While cancer research never stays still for long, the past year saw major shifts in the application of genomic tools to the field. Last fall, The Cancer Genome Atlas published initial results of its first programs in glioblastoma multiforme; around the same time, the Tumor Sequencing Project Consortium reported its early findings from an international study of lung adenocarcinoma. On the technology side, next-gen sequencing continues to revolutionize the field as it has radically altered the types of questions scientists can ask of precious tumor samples.
As we considered some of the most innovative work in cancer research during the past 12 months for this article, one thing became very clear: gone are the days that scientists relied on a single type of technology. In the stories that follow, you'll see that researchers are finally able to deploy many tools and gather a wide range of data sets as they attempt to find a chink in cancer's armor. As Dana-Farber's Matthew Meyerson told us, "This is really the beginning of systematic and complete cancer genome characterization."
Pathway analysis: Cancer Atlas Team Turns Up New Data in Glioblastoma
The Cancer Genome Atlas, which aims eventually to catalog all major cancer-causing genomic alterations, published results last year in Nature from its first genome-wide, integrative analysis of glioblastoma multiforme. Performing a range of studies, including DNA copy number, gene expression, DNA methylation, and SNP analysis, on 206 glioblastoma samples — this is the most common type of adult brain cancer and the first cancer studied under TCGA — the consortium of 18 different institutions found a host of new and confirmed genetic changes.
"A specific goal of TCGA [is] to generate a reference data set," says Lynda Chin, co-chair of the committee that wrote the paper. "People tend to downplay that part of it, which I actually think is the most important component."
Matthew Meyerson, also a co-chair, adds, "I think it's fair to say that there's never been in all of cancer genomics anything that is a reference set at all like the TCGA because of its comprehensive nature." Chin and Meyerson are both affiliated with the Dana-Farber Cancer Institute and Harvard Medical School.
After sequencing 601 genes in the GBM samples and matched control tissue, they found three significant genetic mutations that had not previously been reported to be common in GBM: NF1, ERBB2, and PIK3R1, the last a gene that influences activity of an enzyme called PI3 kinase that is deregulated in many cancers. As drug targets, PI3 kinase inhibitors can be affected by whether patients have mutated forms of the genes.
The scientists also found three pathways that were changed in more than three-fourths of the GBM tumors: the CDK/cyclin/CDK inhibitor/RB pathway, involved in the regulation of cell division; the p53 pathway; and the RTK/RAS/PI3K pathway, which is involved in the regulation of growth factor signals. "Within GBM, the dogma is that p53 is not common in primary GBM; rather it is a signature of secondary GBM," Chin says. "What TCGA data now definitely show is that it is not unique to secondary GBM, it is actually a common mutation in primary GBM, [and] that's an example of something that can have important implication on the biology of the disease."
Meyerson says that while there were hints that three major pathways are almost always universally altered in GBM, it is TCGA data that now prove this to be true.
Physicians already know patients with GBM tumors that have an inactivated, or methylated, MGMT gene respond better to temozolomide, a chemotherapy drug commonly used to treat GBM. TCGA also found that in patients with MGMT methylation, this type of therapy could lead to mutations in mismatch repair genes. These could cause recurrent tumors with abnormally high numbers of DNA mutations and that may be resistant to chemotherapy treatment. "If I had to name one specific thing that might in the short term have some clinical indication, it would probably be the more clear elucidation of one mechanism of resistance to temozolomide because it is a standard of care today for GBM patients," Chin says. "Every patient eventually develops resistance and [the cancer] recurs."
Meyerson thinks that detailing the PI3 kinase mutations is a major new finding. "The PI3 kinase catalytic subunit gene was already known to be mutated in cancer, but nobody had ever found cancer-causing mutations in the regulatory subunit," he says. "It defines a new class of patients who might benefit from a new kind of targeted cancer therapy."
He adds, "This is really the beginning of systematic and complete cancer genome characterization, rather than the end." In follow-up studies, Meyerson says, the TCGA team plans to sequence for structural variation.
Sequencing: Analysis of Rare Cancer Sparks New Hope in Existing Therapeutics
As the study of common cancers has all but taken the genomics community by storm, Marco Marra's lab at the Genome Sciences Center in the BC Cancer Agency has taken a different tack: focusing on rare cancers and how to target them with therapeutics that were designed for other diseases.
The work began as a collaboration with David Kaplan's lab at the Hospital for Sick Children in Toronto, prompted by former Kaplan lab member Olena Morozova, who joined Marra's lab. "We were heavily invested in this concept that DNA sequencing analysis of cancers would be a good thing," Marra says. Morozova suggested turning the sequencing arsenal on neuroblastoma — a major cancer that affects infants but has seen little research progress in the past few decades, Marra says — after Kaplan's lab was able to derive neuroblastoma tumor-initiating cells.
Kaplan provided the cells, which represented both remission and relapse states from one patient, and Marra's group generated "something like a flow cell's worth of data," he says. Morozova dug in on the analysis, finding that the cells expressed markers indicative of a neurological lineage, and that there was also an abundance of B cell markers. Drilling down into the transcriptomic data, Morozova performed a clustering analysis and confirmed "that the cells express B cell markers," Marra says.
Marra, who suspected that this was "some sort of technical artifact as opposed to biological reality," waited for Kaplan to go back to the original cell cultures to validate the findings — and only when that happened did he realize they were on to something. "The exciting thing about that is that these cells were shown to express CD20," Marra says — which happens to be the target of the drug retuximab, creating the possibility that an existing drug could be repurposed to treat a disease that otherwise might never be targeted by a pharma.
Granted, all of this work was done based on a single patient, but Marra says that's part of why this is so encouraging. "Because the technology is so powerful, even an analysis of N=1 can yield an entirely new direction of research," he says. For something like neuroblastoma, he says, clinicians may have "the opportunity to reposition a drug from one sort of cancer to another sort of cancer and make an impact on a field that hasn't seen progress in 20 years."
Marra's team is now running with this approach for other rare cancers, including a series of ovarian cancers, granulosis cell tumors, and a variety of cancers that have no formal definition. The group will be looking at sequencing and expression data, as well as other information, to help figure out whether the molecular content of these cancers "[positions] them or … makes possible their treatment -using things we already have on the shelf," he says.
Nanotechnology: Barcode Arrays for Cancer Diagnosis
As anyone who has recently been to the doctor can attest, getting blood work done is not fun — -often it's a few milliliters taken from your arm — and the turn-around time is sluggish. Rong Fan, a third-year postdoc in James Heath's lab at Caltech, has been working on developing a less invasive, faster barcode chip to survey a large panel of proteins found in blood, with an eye toward diagnosing cancer. "We really want to develop nanotechnology to achieve [a] systems biology approach to diagnose cancer," Fan says.
This integrated blood barcode chip has two functional modules. "Upstream, we have a blood separation module," Fan says. "Downstream, we have a protein barcode, just like a protein microarray, but attached in a barcode fashion."
The chip only needs a finger-prick's worth of blood, and the red blood cells and white blood cells get separated out in the upstream module. This sorting occurs due to the Zweifach-Fung effect, which says that where capillaries branch, blood cells will take the path with the higher flow rate. This effect is seen in capillaries of the arm, Fan says, adding that his team is emulating that phenomenon to flow pure plasma into the downstream portion of the chip.
To create the barcode, Fan and his colleagues patterned single-stranded DNA onto the barcode array. Here, they can pause until it's time to use the chip without worrying about the antibodies denaturing. "We made the device, we can put it on a shelf to store for a few months, no problem. Only when we want use the device, we flow DNA-labeled detection capture antibodies on the barcode area and transform the DNA microarray into an antibody microarray," Fan says. If a plasma protein binds a capture antibody, it is detected through a fluorescent probe.
As they report in a Nature Biotechnology paper, Fan and his colleagues tested the barcode chip by seeing how it could detect human chorionic gonadotropin, a hormone used for pregnancy testing, over a range of concentrations in serum and by assessing serum from breast and prostate cancer patients to see if a panel of markers could differentiate between the cancers. "It demonstrates that this platform really works," Fan says.
The chip is now being tested in clinical trials of glioblastoma. Fan says that so far they have looked at 20 patients, but that they will need a lot more, as well as markers specific for glioblastoma. He adds that a collaborator, the Institute for Systems Biology's Lee Hood, is working on finding brain-specific markers for them to include on this panel.
Disease stratification: ALL Research Highlights Poor Responder Subtype
A large-scale effort spearheaded by the Children's Oncology Group study, composed of researchers from St. Jude Children's Research Hospital, the University of New Mexico Cancer Research and Treatment Center, and the National Cancer Institute, may have struck a crucial blow in the fight against acute lymphoblastic leukemia. Recently researchers announced the identification of mutations in the Ikaros gene that play a role in ALL relapse and that may point the way toward more accurate diagnostic tests.
According to the NCI, this form of cancer is particularly grim because, while cure rates are higher than 80 percent using current treatment methods, only 30 percent of children who experience a relapse live longer than five years. Developing a marker to determine the risk of relapse would greatly assist physicians in customizing individual therapies in order to minimize the recurrence of cancer in children.
"The most significant finding in the study was the identification of a new subtype of ALL that is associated with very poor outcome and that can be identified by having alterations in the Ikaros gene, either by deletions or mutations," says Malcolm Smith, association branch chief for the Pediatrics Cancer Therapy Evaluation Program at NCI. "This new subtype also has a gene expression profile that is very similar to another type of ALL that is associated with poor outcome — BCR-ABL fusion gene ALL or Philadelphia chromosome positive ALL. … There's a linkage between the Ikaros 1 deletion, which is observed both in this new subtype that we identified, and it's also observed in BCR-ABL ALL."
This research was done under the umbrella of NCI's Therapeutically Applicable Research to Generate Effective Treatments initiative, which is geared toward developing effective therapeutic targets for the treatment of childhood cancers. Smith says the biggest hurdle facing a study of this nature may be the identification of the appropriate clinical specimens. "We had a large number of leukemia specimens from uniformly treated patients who were at high risk of treatment failure," he says. "But without that large specimen collection, the discovery couldn't have been made, so it highlights the importance of the tissue bank resources that allow this type of work to be done."
In terms of getting these findings to mean something at the bedside, the researchers are cautiously optimistic, but there is still much work to be done. "For the future, this new subtype might be identified through looking at genomic changes in Ikaros or it could be identified by detecting other biological characteristics that are associated with the Ikaros 1 deletion," Smith says. "So further characterization of a new subtype of Ikaros 1 deletion will be required before it's clear what the best way of clinically identifying this new subtype is."
From his vantage point, Smith says the study is not only notable because of the potential impact of its findings, but also because of its multi-tiered genomics approach. "It is certainly one of the first studies that has integrated the detailed gene expression data, the gene characterization data, the genomic characterization data with the DNA sequencing on a large set of clinical experiments," he says. "I think each of those characteristics contributed to the discovery that was made."
Going forward, he and his colleagues all seem to be in agreement that the TARGET initiative can help break down doors for other types of cancer using the same large-scale approach used for ALL. "The TARGET initiative applies to childhood cancers, tumor specimens or leukemia specimens, integrated genomic analysis, gene expression analysis, and gene sequencing," Smith says. "We have applied this approach to neuroblastoma, another childhood cancer, and we'll be seeking to apply it to other childhood cancers in the near future."
Genomic Rearrangement: At Brown, Raphael Studies the Geometry of Genomes
While waiting for the next-gen sequencing data to come out, Brown University computational biologist Benjamin Raphael has been working with older, paired-end sequencing data from BAC clones to develop new methods to suss out genomic rearrangements — particularly fusion genes — associated with cancer.
To better assess genome rearrangements from paired end data, Raphael and his colleagues developed a new approach. This geometric method scores breakpoints in the genome, where fusion genes often arise, to prioritize candidates for follow-up study. "It's really this two-dimensional geometry problem," Raphael says. "Each end [of the paired ends] gives you a coordinate of a point in space — then you can define where the breakpoints are as a polygon in two-dimensional space, and then you look at the intersections of the polygon and they give you where the breakpoints are. You can overlay genes and decide how confident you are that there's a fusion gene, given the data that you have."
Though they developed this method as a way to prioritize which fusion genes to follow up, Raphael and his colleagues realized that it could be used to compare structural variations. "If you had two different sequences that were purporting to report the same variant, you could, very carefully, write down exactly where the mapped reads were and what you know about the lengths of the reads — they give you these nice polygons — and then just see whether or not they intersect," Raphael says.
This geometric approach could also be used for data coming from next-gen sequencers — what Raphael's been waiting for — though the smaller the fragment sizes are, more will be needed to cover the genome. "What we tried to do is develop this geometrical approach, to plug in any parameters you want for whatever sequencing technology. If you know what the fragment size is, all it does is change the size of the polygons," Raphael says.
Recently, the researchers have also been looking at CGH data to see if they can also identify fusion genes and conserved breakpoints. The scientists are developing a Bayesian chain algorithm to segment the CGH data, output all the ways that the data can be segmented, and determine how many of those ways would have a breakpoint in a particular gene. "We can then use that score across samples and try to see whether or not we have a conserved breakpoint in the given gene," Raphael says. They tested this method by looking for a prostate cancer gene known to be caused by a 3-megabase deletion that results in a fusion gene, finding that breakpoint accurately across the patients.
Raphael and his lab are also currently running this algorithm on candidate breakpoints for glioblastoma-using data from The Cancer Genome Atlas and appear to be finding highly conserved breakpoints.
MicroRNAs: ISB/Hutch Group Finds Potential in New Biomarker Class
The hunt for accurate, easily detectable cancer biomarkers is happening on all fronts — progress has been made in epigenomic, proteomic, and other markers. But recent work from the Institute for Systems Biology and the Fred Hutchinson Cancer Research Center has given a serious boost to the idea of using microRNAs for this task.
The most important finding of recent research into the use of miRNAs for cancer biomarkers, says Muneesh Tewari at ISB, "is that microRNAs in circulation are extremely stable, which is an obvious prerequisite for using them as markers." In his collaboration with scientists at the Hutch, they discovered that miRNAs from a host of tumors could be found in blood, and that even epithelial cancers and solid tumors gave off enough cells that "micro-RNAs reached blood in levels that were detectable," Tewari adds.
While the research is still in fairly early stages, Tewari is encouraged that miRNAs could be "potentially a new class of biomarkers" and that they have some benefits over other kinds of markers. For instance, he says, "A lot of biomarkers are just that — just markers, and it's not clear whether or not they have any functional role." Because miRNAs are known to be regulatory, he notes, the assumption is that they're more likely to be causative than simply a passive marker. In another advantage, Tewari says, conventional wisdom suggests that miRNAs will be far easier to detect in sera as compared to, say, a protein marker, since they're nucleic acids and can be amplified.
Next up for Tewari and his team will be trying to get a handle on the biology behind miRNA markers. One question they have is why miRNAs appear to be so stable: "[We're] trying to understand what's protecting these biomarkers in the blood," he says. "We don't still have a complete answer for that."
In addition, his team will work on developing assays with better sensitivity — Tewari says he expects that current assays can be improved by a couple of orders of magnitude. Also, the scientists will have to get a better grasp of the clinical meaning of miRNAs in the blood. "We don't really know whether microRNAs are elevated in very early stages of the disease," he says. Near term, he adds, even finding a biomarker that would say something useful in an advanced case of cancer would be helpful, but longer term they'll be hoping miRNAs offer a chance to detect cancer in its earliest phases.
Metabolomics: Study Points to Sarcosine As Possible Marker for Prostate Cancer
Prostate cancer is the most frequently diagnosed cancer in men, and being able to find it early can help determine which patients have a more -aggressive form that might progress and spread. To that end, University of Michigan researcher Arul Chinnaiyan used a combination of high-throughput LC-MS and GC-MS to survey the metabolome in prostate cancer patients to find specific metabolites that could indicate likelihood of disease progression.
In work published in Nature this February, Chinnaiyan's team screened more than 1,126 metabolites across 262 samples in tissue from healthy prostate, clinically localized prostate cancer, and metastatic prostate cancer. They found a subset of six metabolites — including sarcosine, uracil, kynurenine, glycerol-3-phosphate, leucine, and proline — to be significantly increased in progression from healthy to clinically localized prostate cancer to metastatic prostate cancer. "What this study does is it begins to now monitor the metabolites, or the small molecules, in [an] unbiased, profiling-based manner so we can begin to approach having a more complete systems view of the different molecular alterations that can occur during cancer progression," Chinnaiyan says. "Essentially, this is one of the first studies that really looks at [prostate] cancer progression, at the metabolites that are altered."
He chose to focus on sarcosine because it was elevated in clinically localized disease and highly elevated in metastatic cancer. "It also mapped to different pathways that seemed to, in general, be elevated in prostate cancer progression, including amino acid metabolism [and] methylation," Chinnaiyan says. Because they also wanted to monitor those metabolites noninvasively — any test needs to be able to find these markers in blood or urine — they took urine samples from patients who were either needle biopsy-positive or biopsy-negative, "and lo and behold, we were able to see sarcosine elevation in general in those patients who were biopsy-positive," he says. Even a negative needle biopsy does not rule out the presence of cancer, Chinnaiyan says.
Using RNAi, the team was also able to show noteworthy in vitro effects of this molecule on cancer progression. By adding exogenous sarcosine or knocking down the enzyme that leads to degradation of sarcosine, the researchers caused benign prostate cells to become cancerous. Knocking down the enzyme that makes sarcosine from glycine blocked cancerous invasion.
Future studies will determine whether the marker could have diagnostic relevance. It is more likely, Chinnaiyan says, that a panel will be needed for lab-approved tests. "At the moment it's pretty early days," he says. "Our working hypothesis is that it would be a marker of more aggressive prostate cancer, but we need to do subsequent studies to show that." Validation will take at least five years and will have to be done with a panel of additional markers, not just sarcosine alone. Chinnaiyan envisions that "when somebody comes in [to the clinic] with an elevated PSA, the idea would be that they would get a sarcosine-like test to see if they have the disease, but more importantly if they have the more aggressive disease."