Lung cancer evolution – a journey through space and time

Not so long ago, the top floor of Cancer Research UK’s London Research Institute housed a library, where the Institute’s 300-odd researchers could look back over decades of archived cancer research, searching for clues left by their predecessors around the world.

Today – as you might expect in the digital era – the library has been banished online; books and journals long gone.

In their place, glinting in the sunlight, stand a series of large, black monoliths, studded with twinkling LEDs: advanced computers employed not to study the paper history of cancer research, but to map out the genetic history of individual tumours themselves.

Capable of drawing up a detailed DNA map of a patient’s tumour in a matter of days, computers like these are allowing cancer’s complexity to be scrutinised in unprecedented detail.

And in new research published today in the journal Science, two research groups – including that of the Institute’s Professor Charlie Swanton – have revealed new genetic insights into one of the most common yet poorly-understood of all cancers: lung cancer.

Reading a cancer’s history

Researchers studying cancer’s DNA have made recent leaps forward in understanding two aspects of the disease. The first has been in unravelling how the genetic chaos inside a tumour accrues over time.

In 2012, for example, Professor Swanton’s team showed that – just as Darwin might have predicted – kidney tumours evolve from a single group of damaged cells, but rapidly diversify as they grow and spread – ending up as a variegated, motley crew of disordered clumps, each harbouring its own unique spectrum of DNA errors.

Others have revealed a similar situation – known as ‘intratumour heterogeneity’ – in other types of cancer, including breast, prostate and ovarian cancers; leukaemia; and glioma brain tumours.

The stark message is that, to cure patients with advanced cancer, the rules governing how tumours evolve need to be worked out urgently.

Fingering the culprits

The second has been to uncover the fundamental processes causing this genetic chaos.

Two classics of this genre emerged in 2009 from the Wellcome Trust Sanger Institute in Cambridge. Focusing on lung cancer and melanoma skin cancer, researchers pinpointed the fingerprints of tobacco smoke and sunlight in the faulty ‘letters’ of a tumour’s DNA.

They followed this up with a tour de force in 2013, discovering at least twenty different chemical processes leaving tell-tale fingerprints in cancer’s DNA.

So researchers now have two powerful weapons to crack open cancer’s secrets: a way to reconstruct a tumour’s genetic history; and the ability to identify the forces that shape it.

But can these two approaches be combined? Can the different processes acting during a tumour’s evolutionary history be unravelled?

Working with colleagues at the Sanger Institute, Swanton’s team – in parallel with another in the US led by Professor Andy Futreal (himself formerly based at the Sanger) – set out to do just that in lung cancer: the world’s biggest cancer killer, and one of the hardest to research.

The search for a smoking gun

Swanton’s team – led by lab researcher Dr Elza de Bruin and biological data expert Nicky McGranahan – collected a series of tumours from seven patients with early stage lung cancer who’d had surgery aimed at curing them. Just as in their kidney cancer study, the team analysed and compared DNA from several different regions of each tumour.

The first two tumours they looked at were from former smokers, who’d given up at least twenty years beforehand.

Both tumours showed all the characteristic diversity and chaos seen in other cancer types – on average, only about 70 per cent of a tumour’s DNA errors were found throughout the whole tumour, the rest being unique to one region or another. This implied that the tumour’s early life was relatively homogenous, with a late explosion of diversity.

And – as expected – they saw the tell-tale faults known as ‘C-to-A’ changes, caused by tobacco carcinogens.

But there was something else going on too. The C-to-A changes were much more likely to be early, common errors than to be unique faults confined to specific regions of each tumour. These late-occurring mutations seemed to be caused by something else: DNA-editing proteins called APOBEC proteins.

“APOBEC proteins are turning out to be one of the most important DNA damaging forces in a range of cancers,” says McGranahan. Under normal circumstances, he explains, APOBECs are thought to play a role in protecting us from viruses by damaging their DNA. But there’s now a wide range of evidence that they get switched on accidentally as cancers develop.

“Although we didn’t suspect APOBECs,” he says, “we weren’t that surprised to find that something other than tobacco was driving cancer development in ex-smokers.”

They saw a very similar pattern in another ex-smoker’s tumour: decades of early tobacco damage, and a late explosion in APOBEC damage.

But the big surprise came when they looked at three tumours from people who hadn’t given up smoking.

“Weirdly, the APOBEC signature was still there. It was really quite surprising. We expected maybe just to see a long history of tobacco-related damage in smokers. But just as we saw in the patients who’d given up, tobacco damage was overtaken by an explosion of APOBEC mutations. It’s clearly something fundamental driving lung cancer,” says McGranahan.

APOBEC’s fingerprints were also present in the tumour DNA from the one non-smoker on the study, which otherwise lacked the smokers’ tobacco-related fingerprint.

Other implications

The team made three other important observations.

First, it looked like the cancer-fuelling ‘driver’ mutations in genes like EGFR and ALK – against which a new generation of therapies is targeted – tended to occur early in a lung cancer’s development.

This contrasts with kidney cancer, where similar mutations tended to occur relatively late on. This is good news, as it suggests that these so-called targeted therapies might be more effective in lung cancer, particularly if used in combination.

Second, in four of the seven samples, it looked like the cancer cells contained not one but two copies of the human genome. This seemed to happen before the APOBEC proteins wrought their chaos, but after most of the tobacco damage. This doubling had previously been seen in other studies, but no-one knew when in a cancer’s development it occurred.

And third – and perhaps most importantly – it’s more evidence that lung cancers take a very long time to develop – maybe more than twenty years.

Put this all together and it has quite profound implications: it might be possible to detect early signs of the disease in the blood long before symptoms develop – something that could make an enormous difference to patient outcomes, as Professor Jacqui Shaw of Leicester University, who studies circulating tumour DNA in breast cancer, explains:

“A growing tumour can shed its DNA into a patient’s blood as its cells die,” she says, pointing to a research study in 2006, which showed that certain key mutations could be spotted in the blood of people who went on to develop cancer.

“At the moment, we don’t know enough about whether early lung tumours shed DNA into the bloodstream – it’s a bit of a black box – but it’s certainly an idea that’s got a huge amount of potential, and one we’re eager to explore” she says.  Swanton’s study gives further impetus to this research.

Across the pond – the US study

But as we mentioned earlier, Professor Swanton’s team weren’t the only ones looking at lung cancer. A US team, led by Professor Andy Futreal, had been conducting very similar research.

Happily, Swanton’s team’s findings agree closely with the US study. Across eleven tumour samples, Professor Futreal’s group found only about three quarters (76 per cent) of a tumour’s mutations were ubiquitous – similar to the 70 per cent Swanton’s team found. And they also spotted the APOBEC signature at work in the lung cancer genome.

But because Futreal’s team were able to follow up the patients on the study, they were also able to tentatively make one extra, important observation. As of 21 months after having their tumours removed, three of the patients’ cancers have come back. And these three patients had the highest degree of overall genetic variation in their tumours.

“This suggests something we’ve long suspected in lung cancer,” says Swanton. “Heterogeneity – the degree of chaos within a tumour – could be a really important marker of whether a cancer might come back. If we can confirm this, we could immediately use it to help these patients, by offering them extra treatments like chemo, radiotherapy, or targeted drugs if appropriate”, he says.

Burning questions

As with all the best research, these studies raise as many questions as they answer:

What are the causes of the genome doubling, and what activates the APOBEC system? If these triggers can be understood, then it raises the possibility of stopping them.

What triggers lung cancer in non-smokers? Only three non-smokers were involved across the two studies – too few for firm conclusions. As smoking rates decline, this question is becoming more and more pressing.

Can the degree of genetic chaos, or heterogeneity, accurately predict whether a patient’s cancer will come back? As discussed above, if confirmed, this could immediately be used by doctors to try to improve treatment for patients.

But most importantly, can lung cancer DNA be detected in the blood before symptoms develop? As we’ve discussed before, lung screening using CT scans can reduce deaths, but is fraught with problems of false-positives and unnecessary treatment. But as Professor Shaw says, a blood-based test used alongside chest scans, could be a game changer, “if its utility is  proven in clinical trials”.

But the size of the prize is huge: a reliable method of detecting lung cancer early could prevent countless thousands of premature deaths. Only 14 in every hundred patients diagnosed with late-stage lung cancer survive for five or more years. For early-stage disease, that goes up to 71 in a hundred.

The idea that lung cancer’s long clinical latency could be exploited for early detection is incredibly tantalising (and fits with the lag between smoking rates and lung cancer rates).

Thankfully, many of these questions will be answered by TRACERx, a large, Cancer Research UK-funded initiative launched last year, which Professor Swanton is leading.

This, as we explained when it launched, is the largest single investment we’ve ever made in lung cancer, and will meticulously analyse tumour samples from 850 patients as they go through treatment – far more than the 18 involved in the two studies we’ve discussed here. TRACERx will bring us much closer to answers to these questions, and – hopefully – finally start to improve survival for lung cancer patients.

For too many years, understanding lung cancer has proved elusive. It’s as if researchers have been stuck in a darkened room, trying to discern the size of the challenge, feeling only tentative suggestions of its true nature.

With studies like these, the lights are slowly flickering into action, and the room’s contents are being revealed. And in the centre of the room are a cluster of large, black, twinkling servers, analysing billions of letters of DNA, yielding up cancer’s inner secrets, and shining a light on its past.

Henry

• Professor Swanton’s study was supported by Cancer Research UK and the Rosetrees Foundation

Reference

• de Bruin EC et al, Spatial and temporal diversity in genomic instability processes defines lung cancer evolution. Science (2014) DOI 10.1126/science.1253462
• Zhang J et al, Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing. Science (2014) DOI: 10.1126/science.1256930

Images

• Public domain image of chest x-ray via Wikimedia Commons
• Public domain image of DNA damage via Wikimedia Commons
• Image of blood samples by Vegasjon (Own work), licensed under CC-BY-SA-4.0, via Wikimedia Commons