May 1, 2013 (Vol. 33, No. 9)
Josh P. Roberts
Find out what new research is being done with biomarkers in this article from our May 1 issue.
Biomarkers by definition indicate some state or process that generally occurs at a spatial or temporal distance from the marker itself, and it would not be an exaggeration to say that biomedicine has become infatuated with them: Where to find them, when they may appear, what form they may take, and how they can be used to diagnose a condition or predict whether a therapy may be successful.
Biomarkers are on the agenda of many if not most industry gatherings, and in cases such as Oxford Global’s recent “Biomarker Congress” and the GTC “Biomarker Summit”, they hold the naming rights. There, some basic principles were built upon, amended, and sometimes challenged.
In oncology, for example, biomarker discovery is often predicated on the premise that proteins shed from a tumor will traverse to and persist in, and be detectable in, the circulation. By quantifying these proteins—singularly or as part of a larger “signature”—the hope is to garner information about the molecular characteristics of the cancer that will help with cancer detection and personalization of the treatment strategy.
Yet this approach has not yet turned into the panacea that was hoped for. Bottlenecks exist in affinity reagent development, platform reproducibility, and sensitivity. There is also a dearth of understanding of some of the fundamental principles of biomarker biology that we need to know the answers to, said Parag Mallick, Ph.D., whose lab at Stanford University is “working on trying to understand where biomarkers come from.” And sometimes, too, accepted wisdom just isn’t so.
For example, there are dogmas saying that circulating biomarkers come solely from secreted proteins. But Dr. Mallick’s studies indicate that fully 50% of circulating proteins may come from intracellular sources or proteins that are annotated as such. “Right now we don’t understand the processes governing which tumor-derived proteins end up in the blood.”
Other seemingly obvious questions include “how does the size of a tumor affect how much of a given protein will be in the blood?”—perhaps the tumor is necrotic at the center, or it’s hypervascular or hypovascular. “The problem is that these are highly nonlinear processes at work, and there is a large number of factors that might affect the answer to that very simple question,” he pointed out.
Their research focuses on using mass spectrometry and computational analysis to characterize the biophysical properties of the circulating proteome, and relate these to measurements made of the tumor itself.
“We’ve observed that the proteins that are likely to first show up and persist in the circulation, on average, are more stable than proteins that don’t,” Dr. Mallick said. “This is something that people qualitatively suspected, but now we can quantify how significant the effect is.”
The goal is ultimately to be able to build rigorous, formal mathematical models that will allow something measured in the blood to be tied back to the molecular biology taking place in the tumor. And conversely, to use those models to predict from a tumor forward to what will be found in the circulation. “Ultimately, the models will allow you to connect the dots between what you measure in the blood and the biology of the tumor.”
Bound for Affinity Arrays
Affinity reagents are the main tools for large-scale protein biomarker discovery. And while this has tended to mean antibodies (or their derivatives), other affinity reagents are demanding a place in the toolbox.
Among these are Affimers, a type of affinity reagent being developed by Avacta. Affimers consist of a biologically inert, biophysically stable protein scaffold containing three variable regions into which distinct peptides are inserted. The resulting three-dimensional surface formed by these peptides interacts and binds to proteins and other molecules in solution, much like the antigen-binding site of antibodies.
Unlike antibodies, Affimers are relatively small (13 KDa), non-post-translationally modified proteins that can readily be expressed in bacterial culture. They may be made to bind surfaces through unique residues engineered onto the opposite face of the Affimer, allowing the binding site to be exposed to the target in solution. “We don’t seem to see in what we’ve done so far any real loss of activity or functionality of Affimers when bound to surfaces—they’re very robust,” said CEO Alastair Smith, Ph.D.
Avacta is taking advantage of this stability and its large libraries of Affimers to develop very large affinity microarrays for drug and biomarker discovery. To date they have printed arrays with around 20–25,000 features, and Dr. Smith is “sure that we can get toward about 50,000 on a slide,” he said. “There’s no real impediment to us doing that other than us expressing the proteins and getting on with it.”
Customers will be provided with these large, complex “naïve” discovery arrays, readable with standard equipment. The plan is for the company to then “support our customers by providing smaller, bespoke, arrays with the Affimers that are binding targets of interest to them,” Dr. Smith foretold. And since the intellectual property rights are unencumbered, Affimers in those arrays can be licensed to the end users to develop diagnostics that can be validated as time goes on.
Around 20,000-Affimer discovery arrays were recently tested by collaborator Professor Ann Morgan of the University of Leeds with pools of unfractionated serum from patients with symptoms of inflammatory disease. The arrays “rediscovered” elevated C-reactive protein (CRP, the clinical gold standard marker) as well as uncovered an additional 22 candidate biomarkers. Some of the latter, when combined with CRP, appear able to distinguish between different diseases such as rheumatoid arthritis, psoriatic arthritis, SLE, or giant cell arteritis.
Sometimes biomarkers are used not to find disease but to distinguish healthy human cell types, with perhaps the most obvious examples being found in flow cytometry and immunohistochemistry. These widespread applications, however, are difficult to standardize, being subject to arbitrary or subjective gating protocols and other imprecise criteria.
Epiontis instead uses an epigenetic approach. “What we need is a unique marker that is demethylated only in one cell type and methylated in all the other cell types,” explained CBO and founder Ulrich Hoffmueller, Ph.D. Each cell of the right cell type will have two demethylated copies of a certain gene locus, allowing them to be enumerated by quantitative PCR.
The biggest challenge is finding that unique epigenetic marker. To do so they look through the literature for proteins and genes described as playing a role in the cell type’s biology, and then look at the methylation patterns to see if one can be used as a marker, Dr. Hoffmueller said. They also “use customized Affymetrix chips to look at the differential epigenetic status of different cell types on a genomewide scale.”
The company currently has a panel of 12 assays for 12 immune cell types. Among these is an assay for regulatory T (Treg) cells that queries the Foxp3 gene—which is uniquely demethylated in Treg even though it is transiently expressed in activated T cells of other subtypes. Also assayed are Th17 cells, “which are especially tricky to detect by flow cytometry because the cells have to be stimulated in vitro,” he pointed out.
Developing New Assays for Cancer Biomarkers
Researchers at Myriad RBM and the Cancer Prevention Research Institute of Texas are collaborating to develop new assays for cancer biomarkers on the Myriad RBM Multi-Analyte Profile (MAP) platform.
The release of OncologyMAP 2.0 expanded Myriad RBM’s biomarker menu to over 250 analytes, which can be measured from a small single sample, according to the company. Using this menu, L. Stephen et al., published a poster, “Analysis of Protein Biomarkers in Prostate and Colorectal Tumor Lysates,” which showed the results of a survey of proteins relevant to colorectal (CRC) and prostate (PC) tumors to identify potential proteins of interest for cancer research.
The study looked at CRC and PC tumor lysates and found that 102 of the 115 proteins showed levels above the lower limit of quantification. Four markers were significantly higher in PC and 10 were greater in CRC. For most of the analytes, duplicate sections of the tumor were similar, although some analytes did show differences. In four of the CRC analytes, tumor number four showed differences for CEA and tumor number 2 for uPA.
Thirty analytes were shown to be different in CRC tumor compared to its adjacent tissue. Ten of the analytes were higher in adjacent tissue compared to CRC. Eighteen of the markers examined demonstrated significant correlations of CRC tumor concentration to serum levels.
“This suggests that they would be good protein markers to follow changes in tumor levels,” reported the researchers, who added that the Oncology MAP 2.0 platform “provides a good method for studying changes in tumor levels because many proteins can be assessed with a very small sample.”
Clinical Test Development with MALDI-ToF
Circulating proteins provide a wealth of potentially useful information that can be used for clinically relevant diagnostic tests. While there have been many attempts to translate results from early discovery work on the serum proteome into clinical practice, few of these efforts have progressed past the discovery phase.
Matrix-assisted laser desorption/ionization-time of flight (MALDI-ToF) mass spectrometry on unfractionated serum/plasma samples offers many practical advantages over alternative techniques, and does not require a shift from discovery to development and commercialization platforms. While MALDI still has had some hurdles to overcome, Biodesix claims it has been able to develop the technology into a reproducible, high-throughput tool to routinely measure protein abundance from serum/plasma samples.
“In our approach to develop tests, we improved data-analysis algorithms to reproducibly obtain quantitative measurements of relative protein abundance from MALDI-ToF mass spectra. These measurements are combined with clinical outcome data using modern learning theory techniques to define specific disease states based on a patient’s serum protein content,” says Heinrich Röder, CTO.
The clinical utility of the identification of these disease states can be investigated through a retrospective analysis of differing sample sets. For example, Biodesix clinically validated its first commercialized serum proteomic test, VeriStrat®, in 85 different retrospective sample sets. “It is becoming increasingly clear that the patients whose serum is characterized as VeriStrat Poor show consistently poor outcomes irrespective of tumor type, histology, or molecular tumor characteristics,” Röder adds.
MALDI-ToF mass spectrometry, in its standard implementation, allows for the observation of around 100 mostly high-abundant serum proteins. “While this does not limit the usefulness of tests developed from differential expression of these proteins, the discovery potential would be greatly enhanced if we could probe deeper into the proteome while not giving up the advantages of the MALDI-ToF approach,” he explains.
Biodesix reports that its new MALDI approach, Deep MALDI™, can perform simultaneous quantitative measurement of more than 1,000 serum protein features (or peaks) from 10 µL of serum in a high-throughput manner. The technology increases the observable signal noise ratio from a few hundred to over 50,000, resulting in the observation of many lower-abundance serum proteins.
Epigenetic qPCR requires only about 2 mL of blood, and samples can be frozen and shipped—giving it a huge logistical advantage for large, multicenter clinical trials. “For about half of the top 20 pharma, we provide this as an immune-monitoring service for clinical trial samples,” Dr. Hoffmueller said.
Molecular signatures are sought out to distinguish among seemingly similar maladies, their discovery performed by looking at samples from patients with known infirmities and outcomes and comparing what can be garnered with data about the patients themselves. Nothing—from lifestyle choices to birth order to ancestral genetics and epigenetics—is off the table. It can take enormous datasets, sophisticated algorithms, and vast computational resources to tease out correlations that may prove useful in the clinic.
Take breast cancer, a disease now considered to be a collection of many complexes of symptoms and signatures—the dominant ones being labeled Luminal A, Luminal B, Her2, and Basal—that suggest different etiologies and offer different prognoses. Yet even these labels are lately considered too simplistic for understanding and managing a woman’s cancer.
Studies published in the past year have looked at somatic mutations, gene copy number aberrations, gene expression abnormalities, protein and miRNA expression, and DNA methylation, coming up with a list of significantly mutated genes—hot spots—in different categories of breast cancers. Targeting these will inevitably be the focus of much coming research.
“We’ve been taking these large trials and profiling these on a variety of array or sequence platforms. We think we’ll get not only prognostic drivers (which are only good if you know what to do with them) but also predictive markers for taxanes and monoclonal antibodies and tamoxifen and aromatase inhibitors,” explained Brian Leyland-Jones, Ph.D., director of Edith Sanford Breast Cancer Research. “We will end up with 20–40 different diseases, maybe more.”
Edith Sanford Breast Cancer Research is undertaking a pilot study in collaboration with The Scripps Research Institute, using a variety of tests on 25 patients to see how the information they provide complements each other, the overall flow, and the time required to get and compile results.
Laser-captured tumor samples will be subjected to low passage whole-genome, exome, and RNA sequencing (with targeted resequencing done in parallel), and reverse-phase protein and phosphorylation arrays, with circulating nucleic acids and circulating tumor cells being queried as well. “After that we hope to do a 100- or 150-patient trial when we have some idea of the best techniques,” he said.
Dr. Leyland-Jones predicted that ultimately most tumors will be found to have multiple drivers, with most patients receiving a combination of two, three, or perhaps four different targeted therapies.
Reduce to Practice
Once biomarkers that may have an impact on therapy are discovered, it is not always routine to get them into clinical practice. Leaving regulatory and financial, intellectual property and cultural issues aside, developing a diagnostic based on a biomarker often requires expertise or patience that its discoverer may not possess.
Andrew Gribben is a clinical assay and development scientist at Randox Laboratories, based in Northern Ireland, U.K. The company utilizes academic and industrial collaborators together with in-house discovery platforms to identify biomarkers that are augmented or diminished in a particular pathology relative to appropriate control populations. Biomarkers can be developed to be run individually or combined into panels of immunoassays on its multiplex biochip array technology.
“Individual biomarkers alone may not be specific enough to support a diagnostic claim,” he noted.
Specificity can also be gained—or lost—by the affinity of reagents in an assay. The diagnostic potential of Heart-type fatty acid binding protein (H-FABP) abundantly expressed in human myocardial cells was recognized by Jan Glatz of Maastricht University, The Netherlands, back in 1988. Levels rise quickly within 30 minutes after a myocardial infarction, peaking at 6–8 hours and return to normal within 24–30 hours. Yet at the time it was not known that H-FABP was a member of a multiprotein family, with which the polyclonal antibodies being used in development of an assay were cross-reacting, Gribben related.
Randox developed monoclonal antibodies specific to H-FABP, funded trials investigating its use alone, and multiplexed with cardiac biomarker assays, and, more than 30 years after the biomarker was identified, in 2011, released a validated assay for H-FABP as a biomarker for early detection of acute myocardial infarction.
Ultrasensitive Immunoassays for Biomarker Development
Research has shown that detection and monitoring of biomarker concentrations can provide insights into disease risk and progression. Cytokines have become attractive biomarkers and candidates for targeted therapies for a number of autoimmune diseases, including rheumatoid arthritis (RA), Crohn’s disease, and psoriasis, among others.
However, due to the low-abundance of circulating cytokines, such as IL-17A, obtaining robust measurements in clinical samples has been difficult.
Singulex reports that its digital single-molecule counting technology provides increased precision and detection sensitivity over traditional ELISA techniques, helping to shed light on biomarker verification and validation programs.
The company’s Erenna® immunoassay system, which includes optimized immunoassays, offers LLoQ to femtogram levels per mL resolution—even in healthy populations, at an improvement of 1-3 fold over standard ELISAs or any conventional technology and with a dynamic range of up to 4-logs, according to a Singulex official, who adds that this sensitivity improvement helps minimize undetectable samples that could otherwise delay or derail clinical studies.
The official also explains that the Singulex solution includes an array of products and services that are being applied to a number of programs and have enabled the development of clinically relevant biomarkers, allowing translation from discovery to the clinic.
In a poster entitled “Advanced Single Molecule Detection: Accelerating Biomarker Development Utilizing Cytokines through Ultrasensitive Immunoassays,” a case study was presented of work performed by Jeff Greenberg of NYU to show how the use of the Erenna system can provide insights toward improving the clinical utility of biomarkers and accelerating the development of novel therapies for treating inflammatory diseases.
A panel of inflammatory biomarkers was examined in DMARD (disease modifying antirheumatic drugs)-naïve RA (rheumatoid arthritis) vs. knee OA (osteoarthritis) patient cohorts. Markers that exhibited significant differences in plasma concentrations between the two cohorts included CRP, IL-6R alpha, IL-6, IL-1 RA, VEGF, TNF-RII, and IL-17A, IL-17F, and IL-17A/F. Among the three tested isoforms of IL-17, the magnitude of elevation for IL-17F in RA patients was the highest.
“Singulex provides high-resolution monitoring of baseline IL-17A concentrations that are present at low levels,” concluded the researchers. “The technology also enabled quantification of other IL-17 isoforms in RA patients, which have not been well characterized before.”
The Singulex Erenna System has also been applied to cardiovascular disease research, for which its cardiac troponin I (cTnI) digital assay can be used to measure circulating levels of cTnI undetectable by other commercial assays.
Recently presented data from Brigham and Women’s Hospital and the TIMI-22 study showed that using the Singulex test to serially monitor cTnI helps stratify risk in post-acute coronary syndrome patients and can identify patients with elevated cTnI who have the most to gain from intensive vs. moderate-dose statin therapy, according to the scientists involved in the research.
The study poster, “Prognostic Performance of Serial High Sensitivity Cardiac Troponin Determination in Stable Ischemic Heart Disease: Analysis From PROVE IT-TIMI 22,” was presented at the 2013 American College of Cardiology (ACC) Annual Scientific Session & Expo by R. O’Malley et al.