Although huge numbers of candidate biomarkers have been identified, “few have been validated,” said Wayne Patton, Ph.D., biochemistry director at PerkinElmer’s (www.perkinelmer. com) Life & Analytical Sciences division. “Meaningful validation will require high-throughput bioinformatics and large sets of data.” In addition to statistical validation, biological/functional validation will also be essential.
Speakers at a Cambridge Healthtech Institute conference on protein biomarkers emphasized several key points: the field of biomarker discovery is advancing, but there is a big gap between biomarker identification and translation to clinical applications; differential protein-expression efforts need to dig deeper into the proteome to identify low-abundance proteins and peptides that may have greater disease specificity; and ultimately the use of biomarkers for disease diagnosis, monitoring, prognosis, and therapeutic decision-making will depend on panels of biomarkers, patterns of protein expression, and technology platforms capable of multiplexed protein analysis.
Furthermore, for clinical applications, biomarker identification will have to move beyond a reliance on MS to higher-throughput, automatable assay formats.
John M. Hevko, Ph.D., senior field applications specialist at Applied Biosystems (ABI; www.appliedbiosystems.com) described the time and cost bottleneck that exists between biomarker discovery and biomarker validation.
“Discovery generates too many candidate biomarkers to validate,” said Dr. Hevko, citing the high cost of validation assays and the long lead time needed to develop them and to generate the necessary reagents.
Dr. Hevko said ABI’s MRM Initiated Detection And Sequencing (MIDAS™) Workflow technology enables rapid development of multiple reaction monitoring (MRM) assays. The assays can then be used to verify which biomarkers are most promising and should move forward into the validation process. This verification process helps to minimize the numbers of antibodies and synthetic peptides needed for biomarker validation.
Dr. Hevko described a current study of lung cancer metastasis and the need for biomarkers for early detection and differentiation between metastatic and nonmetastatic cancer. In the study, a cell line was transformed to overexpress ErbB2 receptor tyrosine kinase in order to create a model to study adhesion and invasion, the first two steps in cancer metastasis.
iTRAQ™ reagents were then used for protein labeling. The iTRAQ are multiplexed, amine-specific, stable isotope reagents for biomarker discovery that Dr. Hevko said can label all peptides in up to four different biological samples, enabling simultaneous identification and quantitation, both relative and absolute, while retaining important post-translational modification information. The next generation of these tags will be related non-isobaric tags (mTRAQ™) that are more compatible with biomarker verification MRM assays in complex mixtures.
After labeling, a comparison of metastatic and nonmetastatic lung cancer samples, using a global, nontargeted approach to protein-expression analysis, yielded 1,494 proteins, of which 27 exhibited a greater than 2-fold decrease in ErbB2 expression. Fifteen proteins exhibited a greater than 2-fold increase in expression. These candidate biomarkers were then prioritized based on biological information with a focus on proteins involved in cell structure and motility and are now undergoing verification by MRM analysis.
Model for Diagnostic Assays
Ventana Medical Systems (www.ventanamed.com) is leveraging its chemistry on glass slides and automated slide-staining technology (Discovery® XT) in collaboration with pharma companies to develop companion diagnostics for targeted therapeutic products.
Thomas Grogan, M.D., professor of pathology at the University of Arizona and CMO at Ventana, provided examples of existing assays: a single-analyte protein diagnostic for c-KIT developed with Novartis to detect gastrointestinal stromal tumor (GIST) and identify patients who may qualify for Gleevec® therapy; a multiparameter gene and protein diagnostic (Genentech) to detect HER-2 in breast cancer; and a multiplexed protein diagnostic in development to measure epidermal growth factor receptor (EGFR) in colon and lung cancer.
c-KIT is a type-3 tyrosine kinase receptor comprised of an extracellular domain, a hydrophobic transmembrane domain, and an intracellular tyrosinase domain. PATHWAY® c-KIT is a rabbit monoclonal antibody-based diagnostic directed against the C-terminal domain of the c-KIT protein that recognizes both wild-type and mutant forms of c-KIT on GIST cells with no cross-reactivity with lysates from homologous receptor tyrosine kinases. Assay validation revealed a 93% correlation with GIST tumors, according to Dr. Grogan.
Due to the heterogeneity of breast cancer, a diagnostic based on a single analyte is probably not enough, said Dr. Grogan. HER-2 is over-expressed in about 25% of breast carcinomas and is an important biomarker because it is predictive of a negative or positive response to chemotherapeutic agents, such as trastuzumab (Herceptin), and of resistance to tamoxifen.
In order to minimize variability in clinical lab-to-lab testing for the Her2/neu gene, “the ultimate goal is to combine Her2/neu gene and protein detection,” reported Dr. Grogan. Protein detection alone would yield indeterminate results about a third of the time, he added, whereas a gene-based diagnostic could definitively determine gene amplification and over-expression.
EGFR, another member of the HER tyrosine kinase family, when activated triggers a cascade of molecular events resulting in cell proliferation, angiogenesis, metastasis, and inhibition of apoptosis, and has been implicated in colon and lung carcinomas.
However, EGFR alone has not been shown to have predictive value for response or nonresponse to a particular treatment. Due to the complexity of the HER-receptor family signaling, predicting responsiveness to targeted therapies will require multiplexed, quantitative methods for measuring multiple activation pathways in parallel in tumor samples.
Targeting Rare Proteins
A key challenge in biomarker discovery is identifying and quantifying typically low-abundance proteins in complex mixtures and at relatively high-throughput. Jeffrey S. Patrick, Ph.D., a scientist in the department of integrative biology at Eli Lilly (www.lilly.com), described the use of LC/MS/MS tools for multiple reaction monitoring to measure targeted protein biomarkers. The traditional global approach to biomarker discovery has evolved to a more targeted methodology in which researchers can zoom in on peaks of interest using an extracted ion chromatogram.
Comparing the advantages and limitations of ion trap MS versus triple quadrupole (QqQ) MS for targeted proteomics, Dr. Patrick pointed out that, while ion traps are sensitive and yield a broad range of protein fragmentation making them well-suited for protein identification, they have diminished selectivity. In contrast, QqQs offer “moderate sensitivity and outstanding selectivity,” with less noise and an enhanced, focused signal. They transmit only protein fragments of defined mass to the detector and are capable of multiplexed analysis of multiple proteins or peptides simultaneously, said Dr. Patrick.
As an example of a targeted proteomics strategy, Dr. Patrick described the quantification of orosomucoid (alpha-1-acid glycoprotein) in rat serum and its role as a biomarker for parathyroid hormone (PTH) activity and bone formation. PTH stimulates bone turnover, while GSK-3 inhibitor induces bone formation in osteopenic rats. Treatment with intermittent or continuous PTH or with GSK-3 inhibitor yielded a significant change in orosomucoid levels over time compared to control animals.
Dr. Patrick demonstrated how the use of internal peptide standards on ion trap MS allowed for rapid, rough quantification of orosomucoid peptides following trypsin digestion of bone extract. Orosomucoid was not detectable in bone extract using a global proteomics approach. In contrast, the targeted proteomics method enabled biomarker detection with a sensitivity down to about 50 ng/mL of protein.
Targeted methods also enable increased cycle times (15–30 minutes versus 120 minutes for global methods), quantification, and reduced sample needs (100 nL of serum for the orosomucoid experiments). They also eliminate the need for antibody development.
Focusing on proteomics sample preparation, Lisa Bradbury, R&D director for proteomics at Pall (www.pall.com), noted that because available “technologies are not robust,” attention to detail is crucial. Proteomics experimental design should include decisions regarding how to reduce the complexity of the sample (depletion, fractionation, digestion, tagging, affinity capture, and optional clean-up) and how to analyze the sample (LC-MS, LC-MS/MS, MS/MS, MALDI/SELDI MS, 2-D gel electrophoresis, FTMS).
She emphasized the importance of evaluating the reproducibility of methodologies upfront and deciding “where you can afford to lose proteins.” In general, with each additional step, reproducibility will decrease and loss of proteins will increase. However, how much will depend on the individual method used.
Decisions regarding proteomic methods should be made in the context of the biological questions and the sample set. “For some researchers, higher throughput is more important due to the number of samples that will be required for statistically significant data, while for others, identifying as many proteins as possible is important,” said Bradbury.
Bradbury discussed Pall’s Enchant™ kits, which deplete abundant proteins from serum or plasma samples. The second generation Enchant Multi-Protein Affinity Separations Kit is intended for use with human samples only and includes ligands for specific capture of IgG and albumin. One or both ligands can be used in a particular experiment in user-defined ratios.
The materials are disposable to minimize risk of cross-contamination. Starting sample sizes for dual albumin/IgG depletion are 5–50 µL of plasma or serum. Typical depletion efficiency is 97–99%. The kits can be used in the presence of denaturants added to release bound proteins. Percent depletion remains above 97% in the presence of urea or up to 3M of CHAPS. Alternatively, users can capture albumin on the column and then add denaturants, thereby isolating the albuminome in the eluant.
Zooming in on Peptides
The carrier protein hypothesis and the field of albuminomics focus attention on albumin-bound peptides, which appear to have a longer circulating half-life than unbound peptides and could serve as a source of valuable biomarkers. Dr. Patton of PerkinElmer described the types of proteomic tools being used in ongoing work to identify albumin-bound peptide biomarkers for early detection of ovarian cancer.
Studies of protein expression comparing serum samples from patients with and without stage I ovarian cancer revealed both up and down regulation of peptides associated with the coagulation cascade—likely resulting from general inflammatory processes associated with the presence of cancer. The peptides identified were not tumor-derived.
The challenge is to “get underneath the coagulation cascade and start exploring tumor-associated peptides,” said Dr. Patton. This requires selective enrichment of low-abundance peptides from the serum. By using selective affinity capture methods interfaced with MS/MS, researchers can target the phosphopeptide component of the albuminome.
PerkinElmer’s high-throughput workflow includes ProXPRESSION™ and Phos-trap™ Proteomics Enrichment Kits for microscale fractionation of complex protein and peptide mixtures based on the company’s membrane adsorber and magnetic bead technologies, respectively. The ProXPRESSION utilizes Cibacron Blue for selective adsorption of carrier proteins, while the Phos-trap™ kit is designed for enrichment of phosphorylated peptides.
“Phos-trap is about three orders of magnitude more sensitive than other tools on the market,” said Dr. Patton. Based on metal oxide affinity chromatography, it incorporates a thin-film titanium coating interfaced with magnetic beads to achieve high capacity and selectivity. The 96-well format is compatible with automated liquid-handling instruments, and the kit can enrich for and detect fewer than 100 fmol of phosphopeptides.
Hans-Dieter Zucht, Ph.D., CTO of Digilab BioVisioN (www.biovision-discovery.de), discussed the characterization of protease inhibitors.
Proteolytic processing of peptides by proteases, resulting in peptide degradation and maturation, is one of the most important forms of post-translational modifications, according to Dr. Zucht. The peptidomes are synthesized as precursor molecules, which incorporate modules, soluble mediators, and tethered ligands.
In their role as bioactive messengers, peptides, such as hormones, growth factors, and cytokines, are ideal candidates for biomarkers that reflect the proteolytic process, Dr. Zucht said. Examples of precursor molecules would include proinsulin and procalcitonin. Proteolysis may generate or destroy a peptide’s bioactivity, render it receptor-specific, or cleave modules from the peptide, for example.
Dr. Zucht asserted that peptidomics and an understanding of proteolytic pathways and networks, combined with bioinformatic tools, can help identify components of target molecules that may improve drug activity or pharmacokinetics when cleaved off or that should not be removed due to the risk of introducing unwanted side effects.
BioVisioN’s Peptidomics® technology includes peptide profiling (sample prep, HPLC, and MALDI-ToF MS), Differential Peptide Display® (bioinformatics, data, mining, and statistical analysis), and peptide identification (multidimensional chromatography, nano-electrospray ionization quadrupole time-of-flight tandem MS, MALDI-ToFToF-MS/MS, or Edman degradation sequencing). The platform can perform peptide extraction on sample volumes 50 µL to 5 mL and has a limit of detection of 100 x 10–12 mol/L.