The emotion and anxiety aroused by a single word – ‘cancer’ – spans ages, sexes, nations, races and classes.
But as we understand more about the disease, the idea that cancer is a single, common enemy, is increasingly being challenged.
In late 2009, the publication of the first complete cancer genomes showed the extraordinary chaos present in the DNA inside cancer cells. But they also highlighted the molecular differences between different types of cancer – in this case, skin cancer and lung cancer
Other large gene studies have revealed even more differences between types of cancer, but have also increased out understanding of the differences between the ‘same’ cancer type in different people – the foundation of ‘personalised medicine’.
For example, last week a team of Canadian and British researchers, writing in the journal Nature, analysed the DNA from 104 ‘triple-negative’ breast cancers – a particularly hard-to-treat form of the disease.
As this in-depth post on the Respectful Insolence blog describes, they found that no two women’s cancers were alike – there were differences across all the tumour samples. Even a subcategory like ‘triple-negative’ breast cancer doesn’t seem to be a single disease (a point we’ll return to later). And genetic differences also appeared between cells from the same tumour – known as ‘intratumour heterogeneity’.
This point was emphasised a few weeks earlier by researchers at our London Research Institute. They analysed multiple samples from the same patient’s kidney tumour and secondaries (where the cancer had spread to other parts of the body).
No two samples were identical, suggesting that there’s significant variation even inside a tumour. As we discussed in this blog post, it looks like tumours can be highly varied, creating new challenges in the search for personalised medicine.
Which brings us to today’s news, of a landmark Cancer Research UK-funded study published in Nature.
Through intricate genetic analysis, the same British and Canadian researchers, led by Professor Carlos Caldas from our Cambridge Research Institute and Professor Sam Aparicio from the British Columbia Cancer Centre in Canada, have uncovered crucial new information about breast cancer.
Their study group, METABRIC (Molecular Taxonomy of Breast Cancer International Consortium), looked at the patterns of molecules inside tumours from nearly two thousand women, for whom information about the tumour characteristics had been meticulously recorded.
They compared this with the women’s survival, and other information, like their age at diagnosis.
While many other studies have highlighted differences between cancers, the METABRIC study looked at so many tumours that they could spot new patterns and ‘clusters’ in the data.
Their conclusion is that what we call ‘breast cancer’ is in fact at least ten different diseases, each with its own molecular fingerprint, and each with different weak spots.
This is simultaneously daunting and heartening – daunting because each of these diseases will likely need a different strategy to overcome it; and heartening because it opens up multiple new fronts in our efforts to beat breast cancer.
Let’s look at the background to the study, then in detail at what the researchers actually did, what they found, and what this means for the future of breast cancer treatment and diagnosis.
A changing view
Researchers studying breast cancer down the ages have long suspected it’s a complicated disease.
The first clue came more than a hundred years ago. Scientists found that most breast cancers make too much of a protein called the oestrogen receptor (or ‘ER’), and this allowed the hormone oestrogen to drive their growth.
As a result, ‘anti-oestrogen’ drugs like tamoxifen and anastrozole were invented. And now women are routinely given tests for levels of ER in their tumours. This helps decide whether to prescribe these drugs alongside standard treatments like surgery, radiotherapy and chemotherapy.
The progesterone receptor (PR) was discovered next.
Women whose breast cancers are ‘PR+’ are also likely to respond to anti-oestrogen drugs, and those whose tumours are both PR+ and ER+ tend to have the best outlook.
This suggested three ‘types’ of breast cancer –
- ‘double-positive’ cancers that had high levels of both hormone receptors
- Cancers with high levels of either ER or PR
- ‘double-negative’ cancers
But the idea was soon shown to be simplistic. For example, some women with ‘double-positive’ cancer did better on tamoxifen than others. And this variable response was seen for other types of breast cancer. Clearly something else was going on in at a deeper level.
In the late 90s – underpinned by work at our London Research Institute in the 1980s – a new molecule called Her2 was discovered to drive breast cancer in some women. This discovery led to the drug trastuzumab (better known as Herceptin).
And again, women can be tested for Her2, and are ‘Her-2 positive’ if their tumour contains high amounts of the protein. These cancers also tended to have worse outlook (at least until the advent of trastuzumab).
But this wasn’t the whole picture. Again researchers found that some ‘Her2-positive’ women responded to Herceptin better than others.
Today, in the NHS, women are routinely given tests for the oestrogen and progesterone receptors, and for Her2 – and their treatment determined by the results.
For some women, this means their cancer is ‘triple-negative’ – their tumours contain low or normal amounts of ER, PR and Her2, and there are no extra treatments available for them. These cancers also tend to be very aggressive.
All of the tests described above measure the levels of proteins inside tumours. Recently, research has focused on testing which genes are switched on or off inside the cancer cells.
This has led to tests, not yet widely used in the NHS, such as ‘PAM50’. This examines 50 separate genes inside a woman’s tumour, and uses the resulting ‘fingerprint’ to group cancers into four subtypes’:
- Luminal A cancers, which are usually ER+ and/or PR+ – and make up about half of all cases. They tend to have low amounts of Her2. Women with these tumours tend to have the best outlook.
- Luminal B cancers, which again tend to be ER+ and/or PR+, but also Her2+. These have a good outlook (but not as good as luminal A cancers), and account for about 12 per cent of cancers.
- Her2-amplified cancers. About one in ten cancers are ER and PR negative, but have high levels of Her2. These tumours have a poorer outlook than the two types above, but can be treated with trastuzumab (Herceptin).
- Basal-like tumours – these are usually the ‘triple-negative’ cancers mentioned above, and make up about 20 per cent of tumours. They have the least favourable outlook.
There’s a more detailed description of these genetic subtypes here.
Other commercially available gene-based tests are also available, such as Oncotype DX and MammaPrint. There’s more information about these here.
Such gene tests can be used like the protein-based tests – guiding treatment and helping doctors predict how the tumour will behave. But they’re more expensive, and there’s little hard evidence that they outperform the high-quality protein-based tests currently offered by the NHS. Newer is not always better, from a patient perspective.
Overlap and confusion
Researchers following the fates of women diagnosed with different types of breast cancer soon noticed that, occasionally, women with a supposedly ‘poor survival’ form of cancer would respond very well to treatment.
Similarly, some women who ‘should’ respond to drugs like trastuzumab, (because their cancer seemed to contain high levels of Her2 protein), didn’t.
Even the four ‘genetic’ types of cancer can’t reliably predict the best treatment or how a patient’s tumour would behave.
So, as it stands, our current map of breast cancer just isn’t detailed enough. It outlines broad, vague, overlapping continents, rather than showing clear boundaries between countries, counties or cities.
The research published today gives us a newer, better map of breast cancer.
The METABRIC project has produced a “goldmine” of data, according to Professor Caldas.
Over a period of decades, the team has carefully developed a resource of thousands of individual tumours, which can be (and have been) analysed and reanalysed in a number of different ways.
Crucially, each of these tumour samples is linked to detailed information about the subsequent fate of the women they were from. And, just as crucially, a large number of the samples are linked to a ‘matched normal’ sample (in other words, a sample of non-tumour DNA from the same patient).
The METABRIC team used sophisticated DNA hybridisation machines to analyse regions of variation in the DNA of nearly 1000 tumours. These regions, called copy-number variations, occur when a dividing cell makes mistakes, and result in regions of DNA being repeated, or deleted. Both of these alterations can affect gene activity.
By measuring such changes in each of the thousand samples, the team generated a ‘copy-number map’. And because they had the matched normal samples, they could remove any naturally occurring gene variations. As a result, they were able to draw up the first ever map of copy-number alterations specific to breast cancer.
The researchers also looked for other types of ‘single letter’ genetic variation, called SNPs, for each tumour sample.
And to cap it all off, they measured which genes were active in each sample (so-called ‘gene-expression data’), by measuring levels of molecules called RNAs. In all, they measured the levels of more than 30,000 different types of RNA, each corresponding to the activity of a single gene.
Annotating the map
But a map is useless without directions. The METABRIC data allowed unprecedented comparison of gene data – copy-number, SNPs and gene-expression – with the outlooks of the patients the tumours came from. This was done using specialised computer software to sift gigabytes of data, searching for patterns.
What they found was remarkable. Distinct patterns emerged from the data, with certain copy-number aberrations far more likely in women with certain clinical features (such as high levels of the oestrogen receptor, or of Her2), or with distinct, and shared, clinical outlook.
In all, there were ten clear ‘clusters’ in the data – corresponding to ten entirely new ways to classify breast cancer.
To check this finding was accurate, they then painstakingly repeated the analysis on a second panel of nearly 1000 tumours. Again, the same ten subtypes shone out from the data.
Finding the drivers
One thing immediately became clear from the new map. Different ‘clusters’, or subtypes, seemed to be characterised by genetic variations at certain ‘hotspots’.
These were associated with the activity of large groups of other genes. This is what you’d expect if a single genetic mutation was turning on lots of different cell processes associated with cancer.
The team looked in detail at these genetic hotspots, to see if they could find the gene or genes responsible. This analysis showed up some ‘old friends’ – for example the gene that makes the Her2 protein. But it also revealed some entirely new genes that had never previously been linked to breast cancer.
These are potentially excellent targets to develop the next generation of Herceptin-like drugs.
Intriguingly, one of the clusters (Cluster 4, below) contained both ER+ and a sub-set of triple-negative tumours with – paradoxically – a good prognosis. But these tumours didn’t show any large-scale gene defects. Instead, the team spotted subtle deletions, not in cancer genes, but the genes of the immune system.
Checking the original biopsy revealed what was going on. Looked at down the microscope, these tumours were packed full of white blood cells. It seems that, for some women with a supposedly poor outlook, their own immune system somehow comes to their rescue, holding the tumour at bay. This had been observed before, but never in such detail. And certainly not as a distinct ‘type’ of breast cancer.
The ten ‘clusters’
Here’s an overview of the characteristics of each of the clusters identified:
|Cluster||Outlook||Copy number defects||Comparisons and other notes|
|1||Intermediate||Chromosome 17||‘Luminal B’-like, generally ER+|
|2||Poor||2 x faults on chromosome 11||Mixture of luminal A&B|
|3||Good||Very few||‘Luminal A’-like|
|4||Good||Very few, mainly immune system genes||High levels of immune cells in tumour|
|5||Extremely poor||Chromosome 17 (Her2 gene)||Mixture of ‘Luminal B’ and ‘Her2’|
|6||Intermediate||Region of chromosome 8 deleted||ER+, generally Luminal|
|7||Good||Chromosome 16||Luminal A|
|8||Good||Chromosomes 1 & 16||Luminal A|
|9||Intermediate||Ch 8 and/or 20||Luminal/ER+|
|10||Poor 5-year outcome; good long-term outcome if alive at 5 years||Chrs 5, 8, 10 and 12||Basal-like|
It’s important to say that this new way of looking at breast cancer won’t affect the way anyone with breast cancer is currently treated. It advances how scientists approach future research and clinical trials, but there’s not enough information to know how to put this information to use in a way that will benefit patients. The standard tests available on the NHS are sufficient for the drugs currently available, for the time being (although as new treatments emerge, this will change).
And these ten subtypes aren’t the end of the story. Already it looks like ‘cluster 4’ can be further broken down, and Caldas’s team is planning on analysing other aspects of his sample collection to increase the resolution of METABRIC’s map further. He’s currently looking at the levels of tiny molecules called miRNAs in each tumour, adding this data to the computer clusters. This will doubtless yield even more insights.
They also plan to fully sequence 150 genes in each sample, focusing on those most important in breast cancer. Again, this should refine the map even further.
Another priority for the team is to focus on the links between the immune system and cancer. Immunotherapy – treatments that harness the immune system to fight a patient’s cancer – is a hot topic, and has huge potential in breast cancer.
Understanding how some women’s immune systems can fight their own tumour raises the prospects of treatments that can do likewise.
According to Professor Caldas, “the fact that we have collected matched normal blood samples for all the patients is testament to the set-up here in Cambridge – of high-quality, long-term genetic research”.
“That in turn is testament to the investment Cancer Research UK has put in here, to build and nurture an environment where this sort of discovery is possible”.
Other researchers are similarly excited by the new results. Professor Charles Swanton is the Cancer Research UK-funded researcher who discovered variegation in kidney cancer (as mentioned earlier). He called the new results an “extraordinary” finding that took our understanding of breast cancer “to the next level”.
“We’re increasingly seeing that what was previously thought to be a single ‘type’ of cancer can be subdivided into smaller clinically relevant groups through cancer genomics analysis.
“Breast cancer is perhaps the stand-out example of this, and analyses of gene expression over the years have revealed a hidden complexity in this disease,” he added.
He was also quick to point out the international nature of the work: “METABRIC also illustrates the importance of large international collaborations, and the need to include sufficient numbers of patients to be able to work out what’s going on in smaller and smaller patient groups.
“This extraordinary effort is likely to have important implications for clinical trial design in breast cancer and will prime researchers worldwide to define new approaches to treat each subgroup,” he confirmed. His hope, like Professor Caldas’, is that the ultimate end will be improved outcomes for patients.
And speaking of patients, Professor Caldas had this to say:
“I want to stress, this study wouldn’t have been possible without the breast cancer patients who donated their samples and agreed to take part in the study. None of this would have happened without them, and I’m so grateful for their participation.”
METABRIC is a landmark study, and a shift in how we view breast cancer. It will shape future research, including the search for newer, better treatments – particularly for those with the worst outlook. It could also lead to women with the best prognosis being spared the side effects of chemotherapy they don’t need. And the classification system it sketches out will likely form the basis for newer, better ways to diagnose and manage the disease.
Thanks to the generosity of two thousand breast cancer patients who took part, things look more hopeful for their daughters, grand-daughters and great-grand-daughters.
Find out more:
- This week’s headlines: Why a breast cancer ‘blood test’ is still a work in progress.
- For more information about breast cancer, please visit our CancerHelp UK website
- Help support our work to beat cancer
Curtis, C. et al. (2012). The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups Nature DOI: 10.1038/nature10983