St. Jude Children's Research Hospital announced that it is planning to sequence the whole genomes of 600 pediatric tumors in collaboration with Washington University's Genome Center. GT's Ciara Curtin caught up with St. Jude's scientific director James Downing to talk about the project, which he says is "going fantastic."
Genome Technology: How did the idea for this project come about?
James Downing: We had done a lot genome-wide analysis of a range of pediatric cancers in the last four to five years. A number of investigators here at St. Jude had been involved in this work and as we started analyzing that data and making discoveries, we kept a very close eye on sequencing technologies, and on the national and international efforts to characterize cancer genomes. As we analyzed that data, the technology, and the activities of others, it became clear to us that there was real need for a focused effort to characterize the genomes at the sequence level of pediatric cancers.
GT: How did you choose which cancers to focus on?
JD: We're focusing on the major pediatric cancers, so the broad categories: the hematopoietic malignancies, the solid tumors, and the brain tumors. And within our cancer center, we tasked the disease-specific programs to decide which tumors should be sequenced.
The goal is 600 pediatric tumors to be sequenced in a three-year period with matched germline samples for each tumor, so 1,200 total human genomes. Each program brought forward their list and then we had a steering committee, composed of the program leaders. And we then debated across programs, really challenged the programs: why this tumor and not some other tumor within your program? That iterative process then generated the final list and we broke that final list into phases.
We're planning to sequence 50 tumor whole genomes the first year and 50 matched germline, so 100 total whole genomes. We thought the first year we would equally split that between the three tumor types — that we would look at small subsets of tumors within our particular hematological malignancies versus solid tumors and we would analyze data as it comes off to make decisions on how to move forward for years two and three, when the bulk of the samples will be analyzed.
An example is in the hematological malignancies [where] we decided to sequence a particularly aggressive form of leukemia that we see in infants, acute lymphoblastic leukemia. We'll sequence a number of those cases the first year. We'll look at that data and we'll see what that tells us about that particular tumor: are there lots of mutations, are there recurrent mutations? For example, we may only sequence 10 infant leukemias the first year, but we'll have a cohort of about 100 infant leukemias. Any mutation that we detect in those 10, we can immediately see: does it occur at a higher [frequency]? At what frequency in the 100 infant leukemias? So, recurrence determination. But if we find that there are very few mutations in infant leukemia and very few recurrences, it's unlikely that we would them move forward to sequence 200 of the case. It would be too low of a yield.
GT: Are the tumors all from St. Jude's repository?
JD: The first year, every sample that we will do the whole genome sequencing on will be from St. Jude's tumor bank. We have a large tumor bank that has been in existence almost from the start of the institution, and we are approaching our 50th year. It's an incredibly valuable resource because the vast majority of our patients are treated on protocols that were designed here and we have great data on exactly which drugs, surgery, and radiation therapy they were treated with. We have near-complete follow up on those patients, and so it's optimal because we have frozen material, we have germline sample, and great clinical outcome. We have lots of correlative studies on drug doses, medication, and other molecular diagnostic tests, et cetera. That'll allow us to try to assess clinical value of mutations that are detected within the samples.
GT: Why did you choose a whole-genome approach over, say, a targeted sequencing approach?
JD: Our feeling is that in the end that's what everybody is going to want, whole-genome sequencing. The exon sequencing is a cost-saving approach that can give you mutations within exons. It has a great bias against picking up structural alterations that led to inversions, re-arrangements, translocations, et cetera. We know [these] can be a prevalent mechanism of mutagenesis within pediatric cancers. We thought, 'We don't want to miss that.' We know from work prior to initiating this program, there are mutations in promoter region of genes that are affecting their function and have the same clinical significance as mutations that delete exons or mutate key functional residues. We want to make sure we catch those and the vast landscape of other regulatory RNAs, not microRNAs, but these large, intervening non-coding RNAs. We think that these are going to play a significant role in the pathogenesis of cancer. By knowing, by just starting and doing whole genome sequencing and acquiring as much data as can be acquired for each individual tumor sample, we'll end up getting a clearer picture of what the total complement of mutations are within pediatric cancers.
GT: Which technology are you using?
JD: We are, again, phasing this in. We are using the Illumina GAIIx machine and purchased a substantial number of machines, placed either at WashU or here at St. Jude — this was before the announcement of the HiSeqs from Illumina. Those machines, the GAIIxs, are a stable platform that are running in production mode and generating the data right now. The HiSeqs are just really hitting the market and being placed into production mode. A second purchase of machines will be made at the beginning of year two. We have flexibility that we can move to another platform if it becomes available, stable, and reliable for this kind of project.
GT: How do you think this project will have an impact on patient care?
JD: At the first level, it'll help us understand and define the heterogeneity of particular pediatric cancers. The experience from these kinds of studies has clearly shown us that cancer is an extremely heterogeneous disease — that there are particular subtypes even within a single type of cancer. If we take acute lymphoblastic leukemia, there are many different genetic subtypes in there, and those genetic subtypes have clinical features: different responses to chemotherapy, different outcomes. This genome project will help us to better understand that heterogeneity to identify particular subtypes, and I think will give us prognostic information that different mutations will help us decide, 'Is the patient a low-risk patient who needs relatively mild chemotherapy, or a high-risk patient that needs more aggressive therapy?' Some mutations will also turn out to be rational targets against which drug therapy can either be used if the drugs are available, or developed if not available. An example, again in acute lymphoblastic leukemia, is that [in] a small subset, but a very high-risk small subset we recently identified, they have mutations in the JAK family of kinases and there are drugs in development that are being used in other cancers against the JAK kinases and now clinical studies will get developed to see if they have any efficacy in this subgroup of pediatric ALL.
I think it we help us to understand heterogeneity, it will help us to identify so-called biomarkers of drug response or prognostic markers and it may help to identify new rational targets. There are going to be lots of mutations identified through this and there's going to be a lot of work that will need to be done to identify which of those mutations are true drivers and play a functional role in the development of the cancer, which mutations are passengers, and which are somewhere in