Advertisement
Journal of Clinical Oncology  
Search for:
Limit by:
  Browse by Subject or Issue
Home Search or Browse JCO My JCO Subscriptions Customer Service Site Map

Journal of Clinical Oncology, Vol 23, No 22 (August 1), 2005: pp. 4835-4837
© 2005 American Society of Clinical Oncology.
DOI: 10.1200/JCO.2005.02.912

This Article
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a colleague
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Save to my personal folders
Right arrow Download to citation manager
Right arrowRights & Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Robbins, R. J.
Right arrow Articles by Tempst, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Robbins, R. J.
Right arrow Articles by Tempst, P.
Related Articles
Right arrowRelated Article
Social Bookmarking
 Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Facebook   Add to Reddit   Add to Technorati   Add to Twitter  
What's this?

EDITORIAL

Distilling Cancer Biomarkers From the Serum Peptidome: High Technology Reading of Tea Leaves or an Insight to Clinical Systems Biology?

Richard J. Robbins, Josep Villanueva, Paul Tempst

Endocrine Service, Protein Center, and Molecular Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY

In this issue, Mian et al1 use high-resolution mass spectrometry (MS) to define the serum peptidome of patients with melanoma. They trained machine learning algorithms to recognize serum peptidomic patterns that clustered with specific disease states. This strategy enabled them to discriminate stage I melanoma patients from stage IV patients. Furthermore, the peptidomic patterns were able to discriminate stage III melanoma patients who developed tumor recurrences within the first year after diagnosis from stage III patients who remained disease free. An artificial neural network analysis of the peptide patterns was able to correctly identify 80% of patients who eventually experienced recurrence, compared with a 21% prediction rate when these investigators used the conventional tumor marker S100-ß. Where do these data fit in the evolving field of serum proteomics and the clinical need for new tumor markers?

For the last 50 years, oncologists have sought to identify serum markers that would disclose the presence of specific cancers or shed light on the natural history or prognosis of a cancer. The most reliable and validated markers include carcinoembryonic antigen for a variety of cancers, calcitonin for medullary thyroid carcinoma, prostate-specific antigen for prostate carcinoma, thyroglobulin for papillary or follicular thyroid carcinoma, human chorionic gonadotropin or alpha-fetoprotein for germ cell tumors, CA-125 for ovarian carcinoma, CA 15-3 for breast carcinoma, and SCC for squamous cell carcinoma of the cervix.

Physicians have used changes in serum biomarkers to judge response to therapy or to define the rate of disease progression. Many markers that are of limited value in screening for cancer are valuable for monitoring for recurrences (such as prostate-specific antigen and thyroglobulin). To date, virtually all useful markers have been peptides or proteins. Each marker is normally present in subnanomolar concentrations, and immunoassay is normally used to measure the exact level of the marker in serum.

The earliest attempts to analyze serum proteins were able to identify albumin and gamma-globulin peaks because of the fact that they were present in milligram per milliliter quantities. Even at that time, clinicians tried to deduce disease states, such as liver disease or malnutrition, based on albumin to globulin ratios. The advent of gel electrophoresis revealed dozens of new lower abundance proteins, such as transferring apolipoproteins, clotting factors, and members of the complement system. These proteins, along with albumin and globulin, make up approximately 98% of the mass of all serum proteins. The ability to quickly identify and quantify specific proteins using two-dimensional gels resulted in advances in detecting and defining a variety of diseases, but no new classical cancer markers arose from this methodologic advance.

Virtually all commonly used cancer serum markers and cytokines are present in plasma at concentrations below 1 ng/mL, which is below the detection limit of two-dimensional gels and, therefore, requires individual immunoassay for quantitation. Recent advances in MS now enable protein chemists to rapidly and reliably display hundreds of these low-abundance proteins in single spectra generated from microliter volumes of serum. The use of a matrix-assisted ionization method, which places a positive charge on each polypeptide, results in a spectrum in which the position of the peptide or protein is essentially determined by its molecular mass. It is now clear that human serum contains hundreds of small- to medium-sized peptides, which constitutes a new clinical entity known as the serum peptidome. Does this new source of information hold any clues about the health status of the individual? Is the pattern unique to each individual, like a molecular fingerprint? Does this pattern change constantly with physiologic events, or is it stable over time? Can the serum peptidome serve as a new biomarker of human systems biology, revealing changes that correspond to specific disease states such as cancer?

In 2002, a seminal article by Petricoin et al2 appeared in the Mechanisms of Disease section of The Lancet. In this report, the authors presented data supporting the notion that an iterative searching algorithm could identify unique serum proteomic patterns that correlated with the presence or absence of ovarian carcinoma. Serum peptides were captured on proprietary protein chips and then subjected to surface-enhanced laser desorption ionization time-of-flight MS. In this way, the authors were able to discriminate early-stage ovarian carcinoma from benign ovarian disease based on an algorithm that relied on five key serum peptides. Using a training set to develop the algorithm, they found 100% sensitivity and 95% specificity at detecting a masked set of 50 sera from women with ovarian carcinoma and 50 without ovarian carcinoma. Over the next 2 years, more than a dozen additional reports3-6 appeared, confirming the notion that sophisticated machine learning algorithms could correctly group the highly complex serum proteomic patterns into those with or without breast, prostate, or ovarian cancers. Sensitivities and specificities greater than 85% were routinely achieved. These levels of diagnostic accuracy were much higher than the levels ever achievable with conventional serum biomarkers. One report suggested that tumor response could be predicted by serum proteomic profiling.7 It seemed that a revolution in cancer biomarkers was underway, with the serum proteomic pattern (or molecular fingerprint) surpassing the efficacy of single biomarkers.

In January 2004, Baggerly et al8 used the original Petricoin datasets, which were posted on the internet, to replicate their findings. Using three different datasets that were generated with two different Ciphergen protein chips (Ciphergen, Fremont, CA), they found it difficult to reproduce the original results. Inconsistencies related to baseline correction of spectra and suspected improper mass calibration were noted, and they were not able to overlay sets of features from one run onto another of presumably the same sera. The same authors were actually able to use signals from the posted datasets occurring in regions that did not correspond to polypeptides and get compete discrimination between cancer and noncancer samples. They concluded that the high diagnostic accuracy reported in the article by Petricoin et al2 was likely a result of artifacts of sample processing or of calibration shifts in the spectra that produced spuriously discriminatory patterns. In conclusion, Baggerly et al8 pointed out that, for this approach to achieve credibility as a cancer biomarker equivalent, rigorous attention must be paid to preanalytic sample handling, to MS technique, to baseline correction and peak extraction, and finally, to properly validated bioinformatic analyses.9

The lack of uniformity in published techniques has resulted in a lack of uniformity of outcomes. In five different prostate cancer reports, only two of the many discriminatory masses appear in more than a single publication. It is also clear that, with the current technology, many of the peaks do not correspond to true proteins or peptides and may represent machine noise.

Mian et al1 add new data to this growing field. This is the first appearance of an article based on this novel technology in this journal, and the reader must be aware of the strengths and limitations of this approach. Application of state-of-the-art MS to serum peptide profiling is in a state of rapid development. The methods used to collect blood and harvest peptides from serum, the matrices used to charge peptides, the MS hardware and robotic systems necessary to assure reproducibility, and the data extraction and computational systems necessary to recognize the complex patterns are all rapidly evolving. The report from Mian et al1 acknowledges that each of these factors can influence the production of the final serum peptidome, but they have not demonstrated that their signals are stable over time or reproducible by running the same sample on different days. For example, their data rely on a proprietary protein chip that binds a unique subset of serum peptides. If Ciphergen changes the makeup of this chip or stops production, independent confirmation of their findings in the future will be difficult. To their credit, the authors acknowledge that there is considerable uncertainty in the field because, to date, no two labs have confirmed each other's findings.

It is becoming clear that the highly complex serum proteome must be the product of endogenous substrates in conjunction with their proteolytic breakdown products, which are a readout of the repertoire of proteases that exist in plasma or become activated during clotting. Cancer cells may contribute unique peptides or, more importantly, proteases that may result in subtle but signature alterations in the complex equation of hundreds of peptides that can be resolved from human serum. In addition, the host immune response to the cancer may also contribute new proteins, such as cytokines and lymphokines, and new proteases to plasma that may add to or alter the molecular fingerprint of peptides that is presumed to exist in a healthy individual. It is important to point out that the robustness and the stability of an individual's serum peptidome has never been clearly established. We do not know if the complex peptide patterns are constantly changing as a result of minor physiologic events in the person or if they are extremely stable even in the face of major biologic events.

What is clear is that a spectrum of peaks can be easily generated using MS analysis of serum. The fact that considerable artifact and variability are easily induced by minimal alterations in the techniques of blood collection, clotting, protein extraction, matrix application, laser settings, peak extraction, and bioinformatic analysis will require a concerted effort by MS teams who are exploring the possibility that the serum proteome is an accurate reflection of the human systems biology to standardize and validate the data generated. The proof of the potential value of this exciting new approach will be in the ability of disparate groups to reproduce either the entire proteomic patterns generated in other labs or to show that the highly discriminatory peptides have the same amino acid sequences and, thereby, truly qualify as cancer biomarkers. We look forward to new reports using this technology and validation studies to assure its authenticity. At present, this promising new approach is not yet ready for patient classification or treatment choices.

Authors' Disclosures of Potential Conflicts of Interest

The authors indicated no potential conflicts of interest.

REFERENCES

1. Mian S, Ugurel S, Parkinson E, et al: Serum proteomic fingerprinting discriminates between clinical stages and predicts disease progression in melanoma patients. J Clin Oncol 23: 5088-5093, 2005

2. Petricoin EF, Ardekani AM, Hitt BA, et al: Use of proteomic patterns in serum to identify ovarian cancer. Lancet 359: 572-577, 2002[CrossRef][Medline]

3. Adam BL, Qu Y, Davis JW, et al: Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men. Cancer Res 62: 3609-3614, 2002[Abstract/Free Full Text]

4. Li J, Zhang Z, Rosenzweig J, et al: Proteomics and bioinformatics approaches for identification of serum biomarkers to detect breast cancer. Clin Chem 48: 1296-1304, 2002[Abstract/Free Full Text]

5. Wadsworth JT, Somers KD, Stack BC Jr, et al: Identification of patients with head and neck cancer using serum protein profiles. Arch Otolaryngol Head Neck Surg 130: 98-104, 2004[Abstract/Free Full Text]

6. Xiao X, Liu D, Tang Y, et al: Development of proteomic patterns for detecting lung cancer. Dis Markers 19: 33-39, 2003[Medline]

7. Xiao Z, Luke BT, Izmirlian G, et al: Serum proteomic profiles suggest celecoxib-modulated targets and response predictors. Cancer Res 64: 2904-2909, 2004[Abstract/Free Full Text]

8. Baggerly KA, Morris JS, Coombes KR: Reproducibility of SELDI-TOF protein patterns in serum: Comparing datasets from different experiments. Bioinformatics 20: 777-785, 2004[Abstract/Free Full Text]

9. Villanueva J, Philip J, Entenberg D, et al: Serum peptide profiling by magnetic particle-assisted, automated sample processing and MALDI-TOF mass spectrometry. Anal Chem 76: 1560-1570, 2004[Medline]


Add to CiteULike CiteULike   Add to Complore Complore   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us   Add to Digg Digg   Add to Facebook Facebook   Add to Reddit Reddit   Add to Technorati Technorati   Add to Twitter Twitter    What's this?

Related Article

  • Serum Proteomic Fingerprinting Discriminates Between Clinical Stages and Predicts Disease Progression in Melanoma Patients
    Shahid Mian, Selma Ugurel, Erika Parkinson, Iris Schlenzka, Ian Dryden, Lee Lancashire, Graham Ball, Colin Creaser, Robert Rees, and Dirk Schadendorf
    JCO 2005 23: 5088-5093 [Abstract] [Full Text]


This article has been cited by other articles:


Home page
Clin. Chem.Home page
J. Li, Z. Zhang, and D. W. Chan
Validation of Breast Cancer Biomarkers Identified by Mass Spectrometry: Some of the authors of the article cited above respond:
Clin. Chem., April 1, 2006; 52(4): 772 - 772.
[Full Text] [PDF]


Home page
Exp. Biol. Med.Home page
M. W. Duncan and S. W. Hunsucker
Proteomics as a Tool for Clinically Relevant Biomarker Discovery and Validation
Experimental Biology and Medicine, December 1, 2005; 230(11): 808 - 817.
[Abstract] [Full Text] [PDF]


This Article
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a colleague
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Save to my personal folders
Right arrow Download to citation manager
Right arrowRights & Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Robbins, R. J.
Right arrow Articles by Tempst, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Robbins, R. J.
Right arrow Articles by Tempst, P.
Related Articles
Right arrowRelated Article
Social Bookmarking
 Add to CiteULike   Add to Complore   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Facebook   Add to Reddit   Add to Technorati   Add to Twitter  
What's this?

About
JCO
 Editorial
Roster
 Advertising
Information
 Librarians &
Institutions
 Rights &
Permissions
 PDA Services

Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
Terms and Conditions of Use
  HighWire Press HighWire Press™ assists in the publication of JCO Online