| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
1 Department of Laboratory Medicine, Children’s Hospital, 2 Cardiology Division, and Center for Immunology, and Inflammatory Diseases, Massachusetts General Hospital, and the Departments of, 3 Pathology and 4 Medicine, Harvard Medical School, Boston, MA
aAddress correspondence to this address at: Department of Laboratory Medicine, Children’s Hospital, 300 Longwood Avenue, Boston, MA 02115.
Currently available biomarkers, such as those used to diagnose myocardial infarction (e.g., cardiac troponins), were identified in the course of targeted physiologic studies. Similarly, basic investigation of diseases has largely been characterized by studies of isolated molecules in cellular systems. Advances in genomics technologies, however, are beginning to permit characterization of global alterations associated with disease conditions and, in the process, identification of novel biomarkers and pathways. Laterza et al.(1) used such a global survey, and in this issue of Clinical Chemistry they describe their discovery of potential new markers related specifically to brain injury. Now begins the long road toward validation of these markers in clinically relevant human cohorts. Although no serum biomarkers of brain injury for conditions such as stroke are in clinical use, they could ultimately prove to be of enormous clinical usefulness to complement physical and radiologic examinations, both of which can be ambiguous during acute presentations.
Of the multiple genomics applications, perhaps none has garnered more recent attention for biomarker discovery than proteomics. Proteomics offers unique insight into disease because proteins and their bioenzymatic functions largely determine the phenotypic diversity that can arise from a set of common genes. Posttranslational modifications help regulate structure, function, localization, maturation, and turnover of proteins. Because the entire complement of expressed proteins in their various forms can rapidly change in response to environmental cues, the proteome represents the unique ensemble of proteins that reflects the state of the cell or group of cells at a given time, in a particular context under particular stimuli. Thus, the proteome is highly dynamic, in contrast to the stability of the genome.
The one gene–one protein dictum, now no longer tenable, had led few to anticipate the immense magnitude and complexity of the resulting proteome. This complexity, however, is the basis of both great informative potential and analytical challenge, particularly as it applies to the study of human blood to mine for novel biomarkers. The plasma proteome is unique in that it does not represent a particular cellular genome, but instead reflects the collective expression of all cellular genomes. It has thus far been poorly characterized. Twenty-two of the most abundant proteins, including albumin and the immunoglobulins, comprise 99% of the serum proteome mass(2). Many of the biologically interesting molecules relevant to human disease are low-abundance proteins. For example, cardiac markers such as troponin are found in the nanomolar range, insulin in the picomolar range, and tumor necrosis factor 
in the femtomolar range. In all, there are an estimated 10 000 unique proteins in serum, with concentrations spanning a dynamic range of more than 10 orders of magnitude. Because serum has been found to include not only expected circulating proteins, such as albumin and immunoglobulins, but also less expected proteins from all functional classes and cellular localizations, it is hypothesized that the plasma proteome contains representatives of the entire set of more than 300 000 estimated human polypeptide species resulting from splice variants and posttranslational modifications(2).
Present-day proteomics instrumentation cannot yet tackle the sensitivity and dynamic range issues necessary for high-throughput plasma mining, and thus proteomic approaches to date have failed to identify low-abundance markers(3)(4)(5). Encouraging recent advances, however, particularly in the field of mass spectrometry, suggest that proteomics may soon realize its enormous potential for the field of biomarker discovery.
Perhaps because of the daunting complexity of plasma profiling, as well as the recent successful application of transcript profiling to the field of cancer biomarkers(6), Laterza et al.(1) opted for the latter, more mature technology for the identification of novel biomarkers of brain injury. Using transcript profiling of isolated organ preparations, Laterza et al. identified mRNAs that were highly enriched in the brain. It must be noted, however, that the authors used an agnostic strategy regarding potentially important response pathways that are elaborated by other tissues in response to brain injury. Another concern is that only healthy brain tissue was profiled; better biomarker expression might occur in chronically ischemic brain tissue. Despite such potential limitations, the group was able to confirm protein expression for a number of the potential brain-specific biomarkers via Western blots of various tissue homogenates, often with antibodies that they themselves raised. During the early verification process that used these initial reagents, one potential biomarker identified from the screen, visinen-like protein 1 (VLP-1), was found to increase in patient serum after ischemic stroke. Thus, more extensive analysis of this candidate and others in their screen is warranted in future investigations.
After the discovery and verification of the candidate proteins, robust immunoassays must be developed and optimized to evaluate their potential clinical utility. Individual sandwich-based immunoassays using either monoclonal or polyclonal antibodies and nonisotopic labeled antibodies (e.g., alkaline phosphatase, fluorescein, and ruthenium) are used in this process. Although multiplexing technology is designed to simultaneously evaluate several putative biomarkers, at present the optimization of multiple protein assays is seldom achieved(7). Laterza et al. developed an ELISA that enabled the measurement of VLP-1 at concentrations <100 ng/L(1). An initial step in the optimization process is establishing the specificity of available antibodies, using Western blot and immunostaining or other suitable techniques. Variables that affect the performance characteristics, such as the avidity and concentration of the capture and detection antibodies, incubation time and temperature, sample volume and dilution, pH, composition and concentration of the buffer, and the quality of the detecting system, must be closely examined(8). The calibrators must be prepared in a manner similar to that used for patient sample, and the appropriate curve-fitting model for calculation should be determined(9). Only after the assay is completely optimized can analytical performance be properly assessed.
The analytical evaluation comprises several measures including trueness, accuracy, repeatability, and reproducibility, as well as a determination of linearity and limits of detection and quantification. Specific protocols for the assessment of these parameters are available to ensure a proper analytical performance evaluation of the new method(10)(11)(12)(13). Furthermore, the frequency distribution of the candidate protein must be examined in healthy individuals to establish reference intervals to which patient results will be compared(14)(15). Evaluation of the candidate biomarker in control populations will determine whether the distribution of the protein is gaussian or skewed, and whether significant differences in values exist among different age, sex, or racial subgroups. Such information is essential in determining how the reference intervals will be established.
In addition, characterization and control of the preanalytical variability is required. Preanalytical variability encompasses both physiologic [within- and between-subject variation, circadian rhythm, and effects of age, sex, ethnicity, drugs, diseases, and lifestyle factors (smoking, alcohol intake, diet, exercise, and obesity)] and nonphysiologic components [e.g., patient preparation (supine or standing, fasting or nonfasting), specimen type (serum or plasma), and storage conditions]. Standardization of sample collection procedures and appreciation of the limitation of the assay in various physiologic and pathologic conditions will minimize this variability.
Laterza et al.(1) have indicated that, although their findings regarding VLP-1 and stroke are promising, characterizing the clinical usefulness of this marker will be an extensive process. The diagnostic performance of a novel test involves the assessment of its diagnostic accuracy (sensitivity, specificity, likelihood ratio, and ROC curve) and predictability (positive and negative predictive values) in a series of studies, [Phase I (exploratory phase) to IV (outcome phase)], that emphasize different performance characteristics and require different study populations(16).
Once the clinical usefulness of a novel marker has been demonstrated, in vitro diagnostics (IVD) companies will decide whether to pursue it commercially. Technical, medical, financial, and legal considerations will also influence this decision(17). Major IVD companies have immunoanalyzer platforms that enable the transfer of a research immunoassay to an automated one that meets basic clinical laboratory requirements (such as robustness, reagent stability, ease of use, acceptable turnaround time, adaptability to automated immunoassay systems, and low cost). The IVD company will then evaluate the new assay and generate the required validation data to satisfy the regulatory agencies.
A new IVD device must be cleared by the FDA before it enters the US market. For a novel protein to which no predicate device can be identified, FDA premarket approval, rather than 510(k) clearance, is required. However, if the FDA determines that the intended clinical use of the novel marker does not warrant its placement in the riskiest group (Class III, such as cancer diagnostics), then a new, simpler process—de novo or "risk-based" process—can be used(18). This process allows a new analyte to be regulated as in a 510(k) but requires the demonstration of clinical effectiveness. Brain natriuretic peptide, for example, was cleared by the FDA through this new process. Most likely, such a test will be classified according to CLIA as a high complexity test(19).
In summary, the report by Laterza et al. describes a logically constructed approach to the daunting task of applying gene expression techniques to the first steps in the identification of biomarkers. Consideration of the remaining steps underscores the tenuous road to discovery of biomarkers that are useful, a road that will be better illuminated when the complementary powers of transcriptional profiling and proteomics reach similar maturity. Their efforts to date suggest tantalizing novel markers of diseases of the brain and the beginning of a long journey toward the identification of a clinical biomarker and the arduous transition from the research environment to routine clinical practice.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
