A Blog by

World within a tumour–study shows how complex cancer can be

When I used to work at a cancer charity, I would often hear people asking why there isn’t a cure yet. This frustration is understandable. Despite the billions of dollars and pounds that go into cancer research, and the decades since a war on cancer was declared, the “cure” remains elusive.

There is a good reason for that: cancer is really, really hard.

It is a puzzle of staggering complexity. Every move towards a solution seems to reveal yet another layer of mystery.

For a start, cancer isn’t a single disease, so we can dispense with the idea of a single “cure”. There are over 200 different types, each with their own individual quirks. Even for a single type – say, breast cancer – there can be many different sub-types that demand different treatments. Even within a single subtype, one patient’s tumour can be very different from another’s. They could both have very different sets of mutated genes, which can affect their prognosis and which drugs they should take.

Even in a single patient, a tumour can take on many guises. Cancer, after all, evolves. A tumour’s cells are not bound by the controls that keep the rest of our body in check. They grow and divide without restraint, picking up new genetic changes along the way. Just as animals and plants evolve new strategies to foil predators or produce more offspring, a tumour’s cells can evolve new ways of resisting drugs or growing even faster.

Now, we know that even a single tumour can be a hotbed of diversity. Charles Swanton from Cancer Research UK’s London Research Institute discovered this extra layer of complexity by studying four kidney cancers at an unprecedented level of detail. He showed that the cells from one end of the tumour can have very different genetic mutations to the cells at the other end.

These are not trivial differences. These mutations can indicate a patient’s prognosis, and they can affect which drugs a doctor decides to administer.  The bottom line is that a tumour is not a single entity. It’s an entire world.

This is one of the tumours that Swanton worked with. It was removed from the kidney of a patient being treated for cancer London’s Royal Marsden hospital. Swanton’s team took nine samples from different parts of the fist-sized mass, as well as other places where it had spread to – the space around the kidney and the patient’s chest.

Swanton found that even the primary tumour was surprisingly varied. He found 128 mutations among the various samples, but only a third of these were common to all of them. A quarter of the mutations were “private” ones – unique to a single sample.

The tumour had also split down two evolutionary lines. One area – part of R4 in the picture – had doubled its usual tally of chromosomes and seeded all the secondary tumours in the patient’s chest. The other branch had spawned the rest of the primary tumour. Even though this tumour looks like a single mass, whose cells all descended from a common ancestor, its different parts arehave all  evolved independently of one another.

For example, Swanton found that only one known kidney cancer gene – VHL – was consistently mutated throughout the samples. It was the only one present in the “trunk” of the tumour’s evolutionary tree (the blue line in the image below). The two main branches each picked up their own mutations. However, Swanton also found signs of convergent evolution, where different populations hit on the same adaptations on their own. In this case, three separate parts of the main tumour had disabled the same gene – SETD2 – in different and independent ways.

Three other kidney tumours showed the same degree of bewildering variety and convoluted history.

Darryl Shibata, a cancer geneticist from the University of Southern California, compares Swanton’s approach to the standard medical practice of taking a patient’s history. “By listening to a patient, a wise MD can just about predict what they will find on lab tests,” he explains. “An MD who simply orders tests will get back a baffling list of abnormal findings.” Likewise, reconstructing the evolutionary history of a tumour can make sense of the baffling array of mutations it contains. “Nothing in biology makes sense except in the light of evolution,” adds Shibata, quoting the famous biologist Theodosius Dobzhansky.

Swanton’s study has big implications for the way we study cancers, and could help to explain many of the failures and difficulties that clinicians have faced.

Consider the biomarker problem. For decades, scientists have tried to find molecules that could give away the presence of a cancer, how aggressive it is, what its weaknesses are, and more. They’ve identified tens of thousands of candidates, but only a hundred or so are actually used in practice.

Swanton thinks that this is because people often look for biomarkers after taking a single biopsy of a tumour. That’s potentially misleading. The biomarker might only be relevant to a tiny bit of a tumour, rather than the whole thing.  For example, in one of his patients, Swanton found genetic signatures associated with both good and poor prognosis in different parts of a single tumour. “To identify a robust biomarker, that biomarker must be present in all biopsies of the same tumour,” he explains. “A biomarker in the trunk is going to be better than a biomarker in the branches.”

The same goes for treatments. “Personalised medicine” is the next big thing in cancer. This term refers to drugs and treatments that target the specific milieu of faults within an individual patient’s tumour. But of course, a tumour isn’t a single thing. If you target mutations on the branches, you only hit parts of the tumour. Even worse, you leave a big “evolutionary reservoir” of unaffected cells, which can adapt to the drugs you’re using and grow anew. A tumour is like the mythical hydra. If you cut off a head, more will grow back; you need to kill the entire body.

This could explain why many current treatments eventually stop working. “We deal with patients, day in and day out, whose disease initially benefits from chemotherapy but who develop resistance over time,” says Swanton. “The only way you can fathom how a tumour can be one step ahead of the clinician at all times is through this type of diversity and Darwinian selection.” To get around this problem, scientists will need to find the “driver mutations” that are present in all parts of a tumour, in its trunk rather than its branches. These will present the most inviting targets for drugs.

Swanton’s results could also explain why kidney cancer patients fare better if surgeons remove their main tumour, even if their cancer has spread. “It may be that by removing the evolutionary sink of diversity, you improve a patient’s outcomes,” says Swanton. “There’s less material for the tumour to adapt with in response to environmental pressures.”

The study could also cause problems for scientists who are trying to decipher “cancer genomes”. By identifying the full suite of mutations behind any individual tumour, they hope to better understand how cancers develop and what their weaknesses are. But these studies have taken a single biopsy from each patient. And as we now know, that’s often not enough.

The problem is that taking many samples from a tumour is very difficult, and sequencing those extra samples would be very expensive. The alternative is to sequence the one sample over and over. It’s not ideal, but it would reveal small populations of cells that might be distinct from their neighbours and would otherwise stay hidden.  “This is likely to be the workhorse way we get at the problem,” says Andy Futreal from the Wellcome Trust Sanger Institute, who was involved in Swanton’s study.

Even this approach has drawbacks. More sequencing means more expense, and cancer genome projects are already costly affairs. “We may need to sacrifice overall numbers to really understand how deep the rabbit hole is,” Futreal says.

If this is all starting to sound very depressing, there is a glimmer of optimism. Even though cancer is maddeningly complicated, survival rates are still going up. At the moment, half of all cancer patients will survive for at least five years after their diagnosis. Cancer seems like an unbeatable adversary, but we are beating it. Studies like Swanton’s show that victory will not come in one decisive strike, but through a thousand shallow cuts. It will take a lot of time and effort.

Consider what Swanton needed to do to complete this study. He needed to piggyback on a clinical trial of kidney cancer patients and work very closely with the surgeons to ensure that all the samples were collected in the right way. It took four years just to get enough samples, and the man who resected all the tumours, Tim Christmas, passed away during that time. Swanton says, “He was an amazing guy and well recognised for being one of the most talented surgeons in Europe. We see this work as a legacy for him.”

Once he had the samples, Swanton threw technique under the sun at them. He sequenced their genes to look for individual mutations – the equivalent of typos in a book. He checked the activity of different genes.  He looked for big changes in the structure or number of chromosomes. Most similar projects typically focus on the mutations, but that’s like looking for typos when pages of a book have been smudged, or entire chapters have been rearranged.

“We used every possible genomics technique available,” Swanton says. “Even then we are only scratching the surface of the complexity within each cancer. It has been said you could have one whole scientific institute working on deciphering the genomic events in one tumour.” Even dealing with the data from the study was a challenge. “Our servers were clogged up for two to three months,” Swanton says. “We’re going to have to make a big investment on computing and bioinformatics expertise.”

That’s the type of project we need to really understand cancer’s true nature. Swanton’s study is just the beginning. So far, he has analysed four tumours. “We need to repeat our study in tens or hundreds of tumours, not just four,” he says.

Citation: NEJM, tbc.

Image by Katerha

22 thoughts on “World within a tumour–study shows how complex cancer can be

  1. Many Thanks for explaining in such a clear way so that even someone like me with very limited knowledge can understand.

  2. “Even though cancer is maddeningly complicated, survival rates are still going up. At the moment, half of all cancer patients will survive for at least five years after their diagnosis.”

    That’s not true at all. What we see is that we can diagnose cancer much earlier than before, but mortality is going worse. So sadly, the opposite is true 🙁

  3. This is fascinating, and you described what’s going on incredibly clearly. I wonder whether one implication of this research is that we will rely for longer than we hoped on the traditional chemotherapies even if we have an arsenal of targeted therapies. We hoped that targeted therapies would negate the need for them. But maybe the best approach will be to take the primitive, sledgehammer approach with the traditional chemos (because they are relatively undiscriminating) and then somehow sample what remains–which I’m not sure how to do if what remains are traveling single cells–and target that. This research also implies that early detection should be better because there would be fewer generations of cancer cells, and so fewer chances for new mutations to occur. It’s all really interesting to think about. And boy, am I glad Darwin wrote that book.

  4. @Leslie You wrote, “This research also implies that early detection should be better.” Absolutely! The big problem with early detection is actually doing it. There are only good screening techniques for three cancers (two, according to some) and many types only present with symptoms at a relatively advanced stage. The biomarker quest might help us to overcome this problem, but as we see, it’s easier said than done.

    The point about traditional vs. targeted therapies is a good one. My sense is that targeted therapies will be a big thing in the future but there are several rarely discussed obstacles in the way. This is clearly one of them. Cost, and ability to get regulatory approval may be others.

  5. Thanks, Ed. I think the cost problem–or maybe more precisely the return on investment problem–will be a lot bigger than most people assume. As it becomes clearer that there are, for example, not only more than one type of breast cancer, but potentially more than one type within each of those types, and potentially more than one type within each of those types within each patient, one can imagine scenarios in which many if not most of these targeted cancer therapies essentially become orphan drugs–the market for each will actually be relatively small. We’ll need either a new business model or a way of researching and developing these drugs that doesn’t primarily depend on the private sector, I think. I’m hoping that some of the recent research on cancer vaccines opens up new avenues to treatment, but wouldn’t this evolution problem and the orphan drug problem apply there, too? This is what leads me to the ironic thought that the sledgehammer approach may not be left behind, even if all these new targeted therapies do work. The one thing most cancer cells do have in common is rapid reproduction compared to most other body cells–lowest common denominator. Ugh.

  6. Stunning writing. This is absolutely what you do best, Ed. Take a complex study and write it up with style, depth and utter clarity.

    If/when you’re next looking for pieces to submit for prizes or anthologies, can I suggest that this one belongs near the top of the list.

  7. It strikes me that one useful approach, once sequencing technology gets cheap enough, might be to develop a background gene expression profile panel for various tissue types on a per patient basis (liver, colon, breast, thyroid, pancreas, prostate, and so on). If that patient is diagnosed with cancer, the comparison against the healthy gene expression profile, combined with the gene expression profile of the tumour (or regions of the tumour) itself would provide some useful information. After all, no two people have quite the same genes, expressed in quite the same way. Genomic sequencing in and of itself doesn’t quite give the same picture; I think this kind of epigenetic approach is going to prove useful in diagnosis and, possibly, in guiding treatment.

  8. Recently at a talk at MIT there was a guy from a company sequencing tumor genomes. It was certainly daunting, but they are getting actionable data out of the tumors. And like this data, what they found was that there might be 40 things identified that have gone awry.

    But the other part of that point was that it wasn’t hopeless to treat these tumors. Someone in the audience wanted to use this information to predict the demise of pharma–no more blockbusters. But instead the guy with the tumor samples suggested that what’s likely to happen is that people will get a cocktail of treatment as a result of knowing the full scope. Because there are some common and frequent mutations that occur, there are drugs to treat them. Now, I may end up with a different cocktail than someone else with a tumor–but more meds might go out overall with the new data.

    I actually found that a hopeful direction in an otherwise scary story.

  9. 3 years ago I was diagnosed with stage 1 kidney cancer. In my case, early detection came because of a kidney stone. The tumor was resected via a partial nephrectomy. No chemotherapy or radiation was needed as adjuvant therapy, as it was caught so early.

    I understand that the probabilities of its return are very low and I have kept up with check-ups. Had the kidney cancer been caught at stage 4, the prognosis would have been very different.

    My point is that incredible progress has been made in treating cancer as my experience shows: oncology has advanced to the point that it is know that stage 1 kidney cancer is treatable with only an operation, without need for additional intervention via chemo or radiation.

  10. Thanks for clear article. Have they considered crowdsourcing their research / funding ? Do they have a funding site? I’m sure many of us would contribute

  11. Do the cell lines that evolve differ in aggressiveness? Can we tell which mutant cells are going to behave in an indolent fashion vs. those that are likely to metastasize or invade other organs?

    Many of us with advanced cancers simply hope to manage the disease as a chronic illness. But our treatment strategy is to throw things at any growth that happens to show up, as if all growths were equal. Would we be smarter to use treatments which target the most dangerous cell lines?

  12. It appears to me that commenter #3 may have a point. At one of the links you provided, I note the following explanation: “These improvements largely reflect an increasing number of men being diagnosed with very early stage prostate cancer as a result of widespread use of PSA testing (Prostate Specific Antigen). Most men diagnosed at a very early stage will die with prostate cancer but not from it, therefore the survival rate has increased.” http://info.cancerresearchuk.org/cancerstats/survival/siteandsex/

    Age-adjustment of the survival rates seems helpful but not sufficient to isolate real changes in survival probability.

  13. @Mary – Swanton’s lab is funded by Cancer Research UK, the charity I used to work for. They generally don’t allow donations to be directed towards specific labs but I’m sure they would be grateful for one: http://supportus.cancerresearchuk.org/

    @Patient – People are working on discovering biomarkers for aggressiveness, but this runs into the same problem I mention in the piece. If you take a biopsy from an aggressive tumour, and look at the genetic markers within it, how do you know that those markers reflect aggressiveness? Maybe you took them from an indolent area? Swanton’s study is impressive but with only a dozen or so samples per tumour, it doesn’t have the power to identify aggressiveness markers by itself. That’s why it’s important to scale these analyses up.

  14. I did a short review of literature around mechanisms of resistance to cisplatin in 2002 (unpublished). Two things that were very clear were that a) even from the same parent cell line, cell lines derived in different laboratories rarely expressed the same mechanisms of resistance; and b) the strong selection pressures exerted on parent cell lines in order to express cisplatin resistance rarely resulted in just a single resistance mechanism. Of roughly twenty papers taken randomly from the literature, only seven papers used the same three cell lines; all the rest were derived on a per-lab basis. The result of this is that while one paper finds, for example, that over-expression of metallothionein conferred cisplatin resistance, another paper finds quite the reverse correlation. The same is true for ERCC-1 expression levels.

    Granted, many of these papers are from the late 1980s and 1990s (and I confess I haven’t kept up to date with the literature since 2002 — maybe an updated review might be in order), but the recent spectacular and disappointing failure of drugs such as picoplatin in Phase III trials against small cell lung cancer (and the corresponding collapse of Poniard Pharmaceuticals) indicates that the response of tumours to cisplatin and its derivatives is still very poorly understood.


  15. I commend you on a clear, well-written article about a complicated topic. I am a member of an on-line discussion forum maintained to support breast cancer patients. The members of that forum have found your article, and it is making many of them feel discouraged. (It’s an unfortunate consequence of your writing skills that your readers would finally understand something about tumor biology.) I have some background in molecular biology, which made me realize that my oncologist’s enthusiasm about treatment options for my own ER positive, HER2 negative invasive breast tumor might be short-lived. That background also makes me cringe when I read promotional material about high-end “molecular profiling” of tumors as a means of fine-tuning patients’ treatment protocols. That won’t be an effective strategy if we’re dealing with the biological equivalent of Heisenberg’s Uncertainty Principle.

  16. I first read your coverage of this on Nature News–came across it by chance, and was overwhelmed by the implications. When my chemist husband came home I tried, inadequately, to tell him about it, so did a Google news search. I started reading aloud to him the first link that came up–‘this is garbage,’ I said, and tried the next link.And the next. And the next. Some of the articles (from newspapers, news services, medical/science specialty sites) were just badly written. But even those that read well enough focused on this research signaling a setback for personalized medicine, and not crediting the researchers for their key insights into how individual tumors may possess their own, unique evolutionary trees. This is what I knew would interest my husband, a maverick who thinks the chemistry of life will probably never be understood because evolution’s blindfolded ramble through >a billion years’ time created near-infinite (with respect to man’s puny comprehensive ability) complexity, complexity that is everywhere characterized by layers of feedback and connectivity. Finally I re- found your Nature News piece, which we read together, marveling over both the scientific content and the quality of your writing. The whole experience was a shocking, fully electrifying, demonstration of how bad most science coverage is–and how very, very good it can be.

    Then, of course, I wanted to find out about you, eventually getting ’round to this page, which treated the work in greater depth, placed it in a broader perspective, and drew an immediate and enormous ‘AHA!’ from me. In Nature News you wrote, ‘These results show that any one biopsy presents only a keyhole view of a much bigger landscape’–a FANTASTIC metaphor that on its own prompted more than five minutes of discussion in this household. X-ray crystallography, for example, gives a worse picture than even a biopsy, since x-ray structures show a position that never exists in life. And your metaphor, itself, we found to be quite perfect. Keyhole views have been around, but yours struck as as singularly fresh and apt, especially with respect to genomics, its outrageous hype, and all the man-lives and dollars thrown at it.

    In science writing, doors and windows are frequently being opened, signaling some new way to access the previously inaccessible. Because of you and your keyhole we were struck by the implicit hubris of door/window comparisons. Keyholes, on the other hand. . .they’re not a way in; the narrow view they do provide is often distorted; and, on the utility scale, the information acquired about what’s beyond them is right up there with ‘an elephant is soft and squishy.’ The ‘AHA!’ came upon seeing the wonderful illustration at the top of this page.

    I’m sorry, not having known about you before, but now, after having made a run at your blog and contributions to Nature News and other periodicals, you top my science-writers-of-choice list. You are doing an enormous public service, and doing it uncommonly well!

  17. A wonderful review from the perspective of somatic mutations. You might want to check out some videos from the great-grandfather of genetics, James Watson…there is an undeniable genetic component in cancers…but even Dr Watson challenges the current scientists to look at the whole picture (many still don’t even know of the seminal work of Warburg and John Hopkins (Pederson) on glucose and glycolosis ..I’ll put my money on cancer as a metabolic disease…one disease where all cancer cells are vulnerable to metabolic manipulations. If you think your brain was spinning before, try on some of America’s other leading cancer researchers…whose studies and conclusions pretty much slam the current paradigm, like Soneschein and Soto from Tufts or Seyfried from Boston College…The world needs a cancer “general” who has the guts and brains to bring in all the research areas, and not continue to put all eggs in the genetic basket

Leave a Reply

Your email address will not be published. Required fields are marked *