Name: Jacqueline Schoumans
Title: Head of Cancer Cytogenetic Unit, University of Lausanne
Background: 2009-present, head of Cancer Cytogenetic Unit and associate professor of medical genetics, University of Lausanne, Switzerland; 2005-2009, professor of medical genetics, Karolinska Institute, Stockholm, Sweden; 1999-2005, cytogeneticist, Karolinska Institute; 1995-1999, head of the cytogenetics laboratory, University Hospital Bergen
NUREMBERG, Germany — With a background in cytogenetics stretching back nearly two decades, Jacqueline Schoumans is well acquainted with using array comparative genomic hybridization to identify the genomic aberrations underlying congenital diseases. When she moved to the University of Lausanne at the end of 2009, though, she took over a lab that was exclusively focused on cancer cytogenetics, typically relying on conventional microscope-based karyotyping or fluorescence in situ hybridization.
A desire to implement array CGH in the cancer cytogenetics setting prompted a lengthy evaluation of different microarrays, and Schoumans and other European cytogeneticists earlier this year published a review of the platforms available to them (BAN 6/12/2012).
After assessing the various offerings on the market, Schoumans opted to work with Oxford Gene Technology to design her own cancer-focused chip, an array that OGT now plans to commercialize. Her lab now uses arrays exclusively to test chronic lymphocytic leukemia cases, though she claims the array can be used for testing other cancer samples as well.
At the European Society of Human Genetics' annual meeting, held here this week, Schoumans discussed the process of designing and validating an array for cancer cytogenetics. BioArray News spoke with her after her presentation. Below is an edited transcript of that interview.
When did you become interested in using chromosomal arrays in cancer testing?
When I moved to Lausanne, I took over a lab that is only doing cytogenetics on cancer samples … and with my experience of introducing arrays into the constitutional field at Karolinska, I also got interested in doing the same for cancer. We are trying to answer the same kinds of questions, basically, but [in cancer cytogenetics] we are a couple of years behind having everything in place, [in terms of] being able to interpret what [certain results] mean and so on. But I do believe that when there is a reimbursement system for [array-based cancer cytogenetic testing] … it will [be adopted] at a faster pace than constitutional CGH has been.
Why did constitutional move faster in the beginning versus cancer? Was it easier to implement?
[In constitutional cytogenetics], there are less complex abnormalities in the genome, so it is easier to interpret. You look for one or two aberrations in constitutional, whereas in cancer, you have a whole scale to consider, so it takes a longer time to understand what [your results] mean, because you want to find an early event in cancer. You want to understand what is recurrent in certain diseases and, in the long run, you want to look for treatments. There are certain treatments that can be implemented but they are not ready yet. The famous success story is [Novartis's Gleevec (imatinib)] for chronic myeloid leukemia, where … patients who have the Philadelphia chromosome resulting from a t(9;22) [translocation] … are successfully treated with [the tyrosine kinase inhibitor] and can survive the disease, but that took 50 years between finding the recurrent chromosome aberration specific for this type of leukemia and [being able] to offer a treatment.
After you made the decision to introduce arrays into the cancer setting, you embarked on an evaluation process.
Part of the experience was from [implementing] constitutional testing at Karolinska, looking at the 385K from Roche NimbleGen and the Affymetrix 250K. So from this experience we went further. There are many different commercially available arrays, so we wanted to choose the one best fitted for our goals. That is why we tested different platforms. Our experience with CGH in constitutional was very positive, but it was also positive when we started to use it for looking at cancer samples. What was lacking was the SNPs. So when the SNPs were put on the CGH arrays, that's when we looked at OGT and the other platforms as well, like Agilent and BlueGnome, that have SNPs on their CGH chips. We tested positive controls, so it wasn't a huge project. We had samples where we had already defined clear copy number variants and regions of loss of heterozygosity and we used the samples to evaluate all the platforms to see if the [CNVs and LOH] would be detected by all of them.
What conclusion did you come to?
We came to the conclusion that the SNP analysis on the CGH+SNP on OGT worked well, but not on the two other companies' [arrays] in our hands with our samples. We did not get the results that we were expecting and we still could detect very low grade mosaicism with the OGT array. We haven't tested the big cohort of CLLs on the Affymetrix array, but that is based on previous experience, where, when you have an abnormality, the dynamic range between your normal baseline and the abnormality is more visible when you use CGH compared to SNP arrays.
What OGT array did you use?
We did the evaluation on OGT's ISCA+SNP array. But we had planned from the start to optimize our own design, because we had special needs; you look for other genes in cancer than in constitutional diseases. The ISCA design was not convenient for us. But we first tested it on the analysis of LOH, because if the LOH had not worked, we wouldn't have bothered to make our own design. But OGT worked very well, better than the other providers, so that is why we went along with OGT and made our own design.
How did you arrive at the design?
We designed the array on [Agilent Technologies'] 4x180K platform, with coverage of 1,500 genes for each array. To obtain the content, we went through all databases we could find with cancer genes. We hardly excluded anything. For each gene, we have coverage of one probe per exon, that's not very dense, but from different publications, we knew of other genes. So for 17 selected genes, we made even denser coverage with four probes per exon, to have the sensitivity to detect single-exon aberrations, especially in genes where very small aberrations are known to give a poor prognosis. These kinds of aberrations are often found using other methods like [multiplex ligation-dependent probe amplification] or qPCR. So we wanted our array [to] also replace these kinds of tests. The SNPs were chosen by OGT. They tested a lot of SNPs that are found very frequently in the general population, and whether they hybridize well, and their location. The design was optimized for all cancers, but we implemented for diagnostics in CLL, because that is the easiest cancer to interpret, because it is less complex. We plan to start at the end of the year to apply it for multiple myeloma.
What is the main difference between OGT and BlueGnome and Agilent's CGH+SNP arrays?
The key difference is that for the other two, you need to digest the DNA first, because it will cut at the SNP or not cut at the SNP. That is how it detects the genotype. The disadvantage with that is that you cannot use whatever reference you want, you have to have the genotype reference where you know exactly what SNP it has, whether it will cut or not cut. So Agilent and BlueGnome provide a known reference now, but you have to stick to their reference. This is optimized for SNP genotyping, but it is not optimized for CNVs, and it is not optimized for cancer because you do not know what kind of normal variants an individual has. And for cancer we would rather have the matched reference of the same patient who has the cancer. That is the advantage [of OGT's approach], that you are free in terms of what reference you use. The other thing is that we have run into problems with the other platforms when we were running more challenging samples, like multiple myelomas and solid tumors from formalin-fixed, paraffin-embedded samples. So then we could not get good results either.
What are the advantages of CGH+SNP arrays over high-resolution SNP arrays, such as Affymetrix's CytoScan HD?
The major advantage of the OGT array over the Cytoscan HD from Affymetrix is that it enables one to use a matched reference from the same patient in the same hybridization. This enables the detection of the acquired abnormalities, copy number aberrations as well as copy neutral loss of heterozygosity, and allows [one] to distinguish it from benign germline aberrations. This is very important information when interpreting the results. With all other platforms, two hybridizations would be needed to get this result, which doubles the costs and the workload.
You said during the presentation that you have replaced karyotyping and FISH with arrays.
We did it this spring. Since last summer we started to run the tests [in] parallel. We did 62 cases of CLL by array and by FISH to be sure that we pick up with arrays what we need to pick up. Since this spring, we have launched it in our clinic and we will systematically perform an array unless a clinician wants another test.
Is it really cheaper to do arrays versus FISH or conventional karyotyping? Arrays have this image of being an expensive technology.
I think the price is overestimated and performance is underestimated for detecting low-grade mosaicism. Arrays are cheaper if you count the labor. Culturing cells for conventional karyotyping, and performing multiple interphase FISH analysis is far more time consuming compared to one single array analysis. We actually had already abandoned karyotyping for CLL. It is not yet possible to use arrays for all hematological malignancies since it does not enable detection of the balanced recurrent translocations. It will thus not fit for all leukemias as well as it does for CLL. But in the case of CLL, we gain a lot of time. From an analysis point of view, it does not take more time to analyze an array compared to all the FISH we were doing, but the interpretation is a little more challenging. So overall, I think we are quicker doing one array compared to the six FISH tests we ran on each sample before. So our array is replacing seven tests [six FISH tests and conventional karyotyping] at the moment.
Are there other labs in Switzerland, and beyond that, in Europe, that are replacing FISH and conventional karyotyping with arrays for cancer testing?
In Switzerland, we are the first to offer this systematically on CLLs. There is another lab in Basel that does it on demand, so they have not changed their routine. In Nijmegen in Holland they use arrays on ALLs, and they published an article last year on their workflow. So arrays are being used in more labs, but the problem is often reimbursement. In Switzerland, these tests are being reimbursed, but I know German labs that would like to do arrays but they have a harder time getting it reimbursed. In Holland, there is reimbursement. In the US it is already used by private labs. So it depends on the country's healthcare system.
Do you think that all labs should design their own arrays for cancer cytogenetics like you have?
That would not be necessary. We designed this array because not many arrays have been designed specifically to use in the onco-hematological diagnostic setting or for solid cancers. Therefore, none of the commercially available arrays fit our goal. Any lab is welcome to use our design if it serves their purpose.