Name: Ian Campbell
Title: Associate professor, Peter MacCallum Cancer Centre
For more than two decades, Ian Campbell has been using the best technology available to understand the molecular genetics of ovarian and breast cancer. As head of the Cancer Genomics Research Laboratory at the Peter MacCallum Cancer Centre in East Melbourne, Australia, Campbell has seen studies move from expression microarrays to whole-exome sequencing, always in the search of disease-causing markers that can be quickly implemented in a clinical setting.
For Campbell, though, no one tool can lead to the handful of genes that will make a difference in diagnostics, prognostics, or therapeutics for cancer patients. Instead, he and colleagues have turned to an integrative genomics approach of overlaying gene expression and copy number data, as well as the outcome of functional screens, to narrow in on those disease-driving genes.
Campbell spoke about his methods at the joint Human Genome Meeting and International Congress of Genetics, held last month in Singapore. Among other topics, Campbell discussed the use of genomic technologies, including Affymetrix's molecular inversion probe, or MiP, assay, commercialized as a service called OncoScan, to discover early biomarkers for breast and ovarian cancer in formalin-fixed, paraffin-embedded samples. BioArray News spoke with Campbell after his talk. Below is an edited transcript of that interview.
Can you give me an overview of what you do and how it relates to the broader purpose of Peter Mac?
Peter Mac is a public hospital, and it is the only cancer-only hospital in Australia. There is a big clinical side to what we do, but we also have the biggest cancer research group in Australia. Lots of institutes have groups that work on cancer, but Peter Mac's research vision is to only work on cancer, and we have about 450 scientists doing that. We cross a large spectrum of diseases and themes — cancer immunology, genomics, genetics, cell biology. The general theme of Peter Mac is that we try to, as much as possible, link in with the clinic, so we have a very close relationship with clinicians. That gives us a more direct interaction with clinical problems and you also get clinicians in the lab. I think that makes our work more focused in terms of what is going to have a real outcome. The aim of my lab is looking for new cancer genes. That is based [on] the genomics of breast cancer and the genomics of ovarian cancer. About half of my lab is focused on somatic genomics — the changes that occur within the tumor. But we are also using exome sequencing to look at lineal cancer genes ... in the germ line. Both of those arms require genomics and SNP arrays and so forth. The aim is to find new genes that will lead to new prognostics and new treatments, or biomarkers to predict which lesions are going to progress. And, on the familial cancer side, we've spent five years looking for new BRCA1-type genes.
Does it happen often that you are identifying a marker via SNP arrays or sequencing that will have a clinical impact?
A lot of Peter Mac's research is translational, but it's still a long way from making it into the clinic. We use primary cancer samples, so we are linking with clinical samples. In my own lab, the closest link I have with the clinic is on the familial cancer side because we are looking for new cancer genes. Once we have identified new cancer genes from whole-exome sequencing, that information can be relatively quickly implemented into the clinic, because a certain gene may predispose a family to breast cancer. The clinician can then test for that gene, and then can advise the family members on their risks. From the genetic predisposition side, from discovery to clinical implementation, that process may take one to three years.
The other side — somatic genetics, looking at tumors — is more long-term. In ovarian cancer, we are looking for genes that are mutated or amplified or deleted. In one study, we used the SNP array to look for regions that are amplified in ovarian cancers [to try to] identify what genes to target, and ... to get drugs that target genes within that amplicon. Those, efforts, looking at new therapies, are probably about three to five years from the clinic, if everything goes fine.
In breast cancer, we are using a lot of MiP arrays. There, we are looking at tumors, ductal carcinoma in situ, in particular, where 20 percent of patients will progress, but the clinicians don't know which 20 percent, so they treat 100 percent of them as if they are going to have invasive cancer. There is a massive overtreatment of DCIS, so we are profiling those to see if there are biomarkers that are strongly associated with whether it progresses or doesn't progress. That could be something where a clinician provides us with a sample, and we would be able to determine if that patient's cancer would progress. The clinician would then be able to modify their treatment of the patent according to that risk. Again, that translation is, if everything goes well, probably about three to five years from implementation.
You discussed an 'integrative genomics' approach in your presentation. What does that term mean exactly?
I guess the place to start would be expression arrays. It was about a decade ago when the first expression arrays came out and people were thinking, "expression arrays are the answer to everything. We'll just run an array and it will show us all of the genes that are important for cancer." It soon turned out that it would be a lot more complicated than that because there were thousands of genes that were overexpressed in the cancer itself. How to know which ones were actually driving it? Then we moved on to the high-resolution and high-density SNP arrays and array comparative genomic hybridization came out. We thought that expression was a loose tool, because all different things can change the expression of a tumor; even sitting in some buffer for a few hours could change it. We thought that a genomic copy number change would be more stable and that it would be more likely to show you which genes are involved. That approach is useful, but you still have a massive number of genes that are within amplicons and deletions. Then we thought to integrate the two, looking at genes within amplicons and genes that are overexpressed. You might have 100 genes in a particular amplicon, but how many of those genes are actually overexpressed? There might be only four genes, so they probably are the ones that are the target of that amplification. But even if you overlay that data, you might still get a few hundred genes that are in regions of amplification and are overexpressed, but they are not necessarily all driver genes. The third thing that we are doing is using a functional screen. We get cell lines that have those amplicons in them, knock them down using siRNA techniques, and see if it has some functional effect on the cell line, or something else. That is currently what we are doing at Peter Mac as an integrative approach — expression, copy number, and then functional screens. Hopefully from that, you can get down to the half-dozen genes that are the real drivers.
Where does sequencing fit into that?
Well, these days you would probably do RNAseq rather than run the expression arrays. But then you would still have to overlay the copy number information and do the functional screening. I guess the other part of next-generation sequencing is looking for somatic mutations in genes. You can sequence the DNA to see if all of those genes actually have activated mutations or inactivated mutations. That would be another overlay. You try to bring all of that information together because any one of those systems, or methods, will provide you with thousands of candidate genes. By combining them all, the hypothesis is that you will be able to drill down to the half-dozen critical genes that are important for that cancer.
We have had two applications of exome sequencing. One was for somatic changes in the tumor. There we are looking for genes that drive the cancer based on the presence of activating or inactivating mutations. We found a few new genes that are important in some types of ovarian cancer. The second side of exome sequencing is the familial side. There, we are sequencing the germ line DNA to look for genes that have clear activating mutations that might explain why a family has a higher occurrence of breast or ovarian cancers. We haven't used next-generation sequencing for things like copy number or translocation. That is still a lot more expensive when compared to using an array.
What kinds of SNP arrays are you using?
We are currently using [Affymetrix's GeneChip SNP 6.0]. We were using the 500K, but as soon as the SNP 6 arrived we adopted it because the price was about the same and the resolution was obviously much better.
And you also have some experience with OncoScan, as you mentioned in your talk.
A lot of our studies for gene discovery have used fresh frozen tumors, good quality DNA. But other studies have tried to identify the [protein-coding genes] behind some of the breast cancers and ovarian cancers, and to identify biomarkers that could predict whether DCIS would stay as a benign disease or progress to invasive disease. The issues there are that the samples are very small, so you get a small amount of DNA, and also that they are formalin-fixed, paraffin-embedded, so the quality is really poor. You can use SNP 6.0, but your results will vary. A lot of these samples are so small, that it might take your [research assistant] half a day to dissect from 40 sections enough DNA to run one chip, so you want to make sure that that sample is going to work. That is why we have been using the MiP array because, in our experience, the vast majority of samples will yield good data. The idea there though is not to identify specific genes but to use it as a fingerprinting tool to look at how the tumor develops.
When it comes to SNP arrays, cancer researchers have many different options. What do you recommend in terms of array technology?
If you are looking at SNP arrays that will give you copy number variation as well as SNP information, there are four or five companies that sell them with various prices. We have stuck with Affymetrix because we know them, we know their arrays perform well, and we have the facilities and know-how to analyze it. But it depends on what your purpose is. If you are looking at gene discovery, and you want really high resolution, then you could use something like CytoScan. That's higher definition, higher resolution than the SNP 6.0, but it also has extra probes for genes that are commonly mutated or altered in various diseases. For other purposes, you might be looking at the relationship between cancers. You are using it as a fingerprinting tool. In those cases, the old 250K will tell you if one tumor is similar to another tumor. You don't need high resolution. You just want to see if they match or not. So, it does really depend on the application. If the DNA quality is high, there is a whole stack of suppliers that all work just fine. The Affy SNP array as a high-quality array performs no better, no worse than some of the other platforms.
You are listed as a co-author on many large genome-wide association studies. What role does Peter Mac play in those efforts?
Predominately, we are the source of samples. I am part of an ovarian cancer consortium, that's a big GWAS, and I am also part of a breast cancer GWAS. We have meetings twice a year. I am involved intellectually because I attend those meetings, but our main contribution is supplying samples and the information from these samples. Moving forward, from our exome sequencing of families with breast cancer, there is a next phase of these studies where we are looking at rare variants. A lot of us are doing exome sequencing of breast cancer in families, and we are identifying these variants. Now there is another round of GWAS where people are inputting the variants they have identified in all of these studies. That is going to be done in 100,000 cases and controls of breast cancer. The idea there is that some of these rare variants will turn out to be breast cancer predisposing. So, my involvement going forward will be inputting these rare variants that we have discovered in various studies and supplying samples.
In the US, there is concern about funding for research. Are there similar concerns in Australia?
In terms of the funding, about a decade ago, a lot of Australian groups were getting funding from the US, though they were also getting funding from the federal government. What has happened since is that the US funding has nosedived, but the Australian funding has remained the same. In terms of funding from the federal government in Australia, we are now getting about a 20 percent success rate, which is not great, but the US was a 30 or 40 percent success rate, but they are now down to between a 5 and 10 percent success rate. Funding is always an issue. We are always putting in three times the number of grants. If we put in four, maybe we'll get one. It's a bit of a lottery sometimes, because it is so competitive. But, overall, we haven't seen a huge fluctuation in the funding that has been seen in some other countries, particularly the US.