January 1, 2006 (Vol. 26, No. 1)
Technologies, Tools, and Methodologies for Successful Validation and Clinical Development
With the human proteome comprising 10 times more proteins than there are genes in the genome, the complexity of proteomics research and the efforts under way to identify and characterize protein biomarkers of disease and translate those findings into informative and reproducible diagnostic tests is driving the development of new technologies and improved methods for analyzing proteins in biological samples.
“Proteomics: A New Diagnostic Frontier” was the first freestanding conference of the newly formed proteomics division of the American Association of Clinical Chemistry (AACC).
Convened in Washington, D.C., the meeting featured scientists, clinicians, and regulatory officials who spoke on the challenges and opportunities for biomarker discovery and the development of clinical diagnostic tests for diseases ranging from cancer to cardiovascular and kidney disease to sleep apnea.
Linking the varied presentations was an emphasis on the need for robust, high-throughput proteomic tools for biomarker discovery and validation, and strategies for translating those methods into easy-to-implement diagnostic tests for use by clinical laboratories at the level of the community-based hospital.
Leigh Anderson, Ph.D., CEO of the Plasma Proteome Institute (Washington, D.C.), spoke of the promise of proteomics, noting that proteins are not only more closely linked to cell function than either DNA or RNA and are relatively accessible, but also that proteins are potentially capable of defining all disease states.
The development of new protein-based diagnostic tests has slowed dramatically over the past decade, and Dr. Anderson attributed this to a black hole, or critical gap in the diagnostics developmental pipeline, which he defines as marker validation.
Both government and industry have ignored this issue, contended Dr. Anderson. Without adequate funding to validate new biomarkers of disease, they will not make the leap from the lab bench to the clinical laboratory.
Plasma represents the largest repository of human proteins in a single sample. However, the fact that only 10 proteins make up 90% of plasma by mass is confounding efforts to study the plasma proteome. Detecting biologically important, low-abundance proteins that may be masked by or even bound to proteins such as albumin, and comparing their levels in normal versus disease samples, has proved to be a daunting technological challenge.
Novel multidimensional fractionation techniques, including various combinations of 2-D gel electrophoresis, liquid chromatography (LC), and hybrid mass spectrometry (MS), have greatly increased the numbers of proteins accessible for biomarker analysis.
At present, though, there is no consensus on a comprehensive proteomics platform, further complicating the translation of this technology from the bench to the bedside.
William Hancock, Ph.D., chair of the department of chemistry in the Barnett Institute at Northeastern University (Boston), talked about the challenges of plasma fractionation and described his group’s use of combined fractionation techniques, including depletion of abundant proteins and protein enrichment using multiple lectin affinity chromatography (MLAC).
The main challenges in plasma fractionation are the high charge and mass heterogeneity of protein isoforms, the abundant proteases and phosphatases, plasma stability, the presence of apolipoproteins that foul HPLC columns, and the high concentration of albumin, which hinders 1-D and 2-D gel separation strategies.
Dr. Hancock emphasized the importance of looking for silent changes in protein expression, and not only changes in the amount of a protein, but changes in glycosylation as well, which may be associated with a disease phenotype.
His group uses multidimensional liquid chromatography and linear ion trap MS for protein separation and identification. They select for glycosylated proteins by loading a plasma sample onto an MLAC column. The glycosylated proteins will bind to the lectins.
The use of several different lectin moieties allows for the enrichment of different subpopulations of the plasma glycoproteome. Dr. Hancock then loads displacers sequentially onto the column and collects the fractions displaced.
Comparing the displaced fractions in which the proteins appear reveals the glycosylation state of each protein. Dr. Hancock described the use of this technique to analyze breast cancer samples.
His group is also using this technique to study changes in glycoprotein levels following myocardial injury, analyzing patient samples before and after myocardial ablation to search for changes in glycoprotein levels that might signal, for example, inflammation or endothelial dysfunction.
Mass Spectrometric Techniques
Denis Hochstrasser, M.D., professor of medical biochemistry at Geneva University Hospital (Switzerland), described the need for instrumentation that can perform multidimensional protein separation and analysis and provide accurate results rapidly and inexpensively.
One of the main questions in clinical proteomics at present is where to focus biomarker discovery efforts. Plasma may not be the best place to start, in Dr. Hochstrasser’s view, even with fairly sophisticated tools, due to the complexity of its protein make-up.
He suggested beginning the search for disease-related biomarkers in the affected tissuesfor example, for stroke, looking at the spinal fluid for markers of cell death as evidence of brain tissue injury. Once a marker is identified, an immunoassay can then be developed for clinical diagnostic use that detects and measures the marker in plasma.
Dr. Hochstrasser anticipates that clinical labs will require instrumentation and expertise in a combination of immunochemistry and MS techniques, describing MS as “the future of molecular medicine.”
One can imagine being able to take a thin slice of a diseased tissue sample and expose it to laser energy, causing the proteins to fly out of the sample and into a mass spectrometer, which would then identify and analyze the proteins, looking for specific biomarkers to guide diagnosis.
Eleftherioss Diamandis, M.D., Ph.D., head of clinical biochemistry at Mount Sinai Hospital in Toronto, explained the mechanics behind mass analyzers and the use of electric or magnetic fields, generated by ion traps and quadrupoles, for example, to manipulate ionized particles, allowing only one ion to enter the detector at a time.
Mass spectrometers analyze gas-phase ions based on their mass-to-charge ratio. Second-generation mass analyzers combine elements of earlier systems to generate hybrid instruments, such as dual time-of-flight, triple quadrupole, quadrupole-TOF, and triple quadrupole-ion trap instruments.
Like existing techniques, novel strategies in development to break down proteins and separate them for better resolution each have their own advantages and disadvantages in terms of accuracy, resolution, and throughput.
Dr. Diamondis described the advantages of top-down proteomics, analysis of whole proteins, for qualitative applications such as identifying proteins in a complex mixture and studying post-translational modifications (PTMs), versus the advantages of bottom-up proteomics, in which the proteins are first broken down into peptides, which are then analyzed.
The latter approach is useful more for quantitative proteomic applications such as biomarker studies, comparing normal versus disease tissue samples, for assessing changes in protein levels before or after drug treatment, or for time-course studies.
Reagents developed for quantitative proteomics, which label proteins with different colored tags, enable monitoring of multiple parameters in one experiment.
State-of-the-art MS-based protein analysis is far from a perfect science. Dr. Diamondis provided the example of a single sample that was analyzed by six different laboratories. In total, 1,109 proteins were identified, but only 52 proteins were identified by all six labs, and 63% of proteins were identified by only one group.
Nevertheless, he sees MS and protein microarrays as the future “workhorses of high-throughput proteomics.” Describing the protein microarray as the “compact disk of the future,” he envisioned it playing a central role in protein identification and functional studies aimed at evaluating protein-protein interactions.
During a panel discussion, participants identified the most critical issue for clinical proteomics going forward as the reproducibility of results from lab to lab. Whereas peak intensity may vary depending on the MS instrument used, mass accuracy should not.
The panelists also discussed the challenges inherent in working with biological specimens, including storage issues. Whatever is done to a sample once it leaves the patient will modify it, and the potential effects of freezing a sample, breaking open cells, and isolating the proteins, or digesting a protein sample into its peptide components at the bedside, for example, must all be characterized and understood.
Proteomics-based Cancer Detection
Samir Hanash, M.D., Ph.D., director of molecular diagnostics at the Fred Hutchinson Cancer Research Center in Seattle, described efforts to use proteomic technology to identify highly specific tumor antigens present in low amounts in plasma samples.
Dr. Hanash’s group is developing the intact protein analysis system (IPAS), which incorporates immunodepletion to eliminate 90% of plasma proteins, followed by protein labeling and fractionation of intact proteins based on charge using reverse-phase chromatography, and then quantification of individual proteins.
Working as part of a consortiumthe Biomarker Research Initiatives in Mass Spectrometry (BRIMS), a collaboration between Thermo Electron and Massachusetts General HospitalDr. Hanash’s group is formulating methodologies for studying the proteomics of mouse models of adenocarcinoma.
Citing the dynamic range of proteins in plasma as a crucial problem for biomarker discovery, Dr. Hanash stated, “We have been able to drill down to the dynamic range level where you find PSA.” His group is also using protein microarrays to screen for autoantibodies against tumor proteins.
The researchers are organizing a collaborative project that will evaluate serum samples obtained at least a year before a diagnosis of cancer to search for potential markers present prior to clinical manifestations of disease.
Lance Liotta, M.D., Ph.D., professor and co-director of the Center for Applied Proteomics and Molecular Medicine at George Mason University (Fairfax, VA), discussed the role protein microarrays can play in cancer detection and in tailoring cancer therapy. He described cancer as a proteomic disease driven by defective protein signaling pathways.
To satisfy the need for high sensitivity and specificity of a protein-based cancer diagnostic tool, clinicians will have to use panels of markers to identify a diagnostic pattern that can differentiate diseased from healthy cells.
Challenging the wisdom of removing albumin and other large, high-abundance proteins from a sample prior to proteomic analysis, Liotta pointed out that more than 90% of low molecular weight biomarkers are bound to large carrier proteins and need to be disassociated from the carriers prior to sample fractionation.
He referred to a paper in the October issue of Clinical Chemistry that described the proteomic analysis of an ovarian cancer sample, which yielded a series of low abundance proteins as candidate biomarkers.
Dr. Liotta also discussed the application of phosphoprotein microarray-based assays to interrogate the phosphoproteome. These highly sensitive and reproducible assays can be used to differentiate patient survival and response to treatment depending on the phosphorylation state of tumor receptors.
Developing panels of predictive protein biomarkers will require well-characterized starting materials, explained Daniel Chan, Ph.D., professor and director of the Center for Biomarker Discovery at Johns Hopkins University (Baltimore), who emphasized the need for information on how samples are collected, transported, and stored, together with clinical information about the patient, diagnosis, and clinical course.
Ideally, specimen handling and processing should be standardized and automated. Dr. Chan described his group’s work using pH fractionation via anion exchange chromatography and protein microarrays to look for protein fragments or truncated forms of proteins that have clinical value.
Detecting Biomarkers in Urine
“Post-translational modifications are a treasure trove of clinical information,” said Jon Klein, M.D., Ph.D., professor of medicine at the University of Louisville (Kentucky). As an example, he described his group’s efforts to identify tissue markers of renal disease in patients with type 1 diabetes.
They extracted proteins from the kidneys of a mouse model of diabetic nephropathy and subjected the protein mixture to 2-D polyacrylamide gel electrophoresis (PAGE), looking for protein spots that demonstrate differential expression between normal and diseased kidneys and, in particular, coordinated changes in a single pathwayfor example, down-regulation of elastase expression and up-regulation of elastase inhibitor resulting in increased elastin deposition in the kidney.
Dr. Klein’s group demonstrated that protein expression correlated with their clinical findings, which showed increased elastin deposition on staining of the kidneys of animal models of disease and of human patients with diabetic nephropathy. They are now quantifying the cross-linked elastin molecules in urine.
Describing the kidney’s glomerular barrier as a “gift of nature,” because it filters albumin out of urine, Dr. Klein said, “We think urine is a potentially rich source of biomarkers.” His group is exploring new approaches to urinary proteomics, pursuing MS analysis of human samples to study low molecular weight proteins, and focusing on changes in the urinary peptidome due to hyperglycemia. They will compare proteins extracted from urine of children with type 1 diabetes mellitus who do not have kidney disease with urine from the 33% or so of adults with type 1 diabetes who develop renal disease.
The dynamic range of 2-D gels is a limiting factor in protein separation, according to Dr. Klein, and his group is switching to a sequential (ion exchange and reverse phase) liquid chromatography separation technique. They are also transitioning from MALDI-MS/MS to linear ion trap technology for MS acquisition.
Dr. Klein identifies the main research barrier at present as the need for large, well-annotated sample sets. The key challenge for transferring a defined protein biomarker to an assay suitable for use in a clinical lab is the need for rapid prototyping of antibodies.
Saeed Jortani, Ph.D., assistant professor, pathology and laboratory medicine at the University of Louisville (KY), presented his group’s work on identifying biomarkers of sleep apnea.
Dr. Jortani’s group is doing protein profiling of first morning urine samples in affected children and has identified two low molecular weight proteins with diagnostic value that could be used to differentiate between patients with and without obstructive sleep apnea (OSA). A urine-based assay would provide a relatively low-cost testing alternative compared to the need for an overnight sleep study, and offers 97% sensitivity and 92% specificity.
One of the differentially expressed proteins identified by Jortani’s group is urocortin II, which plays a role in the adaptive stress response. Dr. Jortani theorizes that the hypoxemia and hypercapnia associated with OSA induces a stress response that leads to recovery phase proteins being present in morning urine samples.
Catherine Fenselau, Ph.D., professor of chemistry and biochemistry at the Marlene and Stewart Greenbaum Cancer Center at the University of Maryland (College Park), described the use of comparative proteomics to identify differentially expressed proteins in cultured human cancer cells that correlate with drug resistance.
Dr. Fenselau’s group takes a unique approach to simplifying the protein mixture to be analyzedthey separate the organelles and study only the proteins expressed in individual organelles, such as the nucleus.
The group uses 2-D gel electrophoresis for protein separation and relies on two methods to assess changes in abundance of soluble nuclear proteins: isotope enrichment and gel densitometry. Proteolytic 18O labeling and solution IEF separation/HPLC-MS/MS analysis are used for comparative proteomic analysis of insoluble nuclear proteins.
“Highly multiplexed assays will fundamentally change how we think about quality control,” Stephen Master, M.D., Ph.D., assistant professor, pathology and laboratory medicine at the University of Pennsylvania (Philadelphia), told the conference participants.
Dr. Master anticipates the need for panels of diagnostic biomarkers and multiplexed clinical assays that will require large numbers of analytes and necessitate new bioinformatics-based quality control paradigms. Rather than focusing on individual analytes, which would be impractical with massively multiplexed assays, quality control strategies will have to focus on protein patterns.
HUPO and the FDA
Gilbert Omenn, M.D., Ph.D., professor of molecular medicine and genetics at the University of Michigan (Ann Arbor), outlined the long-term scientific goals of HUPO’s Human Plasma Proteomic Project: a comprehensive analysis of plasma and serum protein constituents in people; identification of biological sources of variation (physiological, pathological, and pharmacological) within individuals over time, with validation of biomarkers; and determination of the extent of variation across populations and within populations.
HUPO partnered with BD (www.bd.com) to acquire specially prepared male/female pooled samples (in three separate ethnic pools).
Dr. Omenn described the progress of the project to date and the importance of the next phase, which will include experimentally assessing the reproducibility of results and standard operating procedures to make proteome analysis practical for routine clinical chemistry, as well as submitting lists of proteins to HUPO’s Antibody Production Initiative, identifying and quantifying modified proteins (subproteome analyses), profiling low-abundance proteins, identifying biomarkers of disease, and collaborating on informatics, database creation, annotations, and error estimation.
Steven Gutman, M.D., director of the office of in vitro diagnostics at the FDA, spoke of the regulatory issues surrounding proteomics-based diagnostic tests, including their analytical accuracy, specificity, and sensitivity, and their clinical performance.
Despite the many challenges and uncertainties, he pointed to some good news, including a refined regulatory toolbox that includes pre-IDEs (Investigational Device Exemptions), expedited reviews, de novo classifications, and real-time interactions.
The pre-IDEs provide for a free protocol review process and 60-day turnaround, and an opportunity to educate the FDA on new technologies to prepare for questions that might come up on review and establish scientific impact and plan for disagreements.
Dr. Gutman also spoke about two critical FDA initiatives: the introduction of the STARD (Statement for Reporting Studies of Diagnostic Accuracy) accuracy initiative; and the FDA Data Template, which includes creating an electronic data submissions template aimed at standardizing data input, output, and review.
A prominent component of the FDA’s broad initiative to accelerate the critical path from bench to bedside is the development of biomarkers as diagnostics and tools in drug discovery.