at samples from a handful of patients, only for this to disappear in larger numbers of people.
Professor Swanton and his colleagues had an inkling that these three phenomena could be explained if tumours were, in reality, highly variegated, rather than a homogenous clump of similar cells.
Using the very latest technology, they set out to measure cancer’s ‘variegatedness’ in more detail than anyone has ever previously attempted.
Professor Swanton’s team – a talented group of upcoming scientists – analysed the DNA from a series of tissue samples donated by a patient taking part in a clinical trial at the Royal Marsden Hospital. To do this, they used ‘next-generation’ DNA sequencing – a relatively new technique that’s only become widespread in the last few years.
(As we’ve blogged about before, next-gen sequencing allows researchers to look at the entire DNA sequence in a tissue sample in one go, rather than focusing on a few select regions. In terms of effort, it’s the difference between copying out the Encyclopaedia Britannica by hand, and photocopying it.)
The computers that power LRI’s next-generation DNA sequencer form a huge black monolith that now takes up half of the building’s 7th floor. Swanton and his team used it to analyse the entire genomes of seven individual samples taken from a 10cm wide tumour that had been removed from a patient’s kidney. This was a Herculean effort that saw the machine running flat-out for four months solid.
Each sample was ‘read’ tens of millions of times, and each time the data was compared to the patient’s ‘normal’ DNA to spot the telltale mutations.
They also sequenced the genomes of three other samples from sites the cancer had spread to: one in the fatty tissue around the kidney, and two more from their chest wall.
All in all, this amounted to about 140 thousand million ‘letters’ of DNA, taking up 5.7 Terabytes of computer storage.
The results suggested that there were over a hundred separate mutations across all the samples.
But only about a third of these were present in every single sample. Each sample, even from next-door areas of the original kidney tumour, was unique.
In other words, there were more genetic differences between biopsies from the same tumour than there were similarities. Clearly, at least in kidney cancer, tumours contain an extreme amount of heterogeneity.
Just to be sure, one of Professor Swanton’s team, Dr Andrew Rowan, then painstakingly verified many of these mutations using ‘traditional’ Sanger sequencing – and the results held up. The graphic below illustrates their findings.
Fearsome family trees
But how were all these samples related? Did they arise separately in a linear fashion, all closely related to each other? Or were they – as the researchers suspected – separate parts of the tumour that had evolved independently from a common ancestor?
Swanton’s team fed their mountain of data into sophisticated computer programmes run by talented mathematicians at the Institute, Stuart Horswell and Aengus Stewart, who together with Marco Gerlinger (lead author on the paper), pieced together the tumour’s history.
As they anticipated, the tumour’s ‘family tree’ turned out to be branched, just like Darwin’s ‘tree of life’ itself:
They even found strong evidence that one region from the primary tumour, dubbed ‘R4’, was the ancestor of the three secondary cancers.
This is the first time a tumour’s evolutionary history has been reconstructed in such meticulous detail, and at such depth.
But the researchers didn’t stop there.
Looking from every angle
Although next-generation sequencing is a powerful tool, it doesn’t tell us everything about what’s going on in a cell – not by a long chalk. So the researchers deployed a variety of other techniques to analyse the samples, and to see if things differed in other ways.
First, they looked at whole chromosomes inside cells from each sample – and found these were deranged to different degrees in different samples. This revealed a whole new level of genetic heterogeneity in the cancer.
Next, they compared mutations in specific genes. One mutation – in a known ‘kidney cancer gene’ called VHL – was mutated in exactly the same way across all samples.
But others were mutated differently. Gerlinger’s eagle eye spotted that a gene called SETD2 was mutated in all samples, but the exact nature of this mutation was different, suggesting that this mutation occurred at different times in each region’s development. This is a phenomenon evolutionary biologists call ‘convergent evolution’ (think ‘bat wing’ vs. ‘bird wing’).
They then looked at overall patterns of genetic activity that had previously been linked to patients’ prognosis in kidney cancer.
Most of the tumour samples seemed to show the ‘bad prognosis’ pattern. But the R4 region, and the secondaries, gave off a ‘good prognosis’ signal. This hints at why so many attempts to predict patients’ fates using such tests have faltered – the signal depends where you look in the tumour.
Finally, in work led by another member of the team, Dr Eva Gronroos, they looked at the activity of certain proteins in the different samples. Again, they found substantial differences, most notably in the activity of a protein called mTOR, which affects the response to the cancer drug everolimus. This was a key piece of evidence, and showed that the genetic patchiness had actual real-world implications.
And to cap off their laboratory tour-de-force, they collaborated with the Wellcome Trust Sanger Institute to repeat these experiments, to a greater or lesser degree, in a second patient’s samples, whilst the LRI team examined 30 more samples from four tumours, adding months to the project, but – crucially – meaning they could be extra confident in their results.
Once again, they found that each patient’s tumour was made up of separate regions, and that the metastases were genetically different from the primary tumour.
So what does intratumour heterogeneity mean for future research? Swanton and his colleagues are sanguine about the implications.
“We have a lot more work to do” to understand how all this fits together, he says. Chiefly, he wants to know whether this degree of heterogeneity is found in other types of cancer – he suspects it is.
Dr James Larkin – the Royal Marsden clinician whose patients donated samples for the study – agrees that the study has implications for how we think about personalised medicine. The difference between the primary tumour and the tumour at distant sites can’t be ignored.
“This may be relevant to how we treat kidney cancer with drugs, because the molecular changes that drive the growth of the cancer once it has spread may be different from those that drive the growth of the primary tumour,” he said.
We’re on the right track
But there’s also a positive side to all of this. We’re now a step closer to appreciating the diversity and complexity at every level inside tumours that may lead to drug resistance. And this knowledge will help researchers find ways to stop it from happening in the first place.
Swanton believes that the key is to prevent cancer from becoming variegated in the first place. One way could be to develop new drugs to block the processes that allow the cells in a tumour to become so diverse and messy.
“The next step will be to understand what’s driving this diversity in different cancers, and identify key driver mutations that are common throughout all parts of a tumour,” he told us. But this means years more slog to identify these mutations and prove that they are indeed ‘drivers’ – even in cancers that have spread.
Thankfully there’s also a simpler way.
Genetic diversity develops over time. So aiming to diagnose and treat cancer earlier, when its less diverse, will likely improve things for patients. Take ovarian cancer for instance: 95 per cent of women diagnosed with the earliest stage of the disease are still alive five years later. This plummets to just five per cent when the cancer is diagnosed at its latest stage. We already know that early diagnosis is a successful strategy – thanks to Swanton’s work on heterogeneity, we now understand why on a more fundamental level.
So as well as developing new targeted drugs, we need to get serious about detecting cancer early, before it acquires the resources to evade treatment.
In the immediate future, funded by Cancer Research UK’s Genomics Initiative, Professor Swanton and his colleagues around the UK will continue to map the evolutionary history of more tumours.
If they can pin down the mutations that are commonly present in the ‘trunk’ of tumours’ Darwinian ‘tree’, and marry these up to new targeted treatments emerging from trials worldwide, we’ll have a fighting chance of successful cancer drugs that can really improve things for patients.
And this, allied to structural initiatives like our Stratified Medicine programme, and our work to diagnose cancers earlier through our work with the National Awareness and Early Diagnosis Initiative (NAEDI), mean the next decade is likely to see continued and accelerating progress in our vision of beating cancer.
- Gerlinger, M. et al (2012). Intratumor Heterogeneity and Branched Evolution Revealed by Multiregion Sequencing New England Journal of Medicine, 366 (10), 883-892 DOI: 10.1056/NEJMoa1113205