Printer Friendly

A systematic method for selection of promising serum protein biomarkers to improve prostate cancer ([PCa.sup.1]) detection.

Immunoassays for prostate-specific antigen (PSA) are used routinely as an aid to detect prostate cancer (PCa). PSA permits early detection of PCa but lacks specificity (1, 2). In a typical screening population of African American men [greater than or equal to]40 years of age and other men [greater than or equal to]50 years of age, ~4% of the men with a PSA concentration >4 [micro]g/L, which is the upper reference limit for PSA, will have prostate cancer (1). Most of these cancers can be detected using PSA. At issue, however, is the comorbidity of men with benign prostatic hyperplasia whose age and PSA concentrations are within these ranges. For example, 50% of men >60 years of age have benign prostatic hyperplasia (3), and many will have PSA >4 [micro]g/L. The low specificity of PSA measurement as a test for PCa necessitates many prostate biopsies, of which ~75% are negative and theoretically unnecessary. Additional biomarkers may improve sensitivity and specificity of PSA. We theorized that bioinformatics and analysis of the existing scientific literature could identify promising protein biomarker candidates that, if combined with PSA, might improve clinical sensitivity and especially specificity.

The work flow we used to select candidate biomarkers is diagramed in the flow chart in Fig. 1. We searched Medline and other relevant databases from Nerac Corporation (Tolland, CT) for publications that appeared in 1999-2004 with the terms prostate cancer and markers or biomarkers or early detection or aggressiveness or staging or grading. We also used prostate-specific genes, prostate-specific proteins, and several well-known institutions and investigators in the field of prostate cancer research as search terms. Once compiled, the titles and abstracts were read to find proteins or genes documented to be differentially regulated in tissue or serum or plasma PCa samples compared with normal or benign samples. A list of >2700 citations was narrowed to 600 candidate biomarkers supported by >800 references. This list was cataloged according to the approved gene symbol (AGS) of the Human Genome Organization (HUGO), and candidate sequences were analyzed by a novel algorithm to determine probability of secretion (4,5). The algorithm processed RefSeq protein sequences through an analysis pipeline that (a) predicted N-terminal signal sequence using TargetP 1.1 and SignalP 3.0, (b) identified transmembrane domains: TMHMM 2.0, and (c) resolved N-terminal signal anchors from signal peptides with Phobius. Proteins were also documented for secretion by mining Swiss-Prot database annotation for cellular localization entries (5). The resulting 130 secreted proteins were further assessed by in-depth analysis of the scientific literature.

The large amount of existing scientific literature was specifically analyzed to select PCa biomarker candidates. These data were decreased to a reasonable number of promising candidates by use of multiple, predetermined, clinically relevant criteria. Specific criteria were used to analyze the literature for promising markers: (i) assay type [e.g., ELISA or quantitative reverse transcription polymerase chain reaction (gRTPCR)], (ii) cancer differential (the assay signal) up- or down-regulated in cancerous tissue samples compared with healthy or benign samples, (iii) sample type (e.g., serum, bulk tissue, micro-dissected tissue), (iv) sample numbers, (v) potential use (e.g., whether data indicated that the marker had potential as a marker for early detection or cancer aggressiveness), and (vi) reference type (e.g., meeting abstract, primary publication, review article). Candidate markers were assigned to 4 groups: pursue (P), promising/ discuss (PD), need more information (N), or hold/discard (H) (Fig. 1). This process was necessarily subjective, but several specific standards were consistently used. More weight was given to markers with ELISA data on serum or plasma samples and to markers that were up-regulated rather than down-regulated in cancer tissue compared with healthy samples. Microdissected tissue data were preferred over bulk tissue, and larger or statistically significant sample numbers were preferred over smaller studies or those with nonsignificant findings. Candidate markers associated with both early detection and aggressiveness were favored. The most complete data came from primary papers.


More information was still required to categorize the potential usefulness of a large number of the markers that initially fit into the N category. Some of these markers were further analyzed by determining their relative gene expression across multiple tissues, both normal and cancerous, with a proprietary approach of mining expressed sequence-tagged data (5). Finally, we compared our list of potential genes to microarray expression data obtained from the National Cancer Institute's cancer Biomedical Informatics Grid (caBIG) database.

After we compiled a list of qualified candidates, we performed a search of commercially available reagents and found antigens, antibodies, ELISA reagent sets, and clinical assays for most of the candidates. Antigens as well as antibodies are essential for formulating quantitative ELISAs and are usually needed for reliable interpretation of Western blots.

We used available ELISAs to test for 9 biomarker candidates in serum samples with defined PSA concentrations. Deidentified serum samples were obtained from men with PSA concentrations of 0-4 (n = 22), 4-10 (n = 20), and >10 [micro]/L (n = 16). To assay the samples we used commercially available reagent sets or immunoanalyzers [e.g., hepatocyte growth factor (HGF) cat. no. DH600 (R & D Systems) or Immulite 2000 for acid phosphatase, prostate (ACPP) (PAP cat. no. L2KDAZ) (Diagnostic Products Corp.) and Access immunoanalyzer for PSA (cat. no. 37200) (Beckman Coulter)] according to manufacturers' instructions.

For 7 candidate biomarkers with available antibodies but no available ELISA, we used Western blot analysis to compare paired cancer and normal tissue samples from 10 patients by the method of Towbin et al. (6).

For the 5 candidate biomarkers with no commercially available reagents, we analyzed paired normal and cancerous tissues from 10 PCa patients. We synthesized nucleic acid probes and used gRTPCR performed with SYBR green PCR Master Mix (Applied Biosystems) on the ABI PRISM 7900 HT Real-Time PCR System (Applied Biosystems). Total RNA was reverse transcribed with Superscript III (Invitrogen) and Oligo deoxythymidine primers. PCR was performed on 10 ng of cDNA with 0.2 [micro]mol/L of each primer in 20 [micro]L, according to the manufacturer's suggested cycling conditions. Primers were designed with the computer programs Oligo 6 (National Biosciences) and Primer Express (Applied Biosystems) and were synthesized by Invitrogen. Relative gene expression was determined by use of the comparative threshold cycle ([DELTA]Ct) method, with [beta]-actin as the reference gene. The expression ratio between each tumor and matched normal sample was calculated from the normalized Ct values. We confirmed the amplification efficiencies of the primers by generating a standard curve, and we analyzed the PCR products with a dissociation curve to verify single-product amplification. Human serum and tissue samples were collected under procedures approved by Mayo Clinic's institutional review board. All statistical tests were performed with SAS-JMP6.0 software.

ACPP and HGF were significantly increased in the >10 [micro]/L PSA group compared with the 0-4 [micro]/L group (Dunnet test P <0.05). There was no significant difference between the 0-4 [micro]/L and the 4-10 [micro]/L group for HGF or ACPP, nor were there any substantial differences in the other 7 candidates between the 0-4 [micro]/L and the 4-10 or >10 [micro]/L PSA groups.

The Western blot analyses of the biomarker candidate anterior gradient 2 homolog (AGR2) showed a 3.8-fold median increase in signal at the expected molecular mass in cancer tissues vs the normal tissues of the 10 patients (Table 1). Expression of the other 6 candidate genes did not show a consistent difference in cancer and healthy tissue.

Two of the 5 candidate genes subjected to gRTPCR showed very low expression in prostate tissue. Three candidates showed expression but no significant (Wilcoxon P >0.05) difference in expression between cancer and normal tissue. A lack of differential in gene expression between cancerous and normal tissue may not exclude the possibility of finding different protein concentrations in serum or other biological fluids of persons with PCa compared with those without PCa.

We found that existing scientific literature can be specifically analyzed to select promising PCa biomarker candidates. These data can be reduced to a reasonable number of promising candidates by use of multiple, predetermined, clinically relevant criteria. Use of HUGO-approved gene symbols is crucial for reducing confusion caused by multiple aliases that exist for most genes and the proteins they express. Bioinformatics and publicly available databases can be used to select specific criteria that theoretically are crucial to the success of biomarkers, such as probability of secretion or tissue- and cancer-specific gene expression profiles. Once markers were selected we found a multitude of commercially available reagents such as ELISAs and antibodies that can be used to rapidly assess the validity of candidate biomarkers on relevant patient samples. Therefore, we have formulated an iterative process (Fig. 1) through which we can use all available scientific literature and bioinformatic information to select and then evaluate, in the laboratory, potential serum markers for prostate cancer. As new literature for prostate cancer marker candidates becomes available, we can continually assess the new data and evaluate new candidates through our iterative process. We hope that some of these promising candidates can be combined with PSA to produce a multianalyte test that decreases unnecessary prostate biopsies. Furthermore, this approach will be useful in selecting and screening markers for other cancers and disease states.


(1.) Catalona WJ, Partin AW, Slawin KM, Brawer MK, Flanigan RC, Patel A, et al. Use of the percentage of free prostate-specific antigen to enhance differentiation of prostate cancer from benign prostatic disease: a prospective multicenter clinical trial. JAMA 1998;279:1542-7.

(2.) Thompson IM, Ankerst DP, Chi C, Lucia MS, Goodman PJ, Crowley JJ, et al. Operating characteristics of prostate-specific antigen in men with an initial PSA level of 3.0 ng/ml or lower. JAMA 2005;294:66-70.

(3.) Skolarikos A, Thorpe AC, Neal DE. Lower urinary tract symptoms and benign prostatic hyperplasia. Minerva Urol Nefrol 2004;56:109-22.

(4.) Klee EW, Ellis LB. Evaluating eukaryotic secreted protein prediction. BMC Bioinformatics 2005;6:256.

(5.) Klee EW, Finlay JA, McDonald C, Attewell JR, Hebrink D, Dyer R, et al. Bioinformatics Methods for Prioritizing Serum Biomarker Candidates. Clin Chem 2006; 52:2162-4.

(6.) Towbin H, Gordon J. Immunoblotting and dot immunobinding: current status and outlook. J Immunol Methods 1984;72:313-40.

DOI : 10.1373/clinchem.2006.072959

Judith A. Finlay, [1] * Eric W. Klee, [2] Cari McDonald, [2] John R. Attewell, [2] Deanne Hebrink, [2] Roy Dyer, [2] Brad Love, [1] George Vasmatzis, [2] Thomas M. Li, [1] ([dagger]) Joseph M. Beechem, [1] and George G. Klee [2] ([1] Invitrogen Corporation, Carlsbad, CA; and [2] Mayo Clinic, Rochester, MN; ([dagger]) current affiliation: Roche Diagnostics, Pleasanton, CA; * address correspondence to this author at: Invitrogen Corporation, 5781 Van Allen Way, Carlsbad, CA 92008; fax 760-476-6814; e-mail
Table 1. Western analysis demonstrates that AGR2 is
significantly up-regulated in prostate cancer tissue.

Tissue Patient Tumor Density Fold cancer
type ID no. area, % AGR2 band Median over expression

Cancer 1 85 133 609 96 176 3.8 (a)
 2 70-75 96 176
 3 75 130 544
 4 90 119 659
 5 65 50 771
 6 90 52 175
 7 60 27 841
 8 85 134 733
 9 70
 10 45-50 50 505

Normal 1 5 027 25 272
 2 7 322
 3 35 903
 4 17 084
 5 28 514
 6 29 183
 7 22 030
 9 45 936

(a) Median between normal and cancer is significantly
different by Wilcoxon test; P = 0.0021.
COPYRIGHT 2006 American Association for Clinical Chemistry, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2006 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Abstracts of Oak Ridge Posters
Author:Finlay, Judith A.; Klee, Eric W.; McDonald, Cari; Attewell, John R.; Hebrink, Deanne; Dyer, Roy; Lov
Publication:Clinical Chemistry
Date:Nov 1, 2006
Previous Article:Development and preliminary clinical validation of a high sensitivity assay for cardiac troponin using a capillary flow (single molecule)...
Next Article:Bioinformatics methods for prioritizing serum biomarker candidates.

Related Articles
Putting the "bio" back into biomarkers: orienting Proteomic discovery toward biology and away from the measurement platform.
Analytical validation of serum proteomic profiling for diagnosis of prostate cancer: sources of sample bias.
LC-MS/MS quantification of Zn-[alpha]2 glycoprotein: a potential serum biomarker for prostate cancer.
Urine markers as possible tools for prostate cancer screening: review of performance characteristics and practicality.
Bioinformatics methods for prioritizing serum biomarker candidates.
Evaluation of serum protein profiling by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry for the detection of prostate...
Mass spectroscopy as a discovery tool for identifying serum markers for prostate cancer.
Boosted decision tree analysis of surface-enhanced laser desorption/ionization mass spectral serum profiles discriminates prostate cancer from...

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |