An electron microscopy image of a breast cancer cell

A breast cancer cell Credit: London Research Institute EM Unit

In 2012, our scientists redefined breast cancer.

By analysing samples from nearly 2000 tumours they found that breast cancer was not one, but 10 different diseases, as we explained in our blog post.

Led by Professor Carlos Caldas, based at the Cancer Research UK Cambridge Institute, and Professor Sam Aparicio at the British Columbia Cancer Agency, the team discovered they could tell tumours apart by looking at faults in their DNA. And these differences defined how these breast cancers behaved, from how likely they were to respond to treatment to how likely tumours were to come back after surgery.

Since then the team has tested their rulebook on even more samples collected by labs around the world. And regardless of where the samples came from, the new set of rules held true.

“What we have here is a universal classification of breast cancer,” says Caldas.

And 7 years after the first study was published, the research team have shown that their rulebook stands the test of time. Publishing their latest work in Nature, they’ve found that the group a woman’s breast cancer is initially placed in could predict how they will fare 20 years after diagnosis.

Writing the new genetic rules

Scientists are constantly looking for new ways to tell tumours apart. They’re searching for something that can help them know if a cancer is likely to be aggressive or if it’s likely to come back. Does it need treating? And if so, how?

That something could be a molecule on the surface of cancer cells – like hormone receptors – or a fault nestled in cells’ DNA.

For years doctors have used molecules on the surface of breast cancer cells to reveal what’s buried within. Markers like oestrogen receptors and HER2 have helped patients and doctors to decide what treatment might be best.

Other tests have come in that look at a tumour’s DNA, including one called PAM50. But Caldas and his team began to look for a more accurate way to distinguish between different breast cancers.

“The big thing about our 2012 study was that we were distinguishing between tumours based on their molecular wiring – what’s going on inside the cells – rather than their prognosis,” he recalls.

The team split the cancers into 10 groups – which they call clusters – based on this molecular data.

And Caldas and his team wanted to see if grouping tumours in this way can more accurately predict the future for these patients than current tests.

Using samples from 1,980 women with breast cancer, they plotted the likelihood of the tumour returning (relapse) or spreading (metastasis) based on its molecular information. And then they looked at how well this prediction stacked up against tests like PAM50.

“What we’ve found now is that if you follow patients for up to 20 years after diagnosis, the women have very different propensities to relapse or even for their disease to become metastatic depending on which subtype of breast cancer they have.”

Importantly, separating tumours based on Caldas’s classification was a better predictor of future disease than looking for molecules on the surface of tumours or the PAM50 test.

Refining the rulebook

The long-term data has done more than just confirm the potential value of classifying breast cancer in this way, it’s also refined it.

“We realised that one of the groups, cluster 4, needed to be split,” says Caldas.

Cluster 4 was the largest group in the original study, accounting for around 1 in 5 tumours. Tumours in this cluster don’t have any major damage to their DNA, which Caldas says is why they were originally grouped together.

“The group had breast cancers that were oestrogen receptor positive and some that were triple negative, but they were clustered together based on features in their DNA.”

But over time Caldas and his team realised that despite the genetic similarities, the oestrogen receptor positive and negative tumours had very different clinical outlooks. So, in the latest paper, they’ve split them into 2 different groups – creating 11 different types of breast cancer in total.

Beyond triple negative breast cancer

The team also learned more about triple negative breast cancers. Defined by the lack of hormone receptors or HER2 on the cancer cells’ surface, they’re considered to be one of the most aggressive forms of the disease.

But according to Caldas’s research, not all triple negative breast cancers are the same.

“We can clearly define two groups of triple negative breast cancers. In one group, women have a very good outlook if they haven’t relapsed after 5 years. Unfortunately, women in the other group have a high chance of their cancer returning even after 5 years.”

This was somewhat surprising for Caldas, as in the first study the outlooks were reversed. Women whose cancer was more likely to come back in the long term had a better prognosis in the first 5 years.

The results not only reinforce the importance of following patients up over long periods of time, it also calls into question how trials are run.

“These patients have very different clinical course – but at the moment they’re treated and followed up in the same way. And trials include both groups,” says Caldas.

Caldas believes that these differences should be considered when new treatments are tested.

“We’re already starting to do it,” he adds. Caldas’s colleagues, led by Dr Jean Abraham, are running a clinical trial called PARTNER, where they’re identifying the type of triple negative breast cancer each women has before they enter the study.

“They’re aiming to get equal proportions of women with the two types of triple negative in each arm of the trial,” says Caldas

The trial is testing if adding a targeted drug called olaparib (Lynparza) to standard chemotherapy is better for women with triple negative breast cancer. The trial is also open to women whose tumours have a faulty BRCA gene. And by taking the different types of triple negative cancer into consideration at the start of the trial, Caldas hopes it will tell researchers if there are differences in how the groups should be treated.

The PARTNER trial is a start, but Caldas thinks it needs to go much further.

“We need to take this into account in more clinical trials. To continue to recruit women based on their hormone receptor status is really nonsensical to me.”

Getting personal

What the team discovered is intimately linked with other projects in Cambridge, including the Personalised Breast Cancer Programme. The project aims to tailor treatment to individuals by mapping their DNA.

So far, they’ve recruited 400 women with breast cancer to the programme, all of whom have had their tumours classified into 1 of the 11 subtypes.

“We’re going to be following these women knowing which cluster they’re in from the word go,” says Caldas.

And with 85% of women in the programme enrolled into a clinical trial or a monitoring study, Caldas says they’ll be learning a lot about the best way to treat these women.

The team are also looking to go one step further. The plan is not just to see which treatments work for which group – but to develop drugs specifically for them.

“We’re creating models of as many of the women’s tumours as possible. And we’re going to use this to try and get to the next step, which is identifying therapies that are tailored to each of the clusters.”

What’s next?

While the rulebook Caldas and his team are writing has proved essential reading so far, they’re not done revising it just yet.

“We’re planning to do an even bigger validation in the next few years, looking at the 11 subtypes in around 10,000 cases of breast cancer,” he says.

They’re also looking to develop a test that can be more easily used on the NHS.

So far, samples have been analysed using expensive equipment that isn’t widely available. And they’ve been done using samples collected during surgery and preserved for this research, which isn’t feasible in an NHS setting. To get a test that could be used in the clinic, Caldas says that it must work on tissue samples that are preserved by locking them in paraffin wax.

“We’ve got a Cancer Research UK-funded pathologist in my lab who is trying to develop a test that can be done on paraffin-embedded material affordably. And by that, I mean a test that wouldn’t cost more than £500.”

From research tool to patient benefits

It’s hard not to feel optimistic when hearing Caldas talk about the potential of the new breast cancer rulebook. Defining breast cancer in this way could change how women are treated and followed up in the future. But we’re not there yet.

“What we have at the moment is a research tool, it’s not useful in the clinic yet.”

The most important aim for Caldas and his team now is to figure out if using the test can change the outlook for women with breast cancer for the better.

“What we need to do now is demonstrate that by using the test we can improve a patient’s outcome. And that’s quite a high bar to reach.”

Caldas and his team are planning to implement the test through a programme that’s already used in the UK to help make treatment decisions for breast cancer, called PREDICT. This tool, developed by Professor Paul Pharoah’s team, already uses a variety of information to make treatment recommendations – including the size and grade of the tumour and if it’s positive or negative for HER or oestrogen receptors.

“What we’d essentially be creating is a PREDICT plus,” says Caldas. “We’d add our test on top of what PREDICT already offers and see if it’s better at individualising recommendations for treatment and follow-up.”

They’re not ready to launch PREDICT plus just yet, but it’s a goal they hope to move towards in the coming years.

Caldas and his team have big ambitions – to turn what they’ve learnt into a test that could help give women diagnosed with breast cancer more certainty about their future and treatment options that are right for them.

And while they’ve still got a long way to go, the latest findings prove they’re moving in the right direction.


  • If you have questions about breast cancer you can speak to our nurses on freephone 0808 800 4040.


Rueda, O et al. (2019) Dynamics of breast cancer relapse reveal late recurring ER-positive genomic subgroups. Nature. DOI: 10.17863/CAM.37037