- Email: [email protected]
When does a fingerprint constitute a diagnostic?
Comment
9
10
Nashef SAM, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999; 16: 9–13. Ambler G, Omar RZ, Royston P, et al. Valvular heart disease: generic, simple risk stratification model for heart valve surgery. Circulation 2005; 112: 224–31.
11 12
Baumgartner H. Aortic stenosis: medical and surgical management. Heart 2005; 91: 1483–88. Vahanian A, Palacios IF. Percutaneous approaches to valvular disease. Circulation 2004; 109: 1572–79.
When does a fingerprint constitute a diagnostic? Diagnosis of disease via biomarkers in circulating blood depends on several factors. Markers must be detectable during the preliminary stages of disease, or at least when successful treatment is possible. Tests should be cost effective and readily available at point-of-care, such that the turnaround time for diagnosis is minimised. Preferably, diagnosis should be minimally invasive and, most importantly, must be absolutely specific to the disease to minimise misguided or inadequate therapy. In today’s Lancet, Dan Agranoff and colleagues describe the use of surface-enhanced laser desorption-ionisation time-of-flight mass spectrometry (SELDI-TOF) to identify serum biomarkers from patients with advanced Mycobacterium tuberculosis infections.1
Serum proteomics with this technique provides a pattern-based diagnosis, because multiple peptide and protein signals are detected simultaneously in a single mass spectrum. Because the detector is based on well-established matrix-assisted (MALDI-TOF) massspectrometry methods, there is a bias toward lower mass signals (1–10 kDa), and the approach is often termed peptidomics.2 Furthermore, disease states seem to be characterised by raised levels of cleavage products of abundant serum proteins.3,4 Alternatively, MALDI-TOF mass-spectrometry profiling, which provides spectra of higher resolution, allows for the improved acquisition of higher mass ranges, circumventing the limitations of low mass profiling (figure). In the earliest iterations of
Published Online September 14, 2006 DOI:10.1016/S01406736(06)69379-3 See Articles page 1012
Patient 1
Patient 2
Diseased
Patient 3
Control 1 Control Control 2
Control 3 3000
4000
5000
6000
7000
8000
9000
10 000 m/z
10000
12000
14 000
16000
18000
Increasing molecular weight 1 kDa
10 kDa
20 kDa
Peptidomic or peptide-centric profiling
Proteomic or whole protein-centric profiling
Applications of monitoring low mass range: Naturally occurring peptides Small intact proteins Biological cleavage products Artifacts from sample processing
Applications of monitoring high mass range: Intact or nearly intact proteins Protein isoforms Post-translational modifications Ratios of intact:truncated forms
50 kDa
Figure 1: MALDI-TOF mass spectra of fractionated human serum used for peptide and protein profiling Examples for both low mass and high mass profiling shown. m/z=peptide or protein mass divided by charge state, induced by ionisation.
www.thelancet.com Vol 368 September 16, 2006
971
Comment
SELDI-TOF mass spectrometry for biomarker discovery, identification of the individual components within the signature was not possible,5 although later studies overcame this difficulty with much better algorithmic interpretation of the resulting peptide masses and concurrent use of tandem mass spectrometry for aminoacid sequencing.4 The capital cost of equipment and assays are such that SELDI-TOF mass spectrometry is unlikely to be routinely and, therefore, identification of a suite of markers from within the signature is needed to provide a test amenable to frequent low-cost use. The advantage of SELDI-TOF mass spectrometry is that it does not rely on evidence of a gold-standard biomarker, but rather on combinations of peptide signals. The disease signature given by the serum sample in the mass spectrometer, and compared against a database of controls and test samples, is the potential diagnostic at the preclinical stage. Mass spectrometry is a bioanalytical technique with almost unparallelled sensitivity, and thus provides an ability to mine substantially deeper into the serum or plasma proteomes than has previously been possible. Unfortunately, substantial caveats remain (table).6 The use of serum or plasma is challenging technically. More than 3000 proteins are contained Reasoning Scientifically valid applications Validation studies
Initial screening of biological samples*
Fractionation of samples where proteome is not dominated by subset of proteins (eg, CSF, urine, and tears)† Applications with limited practicality Biomarker discovery in complex biological samples†
Post-translational modification analysis of proteins altered with disease Search for low abundance biomarkers Identification of proteins/peptides Precise protein/peptide quantitation
Antibody labelling of surface to specifically isolate specific proteins and protein complexes Multiple surface chemistries Identify peptide/protein profile changes (presence/ absence) Fast reproducible results Simple sample fractionation Multiple surface chemistries
Interpretation of meaningful data from simple mass spectrometry pattern is difficult‡ Require additional downstream processing to identify changes in profile Limited peak resolution Mass accuracy of proteins/peptide Limited prefractionation does not overcome dynamic range issues of complex samples* Require additional downstream processing (eg, tandem mass spectrometry) Require additional downstream processing (eg, peptide labelling)
*Complex protein samples contain more than 2000 discrete proteins/peptides, needing instrumentation with higher resolution and dynamic range capabilities. Simple samples will contain fewer proteins/peptides. †High abundance proteins/ peptides suppress and mask signal resulting from low abundance proteins/peptides. ‡Variation in ionisation efficiency affects total ion current due to differences in sample components (eg, salts).
Table 1: Caveats to consider in SELDI-TOF based mass spectometry
972
within the plasma protein set of the Human Proteome Organisation (HUPO). However, the dynamic range is such that fewer than 20 of these represent more than 98% of total protein mass.7 A major flaw in studies of SELDI-TOF mass spectrometry is the attempt to find biomarkers from among only the 25–50 most abundant serum or plasma proteins, with only limited prefractionation with selective binding to specific target surfaces with low capacity and subsequent competition for binding sites.8 Therefore the same proteins, and their breakdown products, are identified as markers in diseases with different underlying pathological processes. For example, serum amyloid A (a common marker of inflammation and identified by Agranoff as a major marker in tuberculosis patients) has been identified in ovarian cancer and stroke in plasma, and in renal, prostate, and nasopharyngeal cancers in serum. The question of marker specificity is therefore obvious. Combination with other potential markers (as in Agranoff’s study) provides better specificity. At this point, however, we should be cautious against over-interpretation of the diagnostic capabilities of individual proteins identified within a disease signature until more data are obtained from a wider number of patients. A fair argument would be that the absolute diagnostic specificity depends on a marker being raised in one disease alone. Substantial debate also remains about sample collection, preparation, and interpretation of data for SELDI-TOF mass spectrometry.9,10 Even the relative merits of serum over plasma have been argued, because of the raised protease activity seen after clotting in serum.11 HUPO are attempting to provide operating procedures for serum and plasma proteomics that would provide better standardisation and reproducibility needed to move this technology beyond an analytical tool.11 Tuberculosis is a disease in need of a rapid, sensitive, and specific diagnostic test, replacing slow culturing of M tuberculosis (2–6 weeks for a positive assay), poorly sensitive microscopy-based detection in patients’ sputum, and non-specific serological analyses in which substantial cross-reactivity with other mycobacteria (including previous immunisation with BCG) reduces specificity. Agranoff and colleagues’ study highlights two important facts. First, that microbial infections caused by one organism might specifically alter the serum proteome profile differently from infection with another pathogen, and this is the first study, to our knowledge, to investigate serum responsiveness to a bacterial pathogen with serum www.thelancet.com Vol 368 September 16, 2006
Comment
peptide profiling. Second, a disease in which early diagnosis and therapy is an important prognosticator of clinical outcome can be detected rapidly with this approach. Further evidence of the absolute specificity of these tuberculosis markers, and generation of mass-spectral serum patterns from patients in the early stages of disease, could eventually have a huge effect on the clinical outcomes of tuberculosis infection worldwide by reducing the time between infection, diagnosis, and onset of therapy.
3 4
5 6 7
8
Melanie Y White, Rebekah L Gundry, *Stuart J Cordwell Department of Medicine, Johns Hopkins University, Baltimore, MD 21224, USA (MW, RLG); and School of Molecular and Microbial Biosciences, University of Sydney, Sydney, New South Wales, Australia (SJC) [email protected]
9
10
We declare that we have no conflict of interest. 1
2
Agranoff D, Fernandez-Reyes D, Papadopoulos M, et al. Identification of diagnostic markers for tuberculosis by serum proteomic fingerprinting. Lancet 2006; published online Sept 14. DOI:10.1016/S0140-6736(06)69342-2 Hortin GL. The MALDI-TOF mass spectrometric view of the plasma proteome and peptidome. Clin Chem 2006; 52: 1223–37.
11
Liotta LA, Petricoin EF. Serum peptidome for cancer detection: spinning biological trash into diagnostic gold. J Clin Invest 2006; 116: 26–30. Koomen JM, Li D, Xiao LC, et al. Direct tandem mass spectrometry reveals limitations protein profiling experiments for plasma biomarker discovery. J Proteome Res 2005; 4: 972–81. Petricoin EF, Ardekani AM, Hitt VA, et al. Use of proteomic patterns in serum to identify ovarian cancer. Lancet 2002; 359: 572–77. Plebani M. Proteomics: the next revolution in laboratory medicine? Clin Chim Acta 2005; 357: 113–22. Omenn GS, States DJ, Adamski M, et al. Overview of the HUPO Plasma Proteome Project: results from the pilot phase with 35 collaborating laboratories and multiple analytical groups, generating a core dataset of 3020 proteins and a publicly-available database. Proteomics 2005; 5: 3226– 45. Seibert V, Wiesner A, Buschmann T, Meuer J. Surface-enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF MS) and ProteinChip technology in proteomics research. Pathol Res Pract 2004; 200: 83–94. Barker PE, Wagner PD, Stein SE, Bunk DE, Srivastava S, Omenn GS. Standards for plasma and serum proteomics in early cancer detection: a needs assessment report from the national institute of standards and technologynational cancer institute standards, methods, assays, reagents and technologies workshop, August 18-19, 2005. Clin Chem 2006; 52: 1669–74. Haab BB, Geierstanger BH, Michailidis G, et al. Immunoassay and antibody microarray analysis of the HUPO Plasma Proteome Project reference specimens: systematic variation between sample types and calibration of mass spectrometry data. Proteomics 2005; 5: 3278–91. Rai AJ, Gelfand CA, Haywood BC, et al. HUPO Plasma Proteome Project specimen collection and handling: towards the standardization of parameters for plasma proteome samples. Proteomics 2005; 5: 3263–77.
Compliance with osteoporosis therapy is the weakest link The lifetime risk for fragility fractures due to osteoporosis after the age of 50 years is about 50% in women and 20% in men.1 The resultant high morbidity, mortality, and economic costs for elderly people are well recognised, and have stimulated the development of effective interventions to reduce fracture risk. Despite these advances, however, two challenges have hindered reduction of this public-health burden. We need to better identify individuals with a high risk of fracture to target treatment more cost-effectively. We also need to ensure that patients take their treatment as prescribed (compliance) and for the recommended duration (persistence). Compliance and persistence are poor in at least 50% of patients during the first year of treatment for osteoporosis, and in 80% after 3 years.2–4 Most instances of inadequate compliance and persistence occur within 3 months of the start of treatment, and are associated with higher rates of fracture than those observed when compliance and persistence are good. In over 35 000 women on bisphosphonates, fracture rates in those who did not comply or persist with treatment were 20–30% higher than those persisting with medication www.thelancet.com Vol 368 September 16, 2006
as prescribed at 24 months.4 Higher rates of prescription refills were also associated with lower fracture rates.4 The stage at which inadequate compliance and persistence compromise effectiveness is unknown. However, fracture rates increase when compliance falls below 50%. Extra costs from such fractures probably outweigh savings in drug costs and adversely affect the overall cost-effectiveness of fracture prevention, especially in high-risk populations.5 The problem of poor compliance with treatment for chronic diseases has several puzzling aspects. In trials for prevention of coronary heart disease, individuals who complied poorly with placebo had worse health outcomes than those who were compliant, suggesting that inherent characteristics of poor compliers increase morbidity.6,7 Differences in lifestyle might partly explain this finding, but did not account for all of the observed increase in mortality. The causes of poor compliance and persistence are poorly understood—putative causal factors such as age, previous history of fracture, use of multiple medications, and comorbidities explain less than 10% of variability in compliance.8 We need to know the relative contribution of other potential factors, 973
9
10
Nashef SAM, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999; 16: 9–13. Ambler G, Omar RZ, Royston P, et al. Valvular heart disease: generic, simple risk stratification model for heart valve surgery. Circulation 2005; 112: 224–31.
11 12
Baumgartner H. Aortic stenosis: medical and surgical management. Heart 2005; 91: 1483–88. Vahanian A, Palacios IF. Percutaneous approaches to valvular disease. Circulation 2004; 109: 1572–79.
When does a fingerprint constitute a diagnostic? Diagnosis of disease via biomarkers in circulating blood depends on several factors. Markers must be detectable during the preliminary stages of disease, or at least when successful treatment is possible. Tests should be cost effective and readily available at point-of-care, such that the turnaround time for diagnosis is minimised. Preferably, diagnosis should be minimally invasive and, most importantly, must be absolutely specific to the disease to minimise misguided or inadequate therapy. In today’s Lancet, Dan Agranoff and colleagues describe the use of surface-enhanced laser desorption-ionisation time-of-flight mass spectrometry (SELDI-TOF) to identify serum biomarkers from patients with advanced Mycobacterium tuberculosis infections.1
Serum proteomics with this technique provides a pattern-based diagnosis, because multiple peptide and protein signals are detected simultaneously in a single mass spectrum. Because the detector is based on well-established matrix-assisted (MALDI-TOF) massspectrometry methods, there is a bias toward lower mass signals (1–10 kDa), and the approach is often termed peptidomics.2 Furthermore, disease states seem to be characterised by raised levels of cleavage products of abundant serum proteins.3,4 Alternatively, MALDI-TOF mass-spectrometry profiling, which provides spectra of higher resolution, allows for the improved acquisition of higher mass ranges, circumventing the limitations of low mass profiling (figure). In the earliest iterations of
Published Online September 14, 2006 DOI:10.1016/S01406736(06)69379-3 See Articles page 1012
Patient 1
Patient 2
Diseased
Patient 3
Control 1 Control Control 2
Control 3 3000
4000
5000
6000
7000
8000
9000
10 000 m/z
10000
12000
14 000
16000
18000
Increasing molecular weight 1 kDa
10 kDa
20 kDa
Peptidomic or peptide-centric profiling
Proteomic or whole protein-centric profiling
Applications of monitoring low mass range: Naturally occurring peptides Small intact proteins Biological cleavage products Artifacts from sample processing
Applications of monitoring high mass range: Intact or nearly intact proteins Protein isoforms Post-translational modifications Ratios of intact:truncated forms
50 kDa
Figure 1: MALDI-TOF mass spectra of fractionated human serum used for peptide and protein profiling Examples for both low mass and high mass profiling shown. m/z=peptide or protein mass divided by charge state, induced by ionisation.
www.thelancet.com Vol 368 September 16, 2006
971
Comment
SELDI-TOF mass spectrometry for biomarker discovery, identification of the individual components within the signature was not possible,5 although later studies overcame this difficulty with much better algorithmic interpretation of the resulting peptide masses and concurrent use of tandem mass spectrometry for aminoacid sequencing.4 The capital cost of equipment and assays are such that SELDI-TOF mass spectrometry is unlikely to be routinely and, therefore, identification of a suite of markers from within the signature is needed to provide a test amenable to frequent low-cost use. The advantage of SELDI-TOF mass spectrometry is that it does not rely on evidence of a gold-standard biomarker, but rather on combinations of peptide signals. The disease signature given by the serum sample in the mass spectrometer, and compared against a database of controls and test samples, is the potential diagnostic at the preclinical stage. Mass spectrometry is a bioanalytical technique with almost unparallelled sensitivity, and thus provides an ability to mine substantially deeper into the serum or plasma proteomes than has previously been possible. Unfortunately, substantial caveats remain (table).6 The use of serum or plasma is challenging technically. More than 3000 proteins are contained Reasoning Scientifically valid applications Validation studies
Initial screening of biological samples*
Fractionation of samples where proteome is not dominated by subset of proteins (eg, CSF, urine, and tears)† Applications with limited practicality Biomarker discovery in complex biological samples†
Post-translational modification analysis of proteins altered with disease Search for low abundance biomarkers Identification of proteins/peptides Precise protein/peptide quantitation
Antibody labelling of surface to specifically isolate specific proteins and protein complexes Multiple surface chemistries Identify peptide/protein profile changes (presence/ absence) Fast reproducible results Simple sample fractionation Multiple surface chemistries
Interpretation of meaningful data from simple mass spectrometry pattern is difficult‡ Require additional downstream processing to identify changes in profile Limited peak resolution Mass accuracy of proteins/peptide Limited prefractionation does not overcome dynamic range issues of complex samples* Require additional downstream processing (eg, tandem mass spectrometry) Require additional downstream processing (eg, peptide labelling)
*Complex protein samples contain more than 2000 discrete proteins/peptides, needing instrumentation with higher resolution and dynamic range capabilities. Simple samples will contain fewer proteins/peptides. †High abundance proteins/ peptides suppress and mask signal resulting from low abundance proteins/peptides. ‡Variation in ionisation efficiency affects total ion current due to differences in sample components (eg, salts).
Table 1: Caveats to consider in SELDI-TOF based mass spectometry
972
within the plasma protein set of the Human Proteome Organisation (HUPO). However, the dynamic range is such that fewer than 20 of these represent more than 98% of total protein mass.7 A major flaw in studies of SELDI-TOF mass spectrometry is the attempt to find biomarkers from among only the 25–50 most abundant serum or plasma proteins, with only limited prefractionation with selective binding to specific target surfaces with low capacity and subsequent competition for binding sites.8 Therefore the same proteins, and their breakdown products, are identified as markers in diseases with different underlying pathological processes. For example, serum amyloid A (a common marker of inflammation and identified by Agranoff as a major marker in tuberculosis patients) has been identified in ovarian cancer and stroke in plasma, and in renal, prostate, and nasopharyngeal cancers in serum. The question of marker specificity is therefore obvious. Combination with other potential markers (as in Agranoff’s study) provides better specificity. At this point, however, we should be cautious against over-interpretation of the diagnostic capabilities of individual proteins identified within a disease signature until more data are obtained from a wider number of patients. A fair argument would be that the absolute diagnostic specificity depends on a marker being raised in one disease alone. Substantial debate also remains about sample collection, preparation, and interpretation of data for SELDI-TOF mass spectrometry.9,10 Even the relative merits of serum over plasma have been argued, because of the raised protease activity seen after clotting in serum.11 HUPO are attempting to provide operating procedures for serum and plasma proteomics that would provide better standardisation and reproducibility needed to move this technology beyond an analytical tool.11 Tuberculosis is a disease in need of a rapid, sensitive, and specific diagnostic test, replacing slow culturing of M tuberculosis (2–6 weeks for a positive assay), poorly sensitive microscopy-based detection in patients’ sputum, and non-specific serological analyses in which substantial cross-reactivity with other mycobacteria (including previous immunisation with BCG) reduces specificity. Agranoff and colleagues’ study highlights two important facts. First, that microbial infections caused by one organism might specifically alter the serum proteome profile differently from infection with another pathogen, and this is the first study, to our knowledge, to investigate serum responsiveness to a bacterial pathogen with serum www.thelancet.com Vol 368 September 16, 2006
Comment
peptide profiling. Second, a disease in which early diagnosis and therapy is an important prognosticator of clinical outcome can be detected rapidly with this approach. Further evidence of the absolute specificity of these tuberculosis markers, and generation of mass-spectral serum patterns from patients in the early stages of disease, could eventually have a huge effect on the clinical outcomes of tuberculosis infection worldwide by reducing the time between infection, diagnosis, and onset of therapy.
3 4
5 6 7
8
Melanie Y White, Rebekah L Gundry, *Stuart J Cordwell Department of Medicine, Johns Hopkins University, Baltimore, MD 21224, USA (MW, RLG); and School of Molecular and Microbial Biosciences, University of Sydney, Sydney, New South Wales, Australia (SJC) [email protected]
9
10
We declare that we have no conflict of interest. 1
2
Agranoff D, Fernandez-Reyes D, Papadopoulos M, et al. Identification of diagnostic markers for tuberculosis by serum proteomic fingerprinting. Lancet 2006; published online Sept 14. DOI:10.1016/S0140-6736(06)69342-2 Hortin GL. The MALDI-TOF mass spectrometric view of the plasma proteome and peptidome. Clin Chem 2006; 52: 1223–37.
11
Liotta LA, Petricoin EF. Serum peptidome for cancer detection: spinning biological trash into diagnostic gold. J Clin Invest 2006; 116: 26–30. Koomen JM, Li D, Xiao LC, et al. Direct tandem mass spectrometry reveals limitations protein profiling experiments for plasma biomarker discovery. J Proteome Res 2005; 4: 972–81. Petricoin EF, Ardekani AM, Hitt VA, et al. Use of proteomic patterns in serum to identify ovarian cancer. Lancet 2002; 359: 572–77. Plebani M. Proteomics: the next revolution in laboratory medicine? Clin Chim Acta 2005; 357: 113–22. Omenn GS, States DJ, Adamski M, et al. Overview of the HUPO Plasma Proteome Project: results from the pilot phase with 35 collaborating laboratories and multiple analytical groups, generating a core dataset of 3020 proteins and a publicly-available database. Proteomics 2005; 5: 3226– 45. Seibert V, Wiesner A, Buschmann T, Meuer J. Surface-enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF MS) and ProteinChip technology in proteomics research. Pathol Res Pract 2004; 200: 83–94. Barker PE, Wagner PD, Stein SE, Bunk DE, Srivastava S, Omenn GS. Standards for plasma and serum proteomics in early cancer detection: a needs assessment report from the national institute of standards and technologynational cancer institute standards, methods, assays, reagents and technologies workshop, August 18-19, 2005. Clin Chem 2006; 52: 1669–74. Haab BB, Geierstanger BH, Michailidis G, et al. Immunoassay and antibody microarray analysis of the HUPO Plasma Proteome Project reference specimens: systematic variation between sample types and calibration of mass spectrometry data. Proteomics 2005; 5: 3278–91. Rai AJ, Gelfand CA, Haywood BC, et al. HUPO Plasma Proteome Project specimen collection and handling: towards the standardization of parameters for plasma proteome samples. Proteomics 2005; 5: 3263–77.
Compliance with osteoporosis therapy is the weakest link The lifetime risk for fragility fractures due to osteoporosis after the age of 50 years is about 50% in women and 20% in men.1 The resultant high morbidity, mortality, and economic costs for elderly people are well recognised, and have stimulated the development of effective interventions to reduce fracture risk. Despite these advances, however, two challenges have hindered reduction of this public-health burden. We need to better identify individuals with a high risk of fracture to target treatment more cost-effectively. We also need to ensure that patients take their treatment as prescribed (compliance) and for the recommended duration (persistence). Compliance and persistence are poor in at least 50% of patients during the first year of treatment for osteoporosis, and in 80% after 3 years.2–4 Most instances of inadequate compliance and persistence occur within 3 months of the start of treatment, and are associated with higher rates of fracture than those observed when compliance and persistence are good. In over 35 000 women on bisphosphonates, fracture rates in those who did not comply or persist with treatment were 20–30% higher than those persisting with medication www.thelancet.com Vol 368 September 16, 2006
as prescribed at 24 months.4 Higher rates of prescription refills were also associated with lower fracture rates.4 The stage at which inadequate compliance and persistence compromise effectiveness is unknown. However, fracture rates increase when compliance falls below 50%. Extra costs from such fractures probably outweigh savings in drug costs and adversely affect the overall cost-effectiveness of fracture prevention, especially in high-risk populations.5 The problem of poor compliance with treatment for chronic diseases has several puzzling aspects. In trials for prevention of coronary heart disease, individuals who complied poorly with placebo had worse health outcomes than those who were compliant, suggesting that inherent characteristics of poor compliers increase morbidity.6,7 Differences in lifestyle might partly explain this finding, but did not account for all of the observed increase in mortality. The causes of poor compliance and persistence are poorly understood—putative causal factors such as age, previous history of fracture, use of multiple medications, and comorbidities explain less than 10% of variability in compliance.8 We need to know the relative contribution of other potential factors, 973