By John S. MacNeil and Meredith W. Salisbury
As the second leading cause of death in the United States, small wonder that cancer is also at the top of the list of many institutes’ funding priorities — or that research in the field continues to grow at an extraordinary rate. In response, the genomics world has shown the tremendous range of systems biology by taking virtually every tool at its disposal and aiming it at defeating this disease.
The National Cancer Institute alone had a budget of $4.7 billion for fiscal year 2004, and of that, $3.7 billion was earmarked to fund extramural research. Two of the institute’s categories — human genome research, with almost $267 million, and gene mapping, with $319.3 million — show that this field is a major cash cow for the genomics community.
From proteomics to bioinformatics, gene expression to RNAi, cancer-related research spans the systems biology realm — to the point where many tools vendors now consider cancer science among their largest markets. That status gives researchers in this field unprecedented clout in demanding, and getting, the tools they need.
The critical mass of this community means that “breakthroughs” are now a regular occurrence, and “cutting edge” is a fast-moving target. It also means that it’s a field rife with fascinating people and great research stories, and that’s the point of this article.
In the pages that follow, you’ll find a series of profiles highlighting scientists and the innovations that make them worth reading about. We’ve tried to give you a representative sampling with as many different technologies as possible and people from both the public and private sectors. In addition, we went to directors of some of the major cancer centers, asking what they see as the most promising research path in the battle against cancer. You’ll find their answers inside as well.
Innovation: High-throughput fingerprinting to find genomic rearrangements
British Columbia Cancer Agency
With a background in fingerprinting and an institute mandate to help people with cancer, it seems only natural that research scientist Martin Krzywinski and the rest of the BC Cancer Agency team led by Marco Marra would apply their high-throughput fingerprinting technology to studying lymphomas, a form of cancer known to be associated with recurring rearrangements.
The group completed a small pilot study of this technology in collaboration with scientists at the University of California, San Francisco, and is now moving on to a larger pilot-study with lymphomas. “The hope is to find cryptic rearrangements that have so far gone undiscovered,” Krzywinski says. With today’s methods, such rearrangements are detected by examining karyotypes — a fairly low-resolution, and not terribly accurate, process. The high-throughput fingerprinting technique could improve resolution from the 10 megabase order seen with karyotypes down to about 100 kilobases, Krzywinski says.
The technology to find rearrangements isn’t all that different from regular DNA fingerprinting. Krzywinski and his colleagues start with tumor DNA, create a BAC library containing sections 100 kilobases to 150 kilobases in length, expose the BACs to a restriction enzyme, and then generate a restriction fingerprint from that. The entire genome sequence is digested and run on a gel, and then the team looks for places where a particular BAC might have come from based on size signatures (rather than aligning actual base pairs) from the restriction fingerprinting.
The technology is particularly good for finding genomic rearrangements — and, because it samples the entire BAC instead of just the ends, which happens in the more common BAC-end sequencing technique, “can potentially detect more complex rearrangements,” Krzywinski says.
He adds that the lab has geared up for doing this in a high-throughput way: scientists there generate about 15,000 clones each week and finish on the order of 1 million fingerprints per year. Still, the process is somewhat labor-intensive, so Krzywinski says, “We’re trying to … automate the process even further to try to drive the cost down.”
Krzywinski and the rest of the group are already well into the project, he says. They’re working off of two libraries made from lymphoma samples, and after starting the fingerprinting process in the third quarter of last year, the crew is halfway done with the first library, he adds.
Innovation: Bioinformatics tool combining statistical work with clinical findings
A simple enough change of scenery for bioinformatics firm Incogen proved to be the first step in kicking off what CEO Maciek Sasinowski would later call the company’s “flagship project.”
It began about three years ago, a few months after the firm moved from South Carolina to Williamsburg, Va., when researchers from Eastern Virginia Medical School approached the Incogen team about some trouble they were having analyzing their mass spec data. They were using SELDI mass spectrometry to profile prostate cancer biomarkers, Sasinowski says, “but they realized they did not have the expertise to analyze [the results] in house.” That analysis included the goal of being able to compare data from many disease and normal samples to determine a representative signature which could be used as a cancer diagnostic for new samples coming in.
Part of the problem, Sasinowski says, was sheer volume of data. “The amount of data you get from just one single laser shot [in a mass spec] is massive.” Multiply that by the numbers of samples needed to mine for signature patterns, and the profile was unwieldy.
In the partnership that has ensued, Eastern Virginia has continued to handle the clinical work — expanding to leukemia as well — while researchers at the College of William & Mary have taken on the statistical and experimental analysis, Sasinowski says. “It became obvious that it was really important to provide a software package that would allow these different groups to exchange information,” he adds — and so Incogen’s bioinformatics team has become “the glue in the middle.”
The Incogen staff started out with VIBE, their sequence analysis tool that allows for graphical drag-and-drop pipeline building. Thanks to seed money from Virginia and subsequent phase I and II SBIR grants through NIH (worth $200,000 and $2 million, respectively), this collaborative project has turned into a mainstay at Incogen and led to VIBE MS, a modular mass spec version of the software tool that handles protein analysis tools such as signal processing, background subtraction, and variable selection.
The tool is still in beta, and was scheduled for its 0.2 release last month. Incogen collaborators are using the tool, and Sasinowski says it’s also been released to members of caProteo, a proteomics unit of NCI’s caBIG initiative. With feedback from those users, he anticipates that a final product could be on the market by the end of the year. He adds that his team is starting to look at adding microarray, protein array, and sequence data to the mass spec data and “allow people to combine all of this heterogeneous data.”
Innovation: RNAi screens for elucidating drug mechanism of action
Netherlands Cancer Institute
As one of the first researchers to systematically create RNAi reagents for knocking down large swaths of human genes, René Bernards is well versed on the ins and outs of RNAi probe design. Together with a collaborator at Cancer UK, Bernards obtained funding in 2002 to design three separate shRNA probes for each of 8,000 human genes under investigation. In 2003, Bernards set off on his own to put the RNAi vectors work in his own lab.
Being a researcher at a cancer institute, it’s no surprise that Bernards has chosen to apply his RNAi reagents to studying that disease. But Bernards’ particular choice of research focus involves the pursuit of innovative drug targets and determining the mechanism of action of cancer drugs already on the market. This is important, he says, because there are quite a few anticancer drugs with unresolved mechanisms of action, and understanding how they work could lead to the design of more powerful drugs that only affect the targets important for cancer progression.
To do this, Bernards sets up RNAi screens using cell lines appropriate to the type of cancer under investigation. By systematically knocking out potential targets of the anticancer drug and observing which colonies of cancer cells continue to grow, the researchers in Bernards’ lab can narrow the field of proteins affected by the drug in the hope of piecing together exactly how the compound operates in the cell.
Bernards has already submitted a manuscript on using genetic screens involving histone deacytelase inhibitors to find the molecular pathways they target. In addition, he’s investigating novel p53 regulatory drugs and has found insights into their mechanism of action with the help of RNAi experiments.
Ultimately, Bernards believes RNAi screens will lead to the discovery of “innovative and more intelligent” drug targets in cancer. For example, the holy grail of cancer drug targets, he says, involves a “synthetic lethal interaction” in which the inhibition of one gene product by the cancer drug only kills the cell when another gene product has been been mutated by the cancer cell — such as p53. “Now you come into the realm of what we would call genotype-specific drugs, where the drug is only acting in the context of a specific and defined genetic deletion, and RNAi will be a spectacularly powerful tool to identify them.”
Innovation: Nanoscale microfluidics for HTP tumor analysis or detection
University of California, Berkeley
When Richard Mathies sees scientists using microliter volumes of sample to prepare standard biological assays, he sees incredible waste. “That’s 10,000 to 30,000 times more material than they’ll use [in the experiment],” says the Berkeley professor of chemistry, who contends that this kind of research can be done with 50 to 100 nanoliters. So Mathies has come up with a nanoliter-volume microchip that can perform experiments from the sample prep stage through to the completion of the test.
That chip has a vast array of applications — scientists in Mathies’ lab use them for space exploration, forensics, and nanoparticle synthesis work — and one of the most promising is in better understanding and treating cancer. The chip can be used for “high-throughput genetic analysis of tumors,” such as genotyping, Mathies says. It’s also being developed as a portable device that could be used, say, to test a biopsy from a surgical field and be able to provide rapid results on “where the tumor has invaded, and to what extent” while the operation was still underway.
Mathies says people have been working for years to scale down processes like electrophoretic separation to the microchip level, but no one had successfully found a way to winnow the sample prep part down to the nanoliter level. Mathies and his team came up with an ultra-tiny membrane valve technology that can integrate pumps, mixers, valves, and other hardware needed for sample prep. The chip is also more successful than its predecessors, he adds, because the valves are compatible with aqueous solution — a characteristic not true in many similar MEMS technologies. “We make many hundreds of these valves on a wafer,” Mathies says. “They’re robust, high-density, and completely trivial to fabricate.”
Though the cancer applications of the chip sound far-off, Mathies says they’re actually within reach. The technology is “pretty far along,” he says, and his team has performed “some basic experiments in this area.”
Innovation: Technology to detect presence and density of methylation
It was back in 2002 that Orion Genomics, originally a company that used methylation technology to help analyze plant genomes, started a business unit focusing on molecular diagnostics. Today, CEO Nathan Lakey sees that field as so promising that he says the company is still shifting resources into that division, with the goal of discovering, developing, and marketing diagnostic tests, initially for oncology.
Such diagnostics, based on the company’s core expertise in DNA methylation, could be used for therapy selection or cancer screening, Lakey says. “There’s a lot of hope that methylated biomarkers will be a very strong way and a very sensitive way to select an appropriate therapy for a patient,” he adds.
Orion has two technologies geared toward studying cancer-related methylation. One is used to study whether and how much various parts of the genome are methylated using comparative genomic hybridization. Unmethylated DNA might be dyed green and the entire genome labeled with red; when sections hybridize, the still-red portions of the genome represent methylated DNA — and, Lakey says, “the redder the feature glows, the more methylated it is.” At this point, notes Jeff Jeddeloh, Orion’s director of biomarker discovery and detection, density of methylation (as opposed to presence or absence of methylation) is not well understood. “I think it’s likely gene-specific or maybe disease-specific,” he says of the regions that appear capable of acquiring additional methylation. This work could one day yield a homebrew diagnostic test, says Lakey, adding that such a test might be viable in two to three years.
The second technology is aimed at targeting methylated DNA as a means of identifying diseased genes in a tumor. Orion scientists can destroy all unmethylated copies of a gene in a sample, and then amplify the methylated copies. That’s necessary in cases where the sample might be a biopsy where “maybe one in 1,000 cells are diseased,” Lakey adds. “You’ve got to be able to see the disease pattern in all of that normal DNA.”
Those disease patterns are expected to emerge as Orion moves forward with its discovery platform to measure methylation across the whole genome. Measurements are taken in disease and normal samples in “a suspicion-blind approach. … Then you let the biology tell you which sites are correlated and which are not,” Lakey says. That process has allowed his team to find sites for methylation gain and loss.
To keep the samples coming and move the research along, Orion has “signed several collaborations with leading oncologists, cancer researchers, and leading cancer centers,” Lakey says.
Innovation: A statewide network of affiliates to increase access to genomic technologies
H. Lee Moffitt Cancer Center & Research Institute
One major hurdle that has always faced genomics has been finding a way to cross the threshold into mainstream medicine. But at the Moffitt Cancer Center, Director William Dalton thinks he’s found a way into the physician’s office.
Moffitt is the only NCI-designated cancer center in Florida, where cancer claims more lives than in almost all other states. With 100,000 new cases of cancer in the state each year, it was clear early on, says Dalton, that there was no way for Moffitt researchers to see all of those patients in person. Besides that, he adds, it wouldn’t have been appropriate: “If the patient can gain access to this technology and clinical trials in their own community, it’s far better for the patient.”
That mentality has led to a program known as Total Cancer Care, an initiative at Moffitt aiming to create a network throughout the state that would give smaller communities access to the cancer center’s top-notch technology. “Genomic profiling is a major effort at our institution,” Dalton says. During the past five years, his team has established a network of 14 affiliates — that number is growing, Dalton says — that sees close to a quarter of all cancer cases in Florida.
The idea is that Moffitt will build an information system that would allow physicians, and conceivably even patients, to access data about clinical trial availability as well as genomic profiling technology. “The community will participate and enroll patients, send us the tissue, we’ll do the profiling, and the patient will be treated in their community,” Dalton says. He hopes to capitalize on economies of scale with this system: clinical trials, for instance, might be larger, take less time, and be more useful through the affiliate network.
While the clinical trial arm and the network of affiliates have been established, Dalton says it’s important to note that the center is just getting started in developing the information system needed to bring everything together. To that end, the center is working with partner IBM and expects to do a good chunk of the work in house.
Innovation: Sequencing mutations known to play a role in cancer
It was a kind of perfect storm of pharmacogenomics events that led to a cancer-related business boom for Agencourt Bioscience. With publications last year showing that mutations in the EGFR receptor were indicative of a response to cancer drug Iressa, and the maelstrom over Vioxx, many companies — particularly pharmas — became increasingly aware of the importance of studying the genetic components to drug response, says Agencourt CSO Kevin McKernan. “A lot of companies have kinase inhibitors in the pipeline,” he notes. “We’re working with a lot of the pharmas now … on resequencing genes involved in these pathways.”
That’s meant scaling up the PCR-based resequencing business at Agencourt, which has made strides in the last few years as a production-scale sequencing shop.
Cancer-related business has really taken off, McKernan says. This kind of work has “probably tripled … or more than tripled” in the past year. What was previously a workload balance of about 70 percent plasmid and 30 percent PCR-based resequencing has “completely switched,” he adds.
His team keeps track of the most frequently requested genes — GPCRs, tyrosine kinases, and phosphatases tend to top the list — and of Agencourt’s internal top 10, McKernan says, “I think almost every one of them has been implicated in cancer.” Scientists at Agencourt now receive biopsies on a regular basis from hospitals, biotechs, and pharmas looking for EGFR sequence data. As a result, the company is installing clean rooms, working on whole genome amplification technology to better extract DNA from biopsy tissue, and awaiting CLIA approval.
Resequencing makes a lot of sense for studying cancer, according to McKernan. “A lot of growth with cancer genomes tends to dramatically rearrange” information within cells, he says — making it a difficult disease to study with a traditional microarray or other fixed-order approach. “You’ve got to think about that genome as a whole new genome.”
Innovation: Cell simulation for pathway interactions and drug reactions
For a graduate student in physics who never quite earned that degree, Colin Hill can at least say he finagled something out of his years as a physics PhD candidate at Cornell: his own company. In the summer of 2000, Hill founded Gene Network Sciences with a colleague from Cornell’s physics department, and in the years since has led the company to a fairly respectable position for a startup with 23 employees. In addition to accolades and awards collected over the past four years for the company’s science, in March the company announced a deal with Johnson & Johnson to study the mechanism of action of one of its oncology compounds under development.
From the start, Hill believed the study of cancer should be the primary application for his company’s approach to cell simulation. The company’s technology — a biochemical model of specific cellular systems — can take full advantage of the rich sets of molecular and clinical data gathered from cancer patients, he says. Furthermore, “the complexity of cancer makes it possible for systems biology to have an impact.”
In practice, GNS has taken a two-pronged approach to modeling the mechanisms of cancer drugs: After assembling a network of gene and protein interactions that describe a particular type of cell — a colon cancer cell, for example — Hill’s scientists introduce the parameters describing the drug and look for perturbations that give some clue as to the molecular pathways affected by the drug’s action. “It’s like an animal that will eat anything — the inference engine is not very finicky about input — but the results only provide a coarse-grained picture of what’s happening.” In the second phase of the simulation, GNS researchers tighten the constraints on the model to focus just on the pathways associated with the drug in an attempt to elucidate the drug’s specific mechanism of action.
Hill says it’s still early to cite particular insights into the disease that GNS simulations have made possible, but he does mention unpublished data obtained through a collaboration with Steve Dowdy at UCSD that appears to shed light on how the G1/S transition and the metabolic pathways that control it relate to cell replication — and hence cancer. And if the collaboration with J&J is successful, GNS’s modeling approach could contribute to bringing an oncology to market, Hill says.
In the meantime, Hill says, GNS’s ability to accurately simulate cancer cells will only improve, as clinicians begin relying more heavily on biomarkers to diagnose and track disease, especially in oncology. “It’s only now you’re starting to see some of these genomics approaches being applied in a diagnostic way,” he says. “This is key to our models having that much more impact in clinical drug development, [because we’re] able to get out the right data from human tissue, and not just animal models and in vitro cell lines.”
Innovation: caBIG: open-source applications and repositories for cancer researchers
National Cancer Institute
It’s easy to imagine potential pitfalls for a pilot project consisting of more than 600 scientists, but Kenneth Buetow, director of NCI’s Center for Bioinformatics, says NCI’s caBIG initiative has managed to avoid those and instead make a great deal of progress in its first year.
The Cancer Biomedical Informatics Grid, better known as caBIG, came about as NCI senior leadership “recognized the critical influence and importance bioinformatics plays to the rapidly evolving cancer research paradigm,” says Buetow. They saw a need to “create a really virtual network of interconnected cancer organizations, institutions, and research, data sources, and applications,” he adds.
And so the concept of caBIG — a giant network of cancer research institutes — was announced in July 2003, and in the following month, staff members visited all of the NCI designated cancer centers that expressed interest. By February 2004, the new consortium, consisting of 50 cancer centers and another 30 volunteer organizations, had its first meeting — as well as $20 million in funding for the first year.
Early on, Buetow says, the idea was to make sure that applications and data sources developed were ones that truly met existing needs. To that end, the organizations were divided into workspaces and then into working groups; each group identified its own goals based on those organizations’ skills and needs. (The eight workspaces include: clinical trials management systems; tissue banks and pathology tools; integrative cancer research; data elements and vocabulary; constructing the caBIG architecture; training; strategic planning; and data sharing and intellectual capital.)
Another way to keep things practical is the developer/adopter teaming that is rapidly becoming a key component of caBIG. Application or project developers are paired with groups that want to use such tools or resources, giving them instant and thorough testing as they’re being designed. “Our goal is to be sure, as we develop caBIG infrastructure and applications, that they meet immediate needs of cancer researchers,” Buetow says. Also, he points out, all code developed for caBIG is open source and easily accessible.
The caBIG initiative — what Buetow refers to as “this large, integrated World Wide Web of cancer research” — is on a three-year pilot program, after which it will be re-evaluated by NCI. After its first year, Buetow says, there’s obvious demand for new workspaces, including one for imaging efforts.
“We certainly hope that caBIG … will become a self-sustaining activity,” Buetow says. “We also hope that as this community comes together we’ll be able to create a large enough market in bioinformatics, information technology, and biomedical informatics that vendors will be able to find successful models to work with us.”
Innovation: Gene expression analysis of cancer biology with microarrays
Harvard Medical School
To many in the genomics community, Todd Golub is known as one of its most preeminent practitioners of microarray analysis in cancer. In his nine years at Harvard Medical School and the Dana-Farber Cancer Institute, Golub has won numerous awards — not to mention grant funding — and has risen to director of the school’s cancer program.
But Golub didn’t decide to specialize in gene expression using microarrays because he thought it was necessarily the most powerful approach to studying cancer; rather, the reason was more specific. His lab was interested in looking into the role of recurring chromosome translocations in the onset of acute lymphatic leukemia, and needed a tool for systematically associating specific types of translocations with various manifestations of the disease. “DNA microarrays provided us with the possibility of taking an unbiased look at these translocations [and producing] a molecular taxonomy of cancer more grounded in biology,” he says.
In addition, Golub saw microarrays as a useful tool for delving deeper into the function of the oncogenes represented by the recurring chromosome translocations in leukemia. Generating lists of oncogenes was one thing, he says, but investigating the molecular consequences of these mutations might produce even more insights into cancer. So his lab introduced translocation-causing fusion proteins into cell lines and put DNA microarrays to work capturing the downstream consequences. At the time, “DNA microarrays offered the first approach to doing this,” he says.
One of the more general conclusions derived from this work — and one that has found support in the work of several other groups — is that much of the clinical behavior of cancer is hard-wired into the tumor at the time of diagnosis, Golub says. While acknowledging that the disease does have the potential to evolve in the patient, Golub says, “it means there’s an awful lot we can learn from the tumor at the time of diagnosis.” Ironically, because cancer diagnosis doesn’t routinely involve gene expression analysis, much of that information is lost.
Looking ahead, Golub is one of the principal investigators on a grant to create an Integrated Cancer Biology Program at Harvard. Together with Steve Carr, who will perform phosphoproteomics experiments, and Bill Hahn, who’s in charge of RNAi experiments, Golub will apply his gene expression analysis expertise to investigating the role of activated kinases in cancer progression.
What is the most promising approach in cancer research right now?
Systematically determining the genotypes of all the common cancer types — a critical step to rational classification, drug development, and treatment decisions.
— Harold Varmus, president,
Memorial Sloan-Kettering Cancer Center
I think the most important agenda is to improve molecular diagnostics. Better diagnostics will improve everything else. Especially early detection of cancer, which can save many lives.
— Lee Hartwell, director,
Fred Hutchinson Cancer Research Center
Understanding the basic biology of tumors. Genomic and proteomic profiling is uncovering all kinds of new targets and pathways that will be unique to cancer cells. The era of translational research is really coming up.
— William Dalton, director, H. Lee Moffitt
Cancer Center & Research Institute
Integrative cancer biology — the study of cancer as a complex biologic system — is one of the most promising areas in cancer research today. The complex interactions between cells and their microenvironment, and between the organism and its macro environment, are potential targets for new and more rationally designed interventions to prevent, detect, and treat cancer.
— Dinah Singer, director,
Division of Cancer Biology, NCI
Using genomics, transcriptional profiling, and proteomics to identify novel targets for therapy and combining this with molecular markers to identify patients likely to respond with limited toxicity. This needs to be combined with functional or molecular imaging to identify responders to therapy at an early stage.
— Gordon Mills, professor,
University of Texas MD Anderson Cancer Center
Mechanism-based therapies appear to be the most desirable approach at this time. The techniques that made this possible [are] not genomics or proteomics or any of the omics; it is a combination of good old enzymology, structural biology, cell biology, immunology, virology, genetics, chemistry and pharmacology.
— Anthony Yeung, member,
Fox Chase Cancer Center