For hundreds of years the Scottish Highlands have resounded to the names of their famous clans: MacDonald, Campbell, Fraser, and many more. Each clan is a complex, branching family tree, starting from a single person but evolving over the years into a plethora of related but distinct groups.
Trying to untangle the different branches of a clan is a complicated and painstaking job for genealogists, poring over detailed histories and dusty parish records. But the family trees they construct from this information reveal the story of a clan’s evolution over time.
Now Charles Swanton and his team at the Francis Crick Institute, funded by Cancer Research UK, have carried out a similar painstaking analysis of data from more than 2,500 cancers, covering nine different tumour types.
Their study, published in the journal Science Translational Medicine, reveals the genetic relationships between different groups of cancer cells within an individual tumour, shedding light on the evolutionary processes at work as cancer grows and spreads within the body and how we might harness them to treat the disease more effectively in future.
First, harvest your data
Unfortunately for researchers, tumours don’t come with a neatly detailed family history explaining exactly where they’ve come from. But what they do have is their DNA – which can be just as revealing.
Regular readers may be familiar with Charles’s name, as we’ve written several posts describing his research piecing together the evolutionary histories of kidney, lung and bowel tumours. And we’ve also discussed other studies looking at ovarian, breast and prostate cancers.
In these cases, researchers have looked at DNA from multiple tumour samples taken from a relatively small number of patients, building up detailed ‘family trees’ revealing how groups of cancer cells have changed and diversified over time. To return to the clan analogy, this kind of research is like a historian tracking down old records detailing long-dead ancestors then comparing them to a handful of living family members to work out how everyone might be related.
In their latest study, Charles and his team took a different approach. Rather than looking at multiple samples from a small number of tumours, they turned to The Cancer Genome Atlas. Known as TCGA – a clever play on the four ‘letters’ of the DNA code – this is a huge database compiled by US scientists containing DNA data from more than 5,000 tumour samples from patients around the world.
Unlike the multiple samples used in the studies mentioned above, taken from different regions of the same tumour, different parts of the body and at different points in time, each set of data in TCGA is a mishmash of all the genetic changes that have accumulated over time in just one tumour – usually the primary cancer removed from a patient when they first have surgery for their disease.
Rather than providing a historical picture of individual branches of a family tree, it’s more like a snapshot of all the living members of the various families in a whole clan, all at once.
Using clever computer analysis, Charles’s colleague Nicky McGranahan set to work reconstructing the evolutionary history of thousands of cancers based on 2,500 data ‘snapshots’ from TCGA, covering lung, bladder, breast, bowel, kidney, skin and head and neck cancers, as well as brain tumours.
Counting the clan
It works like this. Imagine you’re looking at a photo of all the people in a very strange clan, made up of several distinct but related families.
You notice that every single one of them has bright blue hair, whereas all the other people in the country are redheads. That tells you that the gene change (mutation) responsible for blue hair must have happened a long time ago, right back when the clan was founded.
Then you spot that all the members of about half the families have six fingers on each hand, while the rest have five. So this genetic alteration must have happened after the hair colour, back when there were only two families in the clan – one family got the six finger mutation, so all their descendants will also carry it, the other didn’t.
Finally, you see that each family within the clan has different coloured eyes – red, yellow, green, purple and more, as well as all kinds of other quirky unique traits. This diversity reveals that the gene changes responsible for this large range of differences must be recent, as they’re specific to each family rather than shared among the whole clan.
Straightaway, this is enough information to build a simple evolutionary tree showing how the families must have split over time, as well as the evolutionary relationships of the underlying genetic faults: the hair gene changed first, then the gene for finger number, then eye colour and everything else, creating a branching and diverse family tree. And it’s the same for the tumour snapshots within the TGCA database.
Trunk or branch?
Using these same principles, Nicky was able to figure out whether certain mutations had happened early on when the cancer first started growing (known as ‘trunk’ mutations) or later events (the ‘branches’ of the family tree) by looking at the proportions of different genetic alterations in the data for each tumour sample.
Although there were variations in the particular genes that were mutated in individual tumour samples across the different cancer types, he noticed a recurring pattern common to many of them.
Certain genetic changes known to be key drivers of cancer growth tended to happen early on in cancer development and were found in all the tumour cells. But sometimes important mutations – including those hit by the new generation of targeted drugs such as vemurafenib (Zelboraf) – affected a smaller proportion of the cancer cells, suggesting they were turning up later on.
This finding helps to explain why targeted therapies often seem to work well for a while – in some cases dramatically shrinking patients’ tumours – only for the disease to return with a vengeance months or even years down the line. In these situations, Nicky and Charles suspect that although the drugs may be targeting many of the ‘families’ of the cancer’s entire clan, some groups of cancer cells are unaffected by the treatment. This explains why cancers can return after treatment, as these remaining drug-resistant cells continue to grow and evolve.
Importantly, it appears that many of the mutations that happen early on during cancer development are in genes that have proved to be extremely difficult to hit with targeted therapies, such as a gene called Ras. Although our scientists and others around the world are working hard on these ‘undruggable’ targets, they’re not there yet.
However, this discovery does suggest that drugs designed to target these mutations could form the basis of effective future treatments that hit most, if not all, of the cancer cells in a patient’s body.
Another interesting thing that the researchers noticed was that many of the genetic mutations happening late on in tumour development bore the characteristic hallmarks of a family of DNA-altering molecules called APOBECs (something they’d noticed before in lung cancers).
APOBECs’ usual job is to protect our cells from viruses by attacking their ‘foreign’ DNA, but as cancer grows, APOBECs ‘go rogue’ and start attacking our own DNA. This creates lots of different mutations that encourage tumour cells to grow, spread and diversify further, generating many more ‘branches’ of the evolutionary tree.
For some tumours APOBECs appear to turn up relatively early, but for many cancers it was a later event. At the moment it’s not clear what triggers APOBECs to kick into action but they’re clearly an important mechanism for generating diversity within many cancers, fuelling rapid evolution and making the disease harder to treat.
We’ll be exploring these intriguing molecules in further depth in an upcoming blog post – we predict that they’re going to become an increasingly hot topic in cancer research in the future.
Opening the evolutionary rulebook
As we’ve followed the growing body of research studying tumour evolution and diversity emerging over the past few years, it sometimes feels as though cancer is a formidable foe, constantly changing within the body as it spreads and becomes resistant to every treatment that’s thrown at it.
But Charles and Nicky’s study shows that there seem to be common patterns to this seemingly unstoppable growth.
Figuring out the order in which certain genetic mutations seem to arise – a few faults first in the ‘trunk’ of the evolutionary tree, then a flurry of branching changes due to APOBECs – is an important step forward in writing the rules of cancer evolution. And although there’s still a huge amount we don’t know, there are several important points that can be drawn from this work.
One of the key take-home messages is that it’s not good enough just to look at a single tumour sample to see if it contains a mutation in a gene that can be targeted with a specific drug. It’s vital to know the proportion of cancer cells within it that carry that fault – all of them, or just a few ‘families’? The earlier the change in the history of that individual cancer, the more likely a targeted drug will be effective at treating it.
Another aspect is the possibility for developing entirely new ways of thinking about treatment. Instead of aiming to completely wipe out populations of cancer cells – leaving any resistant groups of cells free rein to carry on growing – there’s an opportunity to develop better ways of delivering therapy.
Perhaps there are ways of encouraging different groups of cancer cells to compete against each other, eventually eradicating them all. Alternatively, new treatments targeting the processes driving genetic diversity in tumours could force the disease down an evolutionary dead end where it can no longer adapt to treatment.
Another exciting approach is to ‘train’ a patient’s immune system to recognise some of the faulty molecules resulting from genetic mutations in their cancer cells, known as ‘neo-antigens’. Scientists in the US recently published a very early stage study showing that this approach could work in principle, and it will be very interesting to see how this research progresses.
Finally, a clearer understanding of the gene faults that drive cancer could shed light on better combinations of therapies – including drugs, radiotherapy and immunotherapy – that will target every single cancer cell in the body, rather than just a proportion of them.
This research is just the start, and Charles and his team are continuing to fill in the pages of this evolutionary rulebook in ever more detail – for example, through the TRACERx study which is tracking the genetic changes in tumour samples from more than 800 lung cancer patients.
Unpicking the enormous genetic complexity and diversity in cancer is a big challenge, and it’s one we must overcome if we’re to make progress in treating advanced cancers that have spread through the body. But, thanks to research, it’s a problem we’re starting to solve.
- McGranahan, et al. (2015). Clonal status of actionable driver events and the timing of mutational processes in cancer evolution, Science Translational Medicine. DOI: 10.1126/scitranslmed.aaa1408