Proteomic biomarkers inform on diverse areas of biology, from basic research to disease diagnostics and drug approval.

For years the U.S. Food and Drug Administration defined “biomarker” rather narrowly, to suit the purposes of drug discovery and diagnostics organizations under its regulatory umbrella. Yet a whole other world existed in NIH-funded nonclinical research, whose scientific discoveries have led—directly but also indirectly—to the development of so many drugs and diagnostics. So, in 2015, the joint FDA/NIH Biomarkers, EndpointS, and other Tools task force (BEST) promulgated a definition that, at least to the agencies, clarified “language confusion” responsible for misinterpretation of biomarker-related work.

BEST defines biomarker as “a defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or responses to an exposure or intervention, including therapeutic interventions. Molecular, histologic, radiographic, or physiologic characteristics are types of biomarkers. A biomarker is not an assessment of how a patient feels, functions, or survives.” This last point underscores what FDA knows but medical practitioners sometimes confuse: the difference between the presence or absence of a biomarker and the actual disease, which is almost always more complex.

FDA’s 2018 draft guidance, Biomarker Qualification: Evidentiary Framework, explains the Agency’s expectations for qualifying or validating biomarkers used in human diagnostics, and addresses the “lack of a clear, predictable, and specific regulatory framework for the type and level of evidence sufficient to support regulatory decision-making using biomarkers.” According to FDA, developers should base their validation/qualification efforts on biological rationale, data supporting the relationship between a biomarker and clinical outcome, and analytics. Under this model FDA clearly distinguishes between scientific validity and method robustness.

Why proteomics?

Biomarkers exist everywhere within organisms, and may be found in any biological specimen through proteomic, genomic, or metabolomic investigation. “DNA markers are the easiest to identify,” Hugo Gagnon, Ph.D., CEO at PhenoSwitch Bioscience, tells Biocompare, “but only a fraction of DNA transcribes to RNA, and only a fraction of that is put into action as protein. The correlation between RNA and protein is especially poor in many disease states and fast-growing cells. So your problem, with genomics, is that potential genomic biomarkers do not reflect what’s going on in your system.”

Metabolites have been touted as most representative of gene and protein activity. “But metabolites appear and disappear very quickly, and are more difficult to work with than proteins in terms of stability, complexity, and dynamics,” Gagnon says. “This holds even when considering epigenetics for genes, and variation in conformation and post-translational modification for proteins, which can profoundly affect protein activity.”

Proteomics has several advantages over genomics or metabolomics for biomarker development. Only a fraction of DNA sequences contained in a cell are transcribed to RNA, and an even smaller number eventually undergo translation. Noncoding proteins and protein fragments may nevertheless provide insights into the efficacy or toxicity of a candidate drug, or on the status of patients undergoing treatment. But for the coding region, the bad news is that the 30,000 or so sequences destined for proteinhood exist as more than 1 million variants, including sequences that differ only in post-translational modifications.

Despite this known source of variability, biologists remain interested in the proteome because it reflects the functional or operational role genes play, and as such may be thought of —similarly to metabolomics—as a more accurate “snapshot” of a cell or organism’s status. Unraveling the complexity arising from isoforms is no easy task as the existence and prevalence of proteomic biomarkers, like the underlying disease itself, is the result of complex, dynamic processes that include protein trafficking, localization, and protein-protein interactions.

Low success rate

Validation occurs to demonstrate that a process, method, or piece of equipment is suitable for its intended use, and robust enough to withstand a certain amount of deviation from normal. Validation related to processes or engineering takes time and resources, but is usually successful. Proteomic biomarker validation also occurs to prove suitability and robustness, but failure is much more common than for validations related to equipment or processes. Proteomics and genomics studies have uncovered thousands of putative biomarkers. One notable success, prostate-specific antigen, has been used for years to guide the treatment of prostate cancer. Yet fewer than 100 biomarkers have been validated for clinical applications.

Mohammed Sharif, who studies osteoarthritis and rheumatoid arthritis biomarkers at the University of Bristol, has noted that of more than 1,000 putative biomarkers discovered for those conditions not one has achieved regulatory approval for clinical use. Sharif explains the reasons:

  • Lack of standardization and characterization of samples (e.g. collection/storage conditions, clinical/demographic data)
  • Difficulties generating sensitive antibodies for development of immunoassays to support validation
  • Lack of commercial affinity reagents and antibodies
  • Nonreproducibility among proteomic platforms
  • Error introduced by sample processing (e.g. depletion, enrichment, fractionation). “Using poorly characterised samples for proteomic analysis is one of the main reasons proteomic biomarkers have not progressed beyond the initial discovery phase.”

Sharif bristles at the notion that these deficiencies may be overcome through adherence to “good experimental practices,” which include appropriate sample-handling and harmonization of analytical platforms.

Validating proteomic biomarkers involves many technical difficulties compounded by a lack of understanding of underlying disease mechanisms, he says. “Osteoarthritis is a very complex disease. Different studies have applied different characteristics of OA to define the condition.”

One group of studies focuses on structural damage to joints, another focuses on pain or activities of daily living, and others still focus on biomarkers. “Selection bias exists in many studies, while others have difficulty categorizing asymptomatic patients.”

Challenges in validating proteomic biomarkers for diagnostics spills over into drug development itself, as biomarkers that eventually might be suited to the former are often used to screen or optimize candidate drugs. “There are currently no disease-modifying drugs for osteoarthritis, and one of the main reasons for this is a lack of suitable biochemical tests to monitor the efficacy of new drugs,” Sharif adds.

Reducing target bias

Because liquid chromatography and mass spectrometry have been used for decades to analyze proteins and protein fragments, proteomics and its tools are also much further developed than their genomic or metabolomic counterparts. On the minus side, proteomic concentration dynamic range is as high as 1010. While expression levels are also an issue for genomics, known genetic sequences are easily amplified by PCR. No analogous work-around exists for low-abundance proteins (or metabolites).

Although FDA’s definition of biomarker includes visual and imaging endpoints, biochemical biomarkers are currently all the rage. “The trend today is to look for them in easily accessible bodily fluids,” explains Gagnon. Phenoswitch specializes in LC-MS/MS method development for studying and validating ‘omics biomarkers from plasma, urine, and cerebrospinal fluid. Proteomic biomarkers are also found through tissue biopsies and cell fractionation.

Gagnon describes LC-MS/MS as “the only approach to proteomics that does not introduce bias in terms of target selection.” While multiplexed technologies (e.g. protein arrays, bead based arrays) are more and more seen as proteomics technologies, they rely on antibody-target capture, which limits the number of targets per assay and introduces antibody affinity bias. Conversely, label-free LC-MS based proteomics can provide robust and bias-free relative quantification of thousands of low-abundance proteins.

Antibody-based proteomics biomarker validation can be much more sensitive, but its realization occurs at much lower throughput than LC-MS/MS, and it requires a separate discovery and validation effort for each antibody used.

Hero image: In rheumotoid arthritis, the body's immune system attacks its own tissue, including joints. Image from Dreamstime.com