YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

Dramatic advances in identifying protein biomarkers are spurring new hope in cancer diagnostics, expediting detection and easing testing

By Deborah Janssen

April 16, 2004 | The latest statistics from the American Cancer Society make for grim reading. More than 563,000 people will die of cancer this year — that's 1,500 people each day. One of the keys to reducing these numbers hinges on the fact that, if diagnosed before a tumor undergoes metastasis, the five-year survival rate can exceed 90 percent. "The war on cancer cannot be won with better cancer drugs," says Matritech CEO Stephen D. Chubb. "The major success we've had in the past 70 years has been with cervical cancer, and that's because we are able to detect and treat it at a precancerous stage. As a result, mortality can be reduced to virtually zero."

While big pharmas set their sights on the next Gleevec, biotechnology companies and government agencies are undertaking enormous efforts in molecular diagnostics, searching for proteins that may serve as biomarkers for early detection. This could lead to diagnostic tests that are simple, non-invasive, and cost-effective to the patient.

Nuclear Weapons 
Matritech's bladder cancer test marks the first application of the company's nuclear matrix protein (NMP) technology.

Read More 
"One of the most interesting aspects to come out of molecular taxonomy projects, particularly those utilizing microarray analysis, is that cancer is a very heterogeneous disease, perhaps even involving patient-specific signatures," says Emanuel Petricoin, co-director of the Clinical Proteomics Program, a joint venture between the FDA and the National Cancer Institute (NCI). Diagnosing cancer based on histological findings doesn't give the full picture, Petricoin says, which explains why there haven't been any new cancer biomarkers on the market.

In fact, the only current clinical biomarker tests in regular use for cancer are the prostate-specific antigen (PSA) test for prostate cancer, CA-125 for ovarian cancer, NMP22 for bladder cancer, mammography, and the Pap smear. The latter two involve more invasive procedures that many women avoid, thus getting diagnosed with cancer at a later stage. Although the PSA and CA-125 are good-quality single-diagnostic-marker tests that require only a simple blood test, they don't have the reliable predictive value.

 TEST PATTERN: Emanuel Petricoin of the Clinical Proteomics Program has helped develop a diagnostic approach involving pattern recognition. 
In short, it is difficult to find a single marker that has the sensitivity and specificity to achieve clinical utility. Fortunately, the future of cancer biomarkers is bright, largely as a result of advances in proteomics.

Petricoin and the NCI's Lance Liotta are the co-developers of a new diagnostic paradigm: proteomic pattern diagnostics. This concept, which does not require standard immunoassays, is now being independently developed, commercialized, and validated, and could potentially result in a diagnostic blood test for the early detection of ovarian cancer.

The test, called OvaCheck, utilizes pioneering pattern-recognition technology from Correlogic Systems, in Bethesda, Md., which succeeded in detecting 100 percent of ovarian cancer cases, including stage one, the earliest and most curable stage of the disease (Petricoin, E.F. et al. The Lancet 359, 572-7; 2002). OvaCheck requires just a drop of blood, and will soon be developed as a laboratory-based test. Correlogic has a licensing agreement with Quest Diagnostics and Laboratory Corporation of America (LabCorp), the two largest clinical laboratories in the United States, which will make the test available through their laboratories and nationwide collection systems.

Correlogic's approach "looks at subtle changes in protein patterns in the blood, rather than an increase in an individual biomarker," says Peter Levine, president and CEO. From that concept came the development of the company's Knowledge Discovery Engine (KDE), so named because of the way it can extract knowledge hidden in the data, and not merely identify patterns.

Resources to Enhance Biomarker Discovery 

Read More 
The KDE is the heart of Proteome Quest, a sophisticated software tool that builds computational disease state models. "It works like a factory," Levine says. "When you feed the appropriate data into it, a computational model is constructed that is subsequently used to score individual protein patterns from patients that were prepared exactly the same way under the same conditions." The model is then used to observe if certain features or data points match a specific patient's patterns. "This allows us to apply the same technology and processes to different disease states without having to build a separate analytical platform for each disease," Levine says.

Using a genetic algorithm, the KDE is able to sift some 160,000 data points to find the appropriate ones that distinguish a disease state from a healthy state. The process essentially simulates evolution in a computer, which is much faster than nature.

"One can never capture the entire range of expressions of any disease, because there are just too many people," says Ben Hitt, Correlogic's chief scientific officer. No matter how good your algorithm is, Hitt says, there is bound to be at least one patient who is always going to fall outside of it. "Our algorithm has an adaptive feature that is able to recognize if it has seen previous patterns."

Peak Performance 
To assess the increase in diagnostic sensitivity and specificity that would be afforded from high-resolution mass spectra...

Read More 
The hidden patterns approach was validated in a research study conducted with scientists from the NCI, the FDA, and the University of North Carolina's Lineberger Comprehensive Cancer Center to detect prostate cancer. The method has been twice as accurate as the ubiquitous PSA test in detecting prostate cancer, at least in the studies so far.

Reflexing to Help Patients 
Approved by the FDA in 1986, more than 20 million PSA tests are conducted annually in the United States. But the test suffers important clinical limitations, which is why Correlogic and others are working on an improved or complementary PSA test. Typically, if PSA levels exceed 4 ng/ml, guidelines recommend a biopsy to detect possible cancer. But about 100,000 cancers annually arise in patients with PSA levels below 4 ng/ ml, and the test also suffers from a high false-positive rate — not all individuals with PSA levels higher than 4 ng/ml have cancer.

"The PSA test doesn't have the positive predictive value that we would like in the 2-4 ng/ml range," says Patrick Plewman, CEO of diaDexus, a privately held biotechnology company founded in 1997 by SmithKline Beecham (now GlaxoSmithKline) and Incyte Genomics (now Incyte Corp.). Surprisingly, 2 million men fall into this 2-4 PSA range. "Here's the clinical dilemma: You aren't going to biopsy 20 men to find one cancer — that's not considered ethical or cost-effective. We need to be able to improve the biopsy 'yield' in the 2-4 ng/ml range," Plewman says.

DOUBLE VISION: Ciphergen's Eric Fung advocates using a combination of technologies. 
DiaDexus is developing a blood test that would take a patient with a borderline PSA level and "reflex" to a second blood test, which, if positive, would prompt a biopsy. DiaDexus is seeking an external collaborator to validate its internal studies.

Through access to Incyte's LifeSeq database, diaDexus is working to develop a portfolio of cancer diagnostics products, including the PSA "reflex" test. DiaDexus uses proprietary software to reduce the 12 million gene sequences in the Incyte database to a manageable subset of approximately 25,000 cancer-associated gene fragments. This subset is printed as custom oligo arrays by Agilent Technologies, allowing diaDexus to identify the handful of targets that will be explored as putative biomarkers at the protein level.

In 1996, Matritech received FDA clearance to market its NMP22 test for bladder cancer, hailed by CEO Chubb as "the most accurate, fluid-based cancer test ever developed." (See "Nuclear Weapons," page 52.)

Among several tests in development is one for breast cancer. "We are developing our NMP-based test for breast cancer to do what the Pap smear has done for cervical cancer," Chubb says. Cervical cancer was the largest cancer-related killer of women in the 1930s. However, the introduction of the Pap smear test has reduced deaths per capita by about 85 percent. Today, 80 percent of women who die of cervical cancer have not had a Pap smear in five years. "The combination of automating the Pap smear and getting everyone tested could enable us to come close to eradicating this disease," Chubb says.

Interestingly, almost as many women (per capita) are dying of breast cancer today as were in 1930. Despite heightened awareness and the use of mammography, breast cancer is still not detected early enough to reduce fatalities. Matritech has identified another nuclear protein, NMP66, which is present in early-stage breast cancer but not in healthy women. "[We] believe that this protein will be able to reduce those mortality rates," Chubb says.

Deciphering the 'Fragmentome' 
The NCI-FDA Clinical Proteomics Program is pushing the use of mass spectrometry (MS) as a lead clinical diagnostic device. "Mass spec has the ability to 'hunt and sort' through tremendous amounts of information very rapidly from a drop of blood, and allows you to obtain a proteomic portrait of the unmined, low molecular weight (MW) region of the proteome," Petricoin explains.

Previous efforts have relied upon 2-D gel electrophoresis or isotope-coded affinity tag (ICAT) technology for serum analysis, but these techniques primarily target the larger, more abundant proteins. According to Petricoin, 2-D gels cannot effectively separate proteins below 10 kilodaltons, but coincidentally this is the ideal range where MS has the greatest sensitivity and specificity.

"The area where an instrument has its best analytical precision is also the area where there is virtually nothing known about what exists in this low MW portion of the proteome — a region where metabolites, and peptide and protein fragments exist and which could be a diagnostic goldmine," Petricoin says. His group has been developing this method to demonstrate its reproducibility, using high-resolution QSTAR mass spectrometers from Applied Biosystems. Comparing low- and high-resolution MS instrumentation in ovarian cancer samples, Petricoin's team found that the high-resolution instruments were much more accurate and reproducible (see "Peak Performance"). "We believe this occurs because using low-resolution mass spectrometers, only 200 to 300 peaks may be observed. But on a higher-resolution mass spectrometer, the number of peaks can be increased more than tenfold," he says.

Petricoin's group has discovered that the low MW entities uncovered as patterns are actually fragments of proteins and metabolites bound to circulating carrier proteins such as albumin or immunoglobulins, which effectively harvest the rare fragments in disease-specific ways. "The carrier proteins have long circulating half-lives and are in high concentrations, which enable them to gather up all of the biomarkers at once. It's these biomarker fragments that are the diagnostic entities that we're profiling," he says.

These bound biomarker fragments can be amplified thousandfold and are the reason why MS can detect even low-abundance entities within a matter of seconds. "What researchers have logically done in the past is to throw out the high-abundance proteins with the reasoning that the biomarkers exist in some free, uncomplexed state. We analyze what is bound to the carrier proteins as our starting point of analysis both for discovery and for mass spec," Petricoin says.

Many scientists are skeptical that MS could be sensitive enough to detect even the low-abundance diagnostic information. "Now we have an answer for skeptics on why we can get this information in the first place — it's all being 'mopped up' and amplified by carrier protein harvesting," Petricoin says. In fact, the information archive is so complex that hundreds of combinations of any of these entities may be able to predict disease accurately and specifically. "It's why we don't want to assign exact identity to the peaks yet. We need to wait until we have locked in on which of the hundreds of pattern combinations are those with the most robust combination of peaks for a diagnostic test," he says.

The work is being extended with nanotechnology expert Mauro Ferrari at Ohio State University, to develop nanoparticles that can act as mimetics to these carrier proteins. The nanoparticles will have the same binding properties as, say, albumin. When the nanoparticles are placed in a patient's serum collection tube, they will be able to outcompete the natural carrier proteins for biomarkers.

"There is a whole new continent to explore, and the field can be energized and electrified by this discovery — the existence of the 'fragmentome,'" Petricoin says.

Early Detection Makes the Difference 
The National Institute of Standards and Technology (NIST) has a long track record of offering measurement tools to clinical labs, medical manufacturers, hospitals, and drug makers. Now, in collaboration with the NCI, it is aiming to validate early-detection cancer biomarkers. This is the first time the NCI has funded an entire project from discovery to independent validation to clinical application. "Funding a collection of researchers that are working together every day may enable us to fast-forward the process, so the patients in the clinic will get direct benefits as soon as possible," says Peter Barker, project leader of the Cancer Biomarker Validation Laboratory at NIST.

Barker's lab is validating three projects for NCI's Early Detection Research Network (EDRN). The most advanced study involves analyzing the performance and validation of induced chromosome breaks as an index of lung cancer susceptibility. A second study, the first of the NIST cancer detection assays to incorporate a robotics platform, focuses on mitochondrial DNA (mtDNA). Mitochondrial mutations have been found in lung tumors, so Barker's group has been working on ways to sequence the mitochondrial genome from minute samples of DNA. Says Barker: "The patient population required to clinically validate this test is 200 patients (normal and tumor samples), which equates to 400 samples multiplied by 16,569 bases (the mtDNA genome) at 100-percent accuracy. It's like a mini-genome project!"

The third project centers on telomerase, an enzyme that repairs chromosome ends that is switched on in tumor cells. "People have been interested in telomerase activity as a global tumor marker for a long time, but the problem is that the enzyme consists of RNA to some extent, so it degrades very quickly. We are hoping to be able to get a good reading using serum," Barker says.

Ciphergen Biosystems plays a prominent role with the EDRN using its surface enhanced laser/desorption ionization (SELDI) platform technology. "The EDRN is using our platform not only to discover biomarkers but also as an assay platform for actual diagnosis," says Eric Fung, Ciphergen's director of clinical affairs. EDRN can take patient samples and determine which biomarkers are important for prostate cancer diagnosis, run new samples and look for those same peaks, and use that information in a diagnostic setting.

Fung believes the future of medical diagnostics, not just cancer, will focus on looking simultaneously at multiple biomarkers and utilizing several assays in combination. "Technologies that allow us to look at the different ways proteins are processed in the body are going to be very important and will give us better insight into how diseases are different from each other and provide potentially additional diagnostic accuracy," Fung says. "Ultimately, one needs to be able to assay for these multiple biomarkers and processed proteins, and the platform that can do just that will be the successful diagnostic assay platform of tomorrow."

Deborah Janssen is a contributing editor for Bio·IT World. 


For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.