Characterization of Proteins in Human Pancreatic Cancer Serum

Sep 7, 2005 - ... meet the requirements of a good screening test for pancreatic cancer. ...... Yanlei Ma , Jiayuan Peng , Long Huang , Weijie Liu , Pe...
22 downloads 0 Views 379KB Size
Download PDF
Characterization of Proteins in Human Pancreatic Cancer Serum Using Differential Gel Electrophoresis and Tandem Mass Spectrometry Kenneth H. Yu, Anil K. Rustgi, and Ian A. Blair* Division of Hematology/Oncology, Center for Cancer Pharmacology, and Genomics Institute, University of Pennsylvania, Philadelphia, Pennsylvania 19104-6160 Received June 11, 2005

The purpose of this study was to develop techniques for identifying cancer biomarkers in human serum using differential in-gel electrophoresis (DIGE), and characterizing the protein biomarkers using tandem mass spectrometry (MS/MS). A major problem in profiling protein expression by DIGE comes from the presence of high concentrations of a small number of proteins. Therefore, serum samples were first chromatographed using an immunoaffinity HPLC column (Agilent Technologies), to selectively remove albumin, immunoglobulins, transferrin, haptoglobin, and antitrypsin. Serum samples from three individuals with pancreatic cancer and three individuals without cancer were compared. Serum samples were processed using the immunoaffinity column. Differential protein analysis was performed using DIGE. A total of 56 protein spot-features were found to be significantly increased and 43 significantly decreased in cancer serum samples. These spot features were excised, trypsin digested, and analyzed by MALDI/TOF/TOF (4700 Proteomics Analyzer, Applied Biosystems). We identified 24 unique proteins that were increased and 17 unique proteins that were decreased in cancer serum samples. Western blot analysis confirmed increased levels of several of these proteins in the pancreatic cancer serum samples. In an independent series of serum samples from 20 patients with pancreatic cancer and 14 controls, increased levels of apolipoprotein E, R-1-antichymotrypsin, and inter-R-trypsin inhibitor were found to be associated with pancreatic cancer. These results suggest that affinity column enrichment and 2-D DIGE can be used to identify numerous proteins differentially expressed in serum from individuals with pancreatic cancer. Keywords: serum • biomarker • pancreatic cancer • mass spectrometry • albumin-depletion • proteomics • DIGE • immunoaffinity depletion

Introduction Pancreatic cancer is the fourth most common cause of cancer-related mortality in the United States.1 The five-year survival rate is the lowest among all cancers, with estimates ranging from 0.4 to 4%. In 2003, an estimated 30 700 new cases of pancreatic cancer were diagnosed, and an estimated 30 000 patients died as a result of their disease.2 Because of the aggressiveness of this cancer, the inability to diagnose it early, and the current lack of outcome altering therapies, mortality rates from pancreatic cancer are almost identical to incidence rates. The only potentially curative treatment for pancreatic cancer is surgical resection. However, because the disease is generally advanced at presentation, only 10% to 20% of patients are eligible for curative resection. In these patients who undergo pancreaticoduodenectomy, five-year survival is somewhat better, about 20%.3 However, when pancreatic cancers limited to the pancreas are identified at a very early stage (e.g., less * To whom correspondence should be addressed. Center for Cancer Pharmacology, University of Pennsylvania School of Medicine, 854 BRB II/ III, 421 Curie Boulevard Philadelphia PA 19104-6160. Tel: (215) 573-9880. Fax: (215) 573-9889. E-mail: [email protected].

1742

Journal of Proteome Research 2005, 4, 1742-1751

Published on Web 09/07/2005

than 2 cm in diameter) the 5 year survival is higher, as high as 46% in some series.4,5 Thus, identification of very early stage disease, including pancreatic cancer in situ may be the most promising method to reduce pancreatic cancer mortality. Carbohydrate antigen (CA) 19-9 is a commonly used protein tumor marker for pancreatic cancer, however its use is largely limited to following the course of disease.6,7 CA 19-9 is not specific for pancreatic cancer, being expressed in benign conditions such as cholangitis and chronic pancreatitis.8,9 Furthermore, CA 19-9 is not expressed at all by some pancreatic tumors, and in other tumors, it often is not detectable until pancreatic cancer is at a late and incurable stage.10 As such, CA 19-9 does not meet the requirements of a good screening test for pancreatic cancer. Unfortunately, the same is true of a number of other potential targets including CEA, PNA-binding glycoproteins,11 hTert (telomerase catalytic subunit),12 matrix metalloproteinase-2 (MMP-2)13 and, more recently, aberrantly methylated promoter CpG islands of key genes (for example, see Fukushima, et al.14). Although these techniques continue to evolve in sophistication, to date no marker or method has proved adequately robust, or able to detect sufficiently early disease, to significantly impact overall patient survival. 10.1021/pr050174l CCC: $30.25

 2005 American Chemical Society

research articles

Proteins in Human Pancreatic Cancer Serum Table 1. Patient Characteristics of Pancreatic Cancer and Control Serum Samples Used for Pilot DIGE Study age

gender

disease

TNM stagea

AJCC stagea

42 66 57 29 47 37

F M F F M

pancreatic cancer pancreatic cancer pancreatic cancer chronic pancreatitis chronic pancreatitis pseudocyst

T3N1M1 T2N0M0 T2N1M1

IV IB IV

a

Staging criteria defined by the American Joint Committee on Cancer.39

Differential in-gel electrophoresis (DIGE) is an effective technology for separating complex protein samples and for quantifying protein levels between samples. Proteins of interest can then be identified by tandem mass spectrometry (MS/MS). A major problem in profiling serum protein expression by DIGE is the presence of high concentrations of a small number of proteins. The 5 to 10 most abundant serum proteins comprise over eighty-five percent of total serum protein.15 Proteins of interest are likely present at significantly lower concentrations. The most abundant serum proteins make it difficult to run twodimensional gels reproducibly. These abundant proteins also limit the amount of serum that can be loaded onto 2-D gels for analysis, and mask differential expression of lower abundance proteins with similar molecular weights and isoelectric points. In this study, a reliable method for the immunoaffinity removal of the highest abundance proteins from the serum of patients with pancreatic cancer and controls is presented. Samples were then analyzed by DIGE and MS/MS. These techniques have allowed for the identification and characterization of a number of potential protein biomarkers. Differential expression of several of these proteins has been confirmed using traditional immunological approaches.

Materials and Methods Serum Samples Studied. Pooled, normal serum (Sigma, St. Louis, MO) was divided into 50 µL aliquots, and stored at -80 °C until ready for use. These samples were utilized to test the reproducibility of processing samples using the Multiple Affinity Removal Column (Agilent Technologies, Palo Alto, CA). Sera from 3 patients with pancreatic cancer and 3 patients with benign pancreatic disease (chronic pancreatitis and pancreatic pseudocyst) were obtained (Table 1). A pooled sample, consisting of equal amounts of each of the six experimental samples, was made and used as a pooled internal standard. In addition, sera from 20 patients with pancreatic cancer and 14 patients with benign pancreatic or biliary disease were obtained for validation studies (Table 4). Serum samples used for this study were collected from patients with pancreatic cancer thought to be resectable, or other diseases of the pancreas or biliary system that warranted surgery. 10 mL blood samples were collected preoperatively in glass tubes without additive (10 mL BD VacutainerTM No Additive, BD, Franklin Lakes, NJ) and allowed to clot at room temperature for 40 min. Serum was separated by centrifugation at 1500 rpm for 15 min. 1 mL aliquots of serum were taken and stored at -80 °C until ready for use. The time from collection to frozen storage was no more than 60 min. Samples were collected blind to the investigators participating in the study and contained no identifying features that would make it possible to identify the subjects. The study was approved by the Institutional Review Board of the University of Pennsylvania.

Removal of High Abundance Proteins from Serum Samples. Serum samples were processed using a 4.6 × 50 mm Multiple Affinity Removal Column (Agilent Technologies, Palo Alto, CA), which selectively removes albumin, IgG, IgA, antitrypsin, transferrin, and haptoglobin from the serum sample, attached to an EZChrome Elite HPLC (Hitachi High Technologies America, San Jose, CA). This column can process 20 µL of human serum per sample run. Samples were processed according to manufacturer’s instructions. For each sample, a low abundance fraction was collected and buffer exchanged into 10 mM Tris-HCl pH 7.4 using 5000 Da molecular weight cutoff spin concentrators (Agilent Technologies, Palo Alto, CA). Protein quantification was performed using Coomassie protein assay reagent (Pierce Biotechnology, Rockford, IL), absorbance at 595 nm, with a Bradford protein assay using bovine serum albumin as a protein standard. Approximately 90% of total serum protein is removed by this method. Cy-dye Labeling. The rationale for using a pooled internal standard with DIGE to control for gel to gel variation has been previously described in detail.16 A total of 50 µg of serum protein was minimally labeled with one of three CyDye DIGE Fluors (Amersham Biosciences, Piscataway, NJ). Individual serum samples from 3 groups (pooled internal standard, cancer and control) were labeled with Cy2 [3-(4-carboxymethyl)phenylmethyl)-3′-ethyloxacarbocyanine halide N-hydroxysuccinimidyl ester]; Cy3 [1-(5-carboxypentyl)-1′-propylindocarbocyanine halide N-hydroxysuccinimidyl ester]; and Cy5 [1-(5-carboxypentyl)-1′-methylindodicarbocyanine halide N-hydroxysuccinimidyl ester], respectively. The three dyes were designed to ensure that proteins common to each sample have the same relative mobility regardless of the dye used to tag them. CyDyes were reconstituted in anhydrous DMF and combined with samples at a ratio of 400 pmol of CyDye to 50 µg of protein. Labeling was performed on ice and in the dark for 30 min. The reaction was then quenched by incubating with 1.5 µL of 10 mM lysine on ice and in the dark for 10 min. 2-D Gel Electrophoresis and Imaging. Three labeled protein samples (one pooled standard, experimental, and control sample) were combined. One milligram of unlabeled pooled standard sample was separately processed by 2-D gel electrophoresis for purposes of protein identification. Proteins were focused on 18 cm, 3-11 immobilized pH gradient (IPG) strips (Amersham Biosciences, Piscataway, NJ) using an IPGphor focusing apparatus (Amersham Biosciences, Piscataway, NJ). IPG strips were then equilibrated in equilibration buffer (50 mM Tris-HCl, 6M urea, 30% glycerol, 2% SDS) supplemented with 1% DTT to maintain the fully reduced state of proteins, followed by 2.5% iodoacetamide to prevent reoxidation of thiol groups during electrophoresis. Proteins were separated on 10% Tris-glycine gels (ProtoGel, National Diagnostics, Atlanta, GA) using an Ettan DALT II System (Amersham Biosciences, Piscataway, NJ). Samples were run in triplicate. The gels were scanned using Typhoon 9410 Variable Mode Imager (Amersham Biosciences, Piscataway, NJ). Excitation/emission wavelengths for Cy2, Cy3, and Cy5 are 488/520, 532/580, and 633/ 670 nm, respectively. The 2-D gel containing 1 mg of unlabeled pooled standard sample was fixed in 30% methanol, 7.5% acetic acid, then Coomassie stained (Colloidal Blue stain kit, Invitrogen, Carlsbad, CA), and scanned with the same imager, with excitation wavelength of 633 nm. DIGE Analysis. Relative protein quantitation across all cancer and control samples was performed using DeCyder Differential Journal of Proteome Research • Vol. 4, No. 5, 2005 1743

research articles In gel Analysis and Biological Variance Analysis software (Version 4.0, Amersham Biosciences, Piscataway, NJ). The presence of a Cy2 labeled pooled internal standard on every gel allowed accurate relative quantitation of protein spotfeatures across different gels. A total of 2093 protein spotfeatures were analyzed across all serum samples. Student’s t-test and one-way ANOVA were used to calculate significant differences in relative abundances of protein spot-features in cancer sera compared with control sera. In-Gel Digestion. Protein spot-features that were significantly increased or decreased (p < 0.01) in all cancer samples compared to benign samples were chosen for further analysis. Protein spot-features of interest were picked using the automated Ettan Spot Picker (Amersham Biosciences, Piscataway, NJ) into 96-well plates (Sigma, St. Louis, MO). The 2 millimeter diameter gel plugs were washed in Milli-Q water for 15 min, then washed three times in 25 mM NH4HCO3, 50% CH3CN for 30 min while vortex-mixing. Gel plugs were then dehydrated in 100% CH3CN for 10 min while vortex-mixing. The supernatant was removed, and gel plugs were allowed to air-dry for 1 h. 12 µL of 1.5 µM trypsin (Promega, Madison, WI) suspended in 25 mM NH4HCO3 was added, and gel plugs were allowed to re-hydrate for 30 min on ice. Gel plugs were placed at 37 °C and allowed to digest overnight. 96-well plates were then gently centrifuged, and supernatant was taken for MS/MS analysis. MS/MS and Database Analysis. Trypsin peptide solutions were mixed at a 1:1 ratio with 5 mg/mL R-cyano-4-hydroxycinnamic acid (CHCA) matrix in 0.3% TFA, and spotted on stainless steel MALDI sample plates (Applied Biosystems, Framingham, MA). Peptide mixtures were then analyzed using MALDI/TOF/TOF (4700 Proteomics Analyzer, Applied Biosystems, Framingham, MA). Protein identification was performed using Global Proteome Server Explorer software (Applied Biosystems, Framingham, MA) utilizing the NCBI Reference Sequence (RefSeq Release 5) human protein database. An identification was assigned to a protein spot feature if the protein score was calculated to be greater than 50, correlating to a confidence interval of 99%. Western Blot Analysis. Fifteen micrograms of nondepleted serum from each patient was diluted 25-fold in LDS sample buffer (Invitrogen, Carlsbad, CA) and incubated at 60 °C for 10 min. Fifteen micrograms of normal human serum (Sigma, St. Louis, MO) were similarly processed and loaded onto each gel. Samples were separated on 4-12% Bis-Tris gels (NuPAGE Novex gels, Invitrogen, Carlsbad, CA) then transferred to nitrocellulose membrane (Invitrogen, Carlsbad, CA). SeeBlue protein standard was used to estimate molecular weights (Invitrogen, Carlsbad, CA). The following primary antibodies were utilized: rabbit anti-human R-1-anti-chymotrypsin polyclonal antibody, rabbit anti-human apolipoprotein E polyclonal antibody and inter-R-trypsin inhibitor rabbit anti-human polyclonal antibody (all from DakoCytomation, Carpinteria, CA). Nitrocellulose membranes were incubated with individual primary antibodies diluted 1:10 000 for 2 h at room temperature. Membranes were then incubated for 45 min with goat anti-rabbit secondary horseradish peroxidase-conjugated antibody (Sigma, St. Louis, MO) diluted 1:10 000. Protein bands were then visualized by incubating membranes with ECL Plus detecting reagents (Amersham Biosciences, Piscataway, NJ). Membrane chemiluminescence was acquired on Typhoon 9410 Variable Mode Imager (Amersham Biosciences, Piscataway, NJ). Quantification of protein expression was performed using Imagequant software (Amersham Biosciences, Piscataway, NJ). 1744

Journal of Proteome Research • Vol. 4, No. 5, 2005

Yu et al.

Figure 1. 2-D DIGE image of raw serum labeled with Cy2 (green) and serum depleted of high abundance proteins labeled with Cy3 (red). The large spot feature at 66kD is albumin.

Chemiluminescence ratios of experimental samples to the normal control sample were calculated, thus controlling for inter-blot chemiluminescence variability. Statistical Analysis. Two-sided, Student’s T-tests were used to analyze differences in protein levels between cancer and control serum. T-tests were also used to determine any effects of patient gender and presence of metastatic disease. A p value of less than 0.05 was considered statistically significant. Linear regression analysis was performed on relative protein levels, serum CA 19-9 levels and patient age. Regression analysis was also performed to determine relationship between relative protein concentration and ECL+ chemiluminescence. Receiver operating characteristic curve analysis to assess the performance of biomarkers in an independent validation set was performed using MedCalc software Version 8 (MedCalc Software, Belgium).

Results Effectiveness and Reproducibility of Immunoaffinity Processing of Serum. Immunoaffinity removal of the six most abundant proteins from serum with the Multiple Affinity Removal Column (Agilent Technologies, Palo Alto, CA) clearly allows visualization of many more proteins. Fifty micrograms of normal human serum and immunoaffinity-depleted serum was labeled with Cy2 and Cy3, respectively. Samples were combined and separated by 2-D gel electrophoresis (Figure 1). On the basis of DeCyder software analysis, 700 protein spot features were identified in the unprocessed serum, compared with 1284 protein spot features in the immunoaffinity-depleted serum. The next objective was to determine the reproducibility of serum processing by immunoaffinity depletion. Three aliquots of pooled, normal human serum were processed using the Multiple Affinity Removal Column (Agilent Technologies, Palo Alto, CA). Fifty micrograms of protein from each sample was labeled with Cy2, Cy3, or Cy5. Samples were combined and separated by 2-D gel electrophoresis. Images were scanned and differences in levels of protein spot features were determined by DeCyder software analysis. A total of 1338 protein spot

Proteins in Human Pancreatic Cancer Serum

Figure 2. DeCyder 2-D DIGE image comparing Cy2 and Cy5 labeled normal serum samples after removal of abundant proteins. Protein spot features increased in Cy5 (blue) or Cy2 (red) labeled sample are highlighted.

features were analyzed across the three control samples. On average, 76 (5.7%) spot features were found to differ by more than 2-fold. The majority of these differences occurred in unfocused areas of the gel, which may not represent true spot features (Figure 2). These findings suggest that immunoaffinity processing introduces a minimal, and therefore, an acceptable amount of variance to protein concentrations in serum. A recent report in the literature comparing five different depletion columns17 independently supports our findings of reproducibility and binding specificity of the Multiple Affinity Removal Column (Agilent Technologies, Palo Alto, CA). Many Proteins are Differentially Expressed in Serum from Pancreatic Patients. In a pilot study, sera from 3 patients with pancreatic cancer were compared to sera from 3 patients with benign pancreatic disease (Table 1). Although it would have been desirable to utilize serum from patients with early disease, at the time of our initial study, only one such sample was available. Samples were processed by immunoaffinity depletion of the six most abundant serum proteins. Differential protein analysis was then performed using DIGE, as previously described. A total of 2093 protein spot-features were analyzed across all serum samples. Fifty-six protein spot-features were found to be significantly up regulated and 43 significantly down regulated in the sera of patients with pancreatic cancer. All spot features of interest were trypsin digested and submitted to MS/MS for identification. We identified 24 unique proteins that were up regulated and 17 unique proteins that were down regulated in serum from pancreatic cancer patients (Tables 2 and 3, respectively). An example of MS and MS/MS spectra used for identification of one of these proteins is shown (Figure 3). Lists of protein accession numbers were analyzed and annotated using Gene Ontology classification. Some proteins can be classified into multiple groups (Figure 4). Proteins involved with a wide variety of biological functions were found to be differentially expressed. Differential Levels Confirmed for Several Candidate Proteins. The proteins inter-R-trypsin inhibitor, R-1-anti-chymot-

research articles rypsin, apolipoprotein E, and complement component 3, were chosen for confirmatory analysis. These proteins were all found to be up regulated in serum from pancreatic cancer patients. DIGE analysis found a 33 kDa protein, identified as apolipoprotein E, to be up-regulated 7.7-fold in the three cancer samples. This protein was the most differentially expressed protein we identified. Western analysis was performed on the six serum samples used for the initial DIGE experiment. In all proteins tested, Western protein bands with molecular weights corresponding to the protein spot features identified by DIGE were clearly overexpressed in the three cancer samples compared with the control samples, confirming our DIGE findings for these proteins (Figure 5). Western Blot Analysis for Relative Protein Quantitation. Antibodies used for protein detection identified several protein bands in addition to the specific protein band identified by DIGE as a potential biomarker. Measurement of protein levels by ELISA would not have accurately reflected levels of our protein of interest. To determine if Western blot analysis could be used to accurately quantitate relative levels of the protein band of interest, 5, 10, 15, 20, and 25 µg of normal human serum (Sigma, St. Louis) were prepared, separated by 1-D gel electrophoresis and transferred to nitrocellulose membrane as described earlier for Western blot analysis. Primary and secondary antibody incubation were performed for rabbit antihuman R-1-anti-chymotrypsin polyclonal antibody, rabbit antihuman apolipoprotein E polyclonal antibody and inter-Rtrypsin inhibitor rabbit anti-human polyclonal antibody as described earlier for Western blot analysis. Protein bands of interest were then visualized by incubating membranes with ECL Plus detecting reagents (Amersham Biosciences, Piscataway, NJ). Membrane chemiluminescence was acquired on Typhoon 9410 Variable Mode Imager (Amersham Biosciences, Piscataway, NJ). Quantification of protein expression was performed using Imagequant software (Amersham Biosciences, Piscataway, NJ). Regression analysis of total amount of protein loaded versus chemiluminescence was performed demonstrating a strong, linear relationship for all antibodies tested (Table 5). Levels of Inter-R-trypsin Inhibitor, R-1-Anti-chymotrypsin, and Apolipoprotein E are Elevated in Pancreatic Cancer Serum. Western analysis was then performed for these four proteins (inter-R-trypsin inhibitor, R-1-anti-chymotrypsin, apolipoprotein E, and complement component 3) in an independent, larger set of serum samples. Sera from 20 pancreatic cancer patients and 14 controls with a variety of benign, biliary, and pancreatic processes were tested (Table 4). Relative protein levels were then quantified using Imagequant software (Amersham Biosciences, Piscataway, NJ), and statistical analysis performed. Levels of all four proteins tested were significantly higher in the serum samples from cancer patients compared with controls (Table 6). Statistical analysis demonstrated that the protein complement component 3 was only weakly discriminatory between cancer and control sets, so this protein was not included in further modeling (data not shown). Statistical Analysis of Relative Protein Levels. Regression analysis was performed to determine if patient age or CA 19-9 could account for differences in relative protein levels seen across all serum samples for inter-R-trypsin inhibitor, R-1-antichymotrypsin and apolipoprotein E. No association was found. Student’s T-tests were performed to determine if levels of these three proteins were associated with the presence of metastases or patient gender. Levels of R-1-anti-chymotrypsin were sigJournal of Proteome Research • Vol. 4, No. 5, 2005 1745

research articles

Yu et al.

Figure 3. Sample MALDI-TOF/MS (A), MS/MS (B) and protein identification (C). The protein spot feature was identified as inter-Rtrypsin inhibitor H4.

nificantly higher in serum from cancer patients with metastatic disease compared to patients with limited disease. Such an association was not found for either inter-R-trypsin inhibitor or apolipoprotein E. Patient gender did not account for differences in levels of any of these proteins (Tables 7 and 8). Discriminatory Ability of Individual Biomarkers. Receiver operating characteristic curves for inter-R-trypsin inhibitor, R-1anti-chymotrypsin and apolipoprotein E were created to study the ability of these proteins to discriminate between pancreatic 1746

Journal of Proteome Research • Vol. 4, No. 5, 2005

cancer and control serum samples in the independent validation set. Thresholds for a positive or negative test were set to maximize the sum of the sensitivity and specificity. Individually, all three biomarkers were reasonably good, with areas under the curve of 0.824 to 0.887. Specificities ranged from 82.4% to 94.1%. Unfortunately, in a disease with a low incidence like pancreatic cancer, such a low specificity would result in too many false positives to make any of these individual proteins a useful biomarker.

research articles

Proteins in Human Pancreatic Cancer Serum

Table 2. List of 24 Proteins Found to Be Up-Regulated in Sera of 3 Pancreatic Cancer Patients Using DIGE and MS protein name

accession no.

MW

pI

fold change

apolipoprotein B precursor inter-alpha (globulin) inhibitor H4 inter-alpha (globulin) inhibitor, H2 complement component 1 inhibitor precursor complement component 3 precursor complement factor B preproprotein alpha-1-antichymotrypsin, precursor zinc finger protein 101 complement component 9 vitronectin precursor kininogen similar to MEGF11 protein apolipoprotein A-IV precursor hypothetical protein FLJ23168 apolipoprotein E nebulin hypothetical protein FLJ37543 plasminogen paraoxonase 1; Paraoxonase [Homo sapiens] structural maintenance of chromosomes 2-like 1 similar to Zinc finger protein 492 structural maintenance of chromosomes 1-like 1 parvalbumin N-myc and STAT interactor

gi|4502153 gi|31542984 gi|4504783 gi|4557379 gi|4557385 gi|4502397 gi|4501843 gi|24475859 gi|4502511 gi|18201911 gi|4504893 gi|42661313 gi|4502151 gi|33943101 gi|4557325 gi|4758794 gi|27734795 gi|4505881 gi|19923106 gi|5453591 gi|41151185 gi|30581135 gi|4506335 gi|4758814

515240.8 103293 106646.8 55118.4 187045.9 85450.5 48606 50306.9 63132.7 54301.2 47852.6 61570.1 45353.4 38248.9 36131.8 772727.2 14806.7 90510.2 39706.2 135696 60733 143144.1 12051 35095

6.61 6.51 6.58 6.23 6.02 6.55 5.79 9.67 5.43 5.55 6.29 9.57 5.33 9.95 5.65 9.1 9.54 7.04 5.08 8.68 9.4 7.51 4.98 5.24

1.56 1.6 1.6 1.6 2.2 1.93 2.5 2.3 1.47 1.94 1.94 2.43 2 2.4 7.7 1.6 1.5 1.93 2.25 1.5 1.7 1.4 1.67 1.8

Table 3. List of 17 Proteins Found to Be Down-Regulated in Sera of 3 Pancreatic Cancer Patients using DIGE and MS/MS protein name

accession no.

MW

pI

fold change

hemopexin similar to Ig alpha-2 chain C region antithrombin ΙΙΙ alpha 2 macroglobulin precursor beta-2-glycoprotein I precursor pregnancy-zone protein nuclei expressed gene 2; nesprin 2; nucleus and actin alpha-2-HS-glycoprotein; Alpha-2HS-glycoprotein coagulation factor X precursor; prothrombinase; factor PEDF serum amyloid P component precursor alpha-1-microglobulin/bikunin precursor H factor (complement)-like3; factor H-related gene 2 transthyretin; prealbumin [Homo sapiens] apolipoprotein L1 isoform a precursor apolipoprotein L1 isoform b precursor FLJ43752 protein [Homo sapiens]

gi|11321561 gi|41203848 gi|4502261 gi|4557225 gi|4557327 gi|4506355 gi|33624848 gi|4502005 gi|4503625 gi|39725934 gi|4502133 gi|4502067 gi|5031695 gi|4507725 gi|21735614 gi|21735616 gi|46409622

51643.3 84871.2 52568.9 163174.9 38286.7 163732.6 795705.4 39299.7 54696.5 46283.3 25371.1 38974 30630.6 15877 43946.9 45889.9 20889.8

6.55 4.63 6.32 6 8.34 5.97 5.26 5.43 5.68 5.97 6.1 5.95 6 5.52 5.6 5.93 11.57

2 1.83 3.14 2.1 2 2.1 2.1 2.8 2.8 1.7 1.8 3 2.17 7.9 1.5 1.5 2

Multivariate Model Utilizing All Three Biomarkers Improves on Individual Biomarkers. A multivariate model was created using all three biomarkers and tested in the validation set of serum samples. Again, thresholds for a positive or negative test were set to maximize the sum of the sensitivity and specificity. This model performed better than any of the

Figure 4. Gene Ontology classification of differentially expressed serum proteins in pancreatic cancer patients.

individual biomarkers, with an area under the curve of 0.959, a sensitivity of 82.6%, and a specificity of 100%.

Discussion The discovery of sensitive and specific serum biomarkers for the early diagnosis of pancreatic cancer remains a daunting challenge. The analogy to finding a needle in a haystack is fitting. The large dynamic range between the least and most abundant of the protein constituents found in serum remains

Figure 5. Western blot analysis confirms DIGE findings of increased protein levels in serum from pancreatic cancer patients for four proteins. Journal of Proteome Research • Vol. 4, No. 5, 2005 1747

research articles

Yu et al.

Figure 6. Receiver operating characteristic curves for the three potential biomarkers identified (A), and a test combining all three biomarkers (B). Sensitivity and specificity are for a cutoff that maximizes their sum. Table 4. Patient Characteristics of Pancreatic Cancer (nos. 1-20) and Control (nos. 21-34) Serum Samples Used to Validate Protein Levels

a

patient no.

age

gender

disease

TNM stagea

AJCC stagea

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

41 77 75 73 59 69 58 75 78 77 61 19 47 76 74 48 72 65 86 66 67 70 45 66 33 56 80 47 37 64 72 63 73 69

F M F M n.d. M M F M M F F F F F M F F M M F M M M F F M F F F M M F M

pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma ampullary adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma ampullary adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic adenocarcinoma pancreatic microcystic serous adenoma benign biliary stricture pseudocyst pancreatic microcystic adenoma acinic cell cystadenoma mucinous cystic pancreatic neoplasm pancreatitis chronic pancreatitis mucinous cystic pancreatic neoplasm mucinous cystic pancreatic neoplasm islet cell tumor pancreatitis pseudocyst chronic pancreatitis

T4N1M1 T4NxM1 T2N1M1 T3N1M1 T4NxM1 T4NxM1 T4NxM1 T4NxM1 T4NxM1 T3N1M0 T3N0M0 T2N0M0 T3N1M0 T2N0M0 T3NxM1 T2N0M0 T3N1M0 T4N0M0 T4NxM1 T4N0M0

IV IV IV IV IV IV IV IV IV IIB IIA IB IIB IB IV IB IIB III IV III

Staging criteria defined by the American Joint Committee on Cancer.39

a difficult hurdle to overcome. The current study approaches this problem by removing the six most abundant proteins from serum samples. This technique proves to be both reproducible and effective in removing these highly abundant proteins, which make up 85-90% of total serum protein by mass. Immunoaffinity-depleted serum samples from patients with pancreatic cancer and controls were then analyzed using DIGE. 41 unique, differentially expressed proteins were identified by 1748

Journal of Proteome Research • Vol. 4, No. 5, 2005

b

CA 19-9b

25 n.d 241 9323 n.d n.d n.d n.d n.d 31 75 n.d 12 629 n.d 10 n.d 61 1265 98 27 n.d n.d n.d 14 2 n.d n.d n.d n.d n.d n.d n.d n.d

The normal range for serum CA 19-9 is 0-37 ng/mL.

MALDI/TOF/TOF-MS. Differences were confirmed by Western analysis in four of these proteins. Further validation studies in an independent set of serum samples from 20 pancreatic cancer patients and 14 controls demonstrated increased levels of three of these proteins, inter-R-trypsin inhibitor, R-1-antichymotrypsin, and apolipoprotein E. Removal of the most abundant proteins from serum remains a controversial technique in biomarker discovery. The removal

research articles

Proteins in Human Pancreatic Cancer Serum Table 5. Regression Analysis of Chemiluminescence versus Protein Concentration primary antibody tested

R2

F

F significance

R-1-anti-chymotrypsin apolipoprotein E inter-R-trypsin inhibitor

0.99088 0.95931 0.98784

326.06 70.72 243.70

0.00037 0.00353 0.00057

Table 6. Comparison of Serum Protein Levels between Patients with Pancreatic Cancer and Benign Disease Relative protein concentration (fold, mean ( SD) protein

cancer

benign disease

P

R-1-anti-chymotrypsin apolipoprotein E inter-R-trypsin inhibitor complement component 3

3.27 ( 3.15 3.44 ( 2.69 2.91 ( 2.17 1.58 ( 0.97

0.80 ( 0.48 0.78 ( 0.44 1.41 ( 0.40 1.01 ( 0.57

0.0028 0.0003 0.0077 0.0364

of these proteins reduces the dynamic range of protein concentrations in the serum sample, which allows for the identification of lower abundance proteins. This removal by immunoaffinity depletion, however, may also specifically or nonspecifically remove other, more interesting proteins that are bound to the abundant proteins.18 Although 2-D DIGE comparing unprocessed serum to immunoaffinity depleted serum demonstrated many more protein spot features in the latter sample, loss of protein spot features is more difficult to assess. To address this question, normal human serum was analyzed with or without immunoaffinity depletion by multidimensional liquid chromatography tandem mass spectrometry (manuscript in press). A total of 120 unique proteins were identified in unprocessed serum, and 224 unique proteins were identified in the depleted serum, confirming our DIGE findings. However, 38 proteins identified in the unprocessed serum could not be identified following depletion, presumably due to nonselective loss. One can conclude, therefore, that removal of these abundant proteins is a double-edged sword. While it would be preferable to study exclusively patients with early stage disease, currently, such a study is not feasible due to the rarity of diagnosis of early stage pancreatic cancer. As such, some markers found in sera from patients with advanced stage pancreatic cancer may not be present in early stage cancer. However, it is reasonable to conclude that some markers detected in advanced stage pancreatic cancer will also be present in early stage disease. Importantly, this approach is likely the only way in which biomarker discovery can be advanced.

Initially, it is likely that a serum-screening test for pancreatic cancer would be best utilized in high risk populations to triage selected patients for confirmatory testing with more invasive modalities such as endoscopic ultrasound, tissue biopsy or exploratory laparotomy. Examples of high-risk patients include individuals with chronic pancreatitis or with suspicious but inconclusive radiological findings. The positive predictive value (PPV) of a test is largely influenced by the prevalence of the disease in the population under study. The prevalence of pancreatic cancer is likely only marginally greater than the annual incidence rate since the disease is rapidly progressive and there is a relatively low rate of discovery of incidental pancreatic cancer at autopsy.19-21 Table 9 shows that a screening test with high sensitivity and specificity could reduce the number of patients who would need to undergo the confirmatory test to identify one pancreatic cancer from approximately 1000 to between 10 and 60 subjects. This paradigm is consistent with the current approach to screening for breast cancer (i.e., mammography followed by image guided biopsy) and prostate cancer (i.e., PSA followed by prostate biopsy). Although a control arm comprised of patients with high risk features is appropriate, measurement of protein levels in a larger set of patients and in subjects without pancreatic disease or with other types of cancer would provide more information about the discriminatory power of any potential biomarkers. One of the proteins that appeared elevated in cancer serum from our pilot study, complement component 3, was not validated in the independent set of serum samples. It is not surprising that one or more of the candidate biomarkers identified in our pilot study would fail validation. This may have occurred for a number of reasons. Only three cancer and control patients were analyzed in the pilot study. Differences among such a small number of samples quite easily could be unrelated to disease state. One of the advantages of using 2-D differential gel electrophoresis, compared to other separation techniques such as multidimensional liquid chromatography, is that different levels of one isoform or one post-translationally modified form of a protein can be ascertained. However, these differences may prove difficult to confirm using standard immunoblotting techniques, due to a lack of specific antibodies. Statistical analysis demonstrated that CA 19-9 levels did not correlate with levels of inter-R-trypsin inhibitor, R-1-antichymotrypsin or apolipoprotein E. As earlier noted, CA 19-9 is not an effective screening marker for pancreatic cancer. In addition, many of the patients recruited for this study had

Table 7. Analysis of Relationship between Biomarker Levels and Metastatic Status or Patient Gender relative protein concentration (fold, mean ( SD)

relative protein concentration (fold, mean ( SD) protein

metastases

limited disease

P

male

female

P

R-1-anti-chymotrypsin apolipoprotein E inter-R-trypsin inhibitor

4.34 ( 3.62 2.27 ( 2.27 3.02 ( 2.42

1.28 ( 0.43 4.48 ( 2.93 2.20 ( 0.84

0.0218 0.0721 0.3444

2.44 ( 3.29 2.04 ( 2.59 2.28 ( 2.13

1.73 ( 1.78 2.42 ( 2.38 2.03 ( 0.89

0.4386 0.6643 0.6524

Table 8. Regression Analysis of Age and CA 19-9 Levels versus Biomarker Levels age

CA 19-9

primary antibody tested

R2

F

F significance

R2

F

F significance

R-1-anti-chymotrypsin apolipoprotein E inter-R-trypsin inhibitor

0.0012 0.0110 0.0013

0.04 0.36 0.04

0.84498 0.55542 0.83908

0.0015 0.0530 0.0211

0.02 0.67 0.26

0.89494 0.42826 0.62000

Journal of Proteome Research • Vol. 4, No. 5, 2005 1749

research articles

Yu et al.

Table 9. PPV of a Screening Test According to Sensitivity, Specificity, and Prevalence of Disease prevalence of pancreatic cancer

sensitivity (%)

specificity (%)

PPV (%)

number needed to test with confirmatory test to diagnose one cancer

.00050 .00075 .00100 .00150 .00200 .00050 .00075 .00100 .00150 .00200

90 90 90 90 90 99 99 99 99 99

97 97 97 97 97 99 99 99 99 99

1.5 2.2 2.9 4.3 5.7 4.7 6.9 9.0 12.9 16.6

67.6 45.4 34.3 23.2 17.6 21.2 14.5 11.1 7.7 6.0

potentially resectable pancreatic cancers, and therefore, normal or only minimally elevated CA 19-9 levels. In this cohort of patients, levels of potential biomarkers would not be expected to correlate well with CA 19-9 levels. Finally, CA 19-9 levels were not obtained at the time of collection for many of these samples, reducing the effectiveness of this analysis. In addition to being significantly elevated in serum from pancreatic cancer patients compared with controls, levels of R-1-anti-chymotrypsin were further elevated in serum from patients with metastatic disease as compared to serum from patients with limited disease. This association with metastatic cancer has been previously observed.22,23 A recent study may shed some light on the mechanisms underlying this association. Levels of R-1-anti-chymotrypsin were found to be increased in a xenograph model of breast cancer that readily metastasized, but not in a model that was weakly metastatic.24 The authors found that R-1-anti-chymotrypsin is readily produced by host tissue in response to the presence of tumor cells, and serum levels were confirmed with immunoblotting. Another major problem often encountered in cancer biomarker discovery is the delineation of proteins that are truly specific for the disease of interest, in this case pancreatic cancer, from proteins that reflect the inflammatory state resulting from the disease. The study was designed with this in mind, using controls with inflammatory conditions of the pancreas and biliary system. Despite this, several of the proteins identified here have been found to be acute phase proteins. For example, studies of the acute phase reaction have demonstrated increased levels of inter-R-trypsin inhibitor25 and R-1anti-chymotrypsin,26 and decreased levels of R1-microglobulin/ bikunin precursor and R2-HS-glycoprotein.25,27 These trends in protein levels were also found in the current study. Some of our results are consistent with a recently published study also investigating serum biomarkers in pancreatic cancer. Koomen et al.,28 utilizing a completely different proteomic profiling approach, identified two of the same proteins as in this study, inter-R-trypsin inhibitor and R-1-anti-chymotrypsin, increased in pancreatic cancer serum. Given the extraordinary local inflammatory reaction that is typically found in pancreatic cancer, identification of proteins related to such inflammation may be inescapable. Similarly, several of the proteins identified in this study have been found to be potential biomarkers for other cancers, for example increased levels of inter-R-trypsin inhibitor and decreased levels of transthyretin in ovarian cancer.29 Rather than ignoring these proteins, studies to identify isoforms or post-translationally modified forms of these proteins specific to pancreatic cancer could lead to the identification of more specific disease biomarkers. Such studies would 1750

Journal of Proteome Research • Vol. 4, No. 5, 2005

be akin to the identification and use of the cardiac form of creatine kinase (CK-MB) as a highly specific marker of myocardial infarction, compared to general creatine kinase, which is released into circulation with any kind of muscle damage. Apolipoprotein E, which was found in this study to be increased in pancreatic cancer serum, has not previously been identified as a marker of pancreatic cancer. The chief role of apolipoprotein E is to transport lipids. Polymorphisms in apolipoprotein E have been associated with increased risk for atherosclerosis30 and neurodegenerative diseases.31,32,33-35 Recent studies are beginning to show important antioxidant properties of apolipoprotein E which may protect against these diseases.36 Other studies have shown that apolipoprotein E can be localized to the pancreas in other primates.37 Elevated levels of apolipoprotein E have been described as elevated in type II diabetes,38 a condition often seen in patients with pancreatic cancer. The diabetic status of patients in this study was not known. It is not surprising that our model utilizing several different biomarkers performed better than any individual biomarker. Pancreatic cancer is a heterogeneous disease, and one would expect that tumors from different patients would produce different tumor biomarkers. The more of these biomarkers that are included in a screening test, the greater the sensitivity and specificity this test would possess. Future efforts will focus on better characterizing the proteins identified in the current study. Identification of specific posttranslational modifications may allow for improved sensitivity and specificity in diagnosing pancreatic cancer. Validation of other proteins identified in the original pilot study as differentially expressed in serum from pancreatic cancer patients will also be pursued. This study also illustrates some of the limitations of current proteomic techniques. Current 2-D DIGE methods are relatively low throughput, not allowing for the comparison of many different samples. This may reduce the discriminatory power of any potential biomarkers identified. Despite removal of the most abundant proteins, this study also demonstrates that improved sensitivity is still required to dig deeper into the serum proteome. Novel labeling techniques and use of multidimensional liquid chromatography coupled with MS/MS are being studied as methods for achieving this sensitivity.

Acknowledgment. We thank Stephen Hahn and Ernest Rosato for their assistance in obtaining study serum samples. This work was supported by NIH Grant Nos. RO1 CA95586, S10 RR019221, and P30 DK050306, the Center for Molecular Studies in Digestive and Liver Diseases (and its morphology, molecular biology, and cell culture core facilities), and an NIH T32 DK007066 training grant in GI basic sciences. References (1) Niederhuber, J. E.; Brennan, M. F.; Menck, H. R. Cancer 1995, 76, 1671-1677. (2) Jemal, A.; Murray, T.; Samuels, A.; Ghafoor, A.; Ward, E.; Thun, M. J. CA Cancer J. Clin. 2003, 53, 5-26. (3) Ahrendt, S. A.; Pitt, H. A. Oncology (Huntingt) 2002, 16, 725734. (4) Birk, D.; Fortnagel, G.; Formentini, A.; Beger, H. G. J. Hepatobiliary. Pancreat. Surg. 1998, 5, 450-454. (5) Tsuchiya, R.; Tsunoda, T. Int. J. Pancreatol. 1990, 7, 117-123. (6) Abrams, R. A.; Grochow, L. B.; Chakravarthy, A.; Sohn, T. A.; Zahurak, M. L.; Haulk, T. L.; Ord, S.; Hruban, R. H.; Lillemoe, K. D.; Pitt, H. A.; Cameron, J. L.; Yeo, C. J. Int. J. Radiat. Oncol. Biol. Phys. 1999, 44, 1039-1046. (7) Ritts, R. E.; Pitt, H. A. Surg. Oncol. Clin. N. Am. 1998, 7, 93-101. (8) Furuya, N.; Kawa, S.; Hasebe, O.; Tokoo, M.; Mukawa, K.; Maejima, S.; Oguchi, H. Br. J. Cancer 1996, 73, 372-376.

research articles

Proteins in Human Pancreatic Cancer Serum (9) Albert, M. B.; Steinberg, W. M.; Henry, J. P. Dig. Dis. Sci. 1988, 33, 1223-1225. (10) Rosty, C.; Goggins, M. Hematol. Oncol. Clin. North Am. 2002, 16, 37-52. (11) Ching, C. K.; Rhodes, J. M. Br. J. Cancer 1989, 59, 949-953. (12) Uehara, H.; Nakaizumi, A.; Tatsuta, M.; Baba, M.; Takenaka, A.; Uedo, N.; Sakai, N.; Yano, H.; Iishi, H.; Ohigashi, H.; Ishikawa, O.; Okada, S.; Kakizoe, T. Am. J. Gastroenterol. 1999, 94, 25132518. (13) Yokoyama, M.; Ochi, K.; Ichimura, M.; Mizushima, T.; Shinji, T.; Koide, N.; Tsurumi, T.; Hasuoka, H.; Harada, M. Pancreas 2002, 24, 344-347. (14) Fukushima, N.; Sato, N.; Ueki, T.; Rosty, C.; Walter, K. M.; Wilentz, R. E.; Yeo, C. J.; Hruban, R. H.; Goggins, M. Am. J. Pathol. 2002, 160, 1573-1581. (15) Pieper, R.; Su, Q.; Gatlin, C. L.; Huang, S. T.; Anderson, N. L.; Steiner, S. Proteomics 2003, 3, 422-432. (16) Alban, A.; David, S. O.; Bjorkesten, L.; Andersson, C.; Sloge, E.; Lewis, S.; Currie, I. Proteomics 2003, 3, 36-44. (17) Bjorhall, K.; Miliotis, T.; Davidsson, P. Proteomics 2005, 5, 307317. (18) Qian, W. J.; Jacobs, J. M.; Camp, D. G.; Monroe, M. E.; Moore, R. J.; Gritsenko, M. A.; Calvano, S. E.; Lowry, S. F.; Xiao, W.; Moldawer, L. L.; Davis, R. W.; Tompkins, R. G.; Smith, R. D. Proteomics 2005, 5, 572-584. (19) Ries, L. A. G.; Eisner, M. P.; Kosary, C.; Hankey, B. F.; Miller, B. A.; Clegg, L.; Mariotto, A.; Feuer, E. J.; Edwards, B. K. “SEER Cancer Statistics Review, 1973-1999”; National Cancer Institute: Bethesda, MD, 2002. (20) Kimura, W.; Morikane, K.; Esaki, Y.; Chan, W. C.; Pour, P. M. Cancer 1998, 82, 1839-1849. (21) Pour, P. M.; Sayed, S.; Sayed, G. Am. J. Clin. Pathol. 1982, 77, 137-152. (22) Higashiyama, M.; Doi, O.; Yokouchi, H.; Kodama, K.; Nakamori, S.; Tateishi, R. Cancer 1995, 76, 1368-1376. (23) Tahara, E.; Ito, H.; Taniyama, K.; Yokozaki, H.; Hata, J. Hum. Pathol. 1984, 15, 957-964. (24) Montel, V.; Pestonjamasp, K.; Mose, E.; Tarin, D. Differentiation 2005, 73, 88-98. (25) Daveau, M.; Jean, L.; Soury, E.; Olivier, E.; Masson, S.; Lyoumi, S.; Chan, P.; Hiron, M.; Lebreton, J. P.; Husson, A.; Jegou, S.; Vaudry, H.; Salier, J. P. Arch. Biochem. Biophys. 1998, 350, 315323.

(26) Thompson, D.; Milford-Ward, A.; Whicher, J. T. Ann. Clin. Biochem. 1992, 29, 123-131. (27) Yoshida, K.; Suzuki, Y.; Yamamoto, K.; Sinohara, H. Biol. Chem. 1999, 380, 95-99. (28) Koomen, J. M.; Shih, L. N.; Coombes, K. R.; Li, D.; Xiao, L. C.; Fidler, I. J.; Abbruzzese, J. L.; Kobayashi, R. Clin. Cancer Res. 2005, 11, 1110-1118. (29) Zhang, Z.; Bast, R. C., Jr.; Yu, Y.; Li, J.; Sokoll, L. J.; Rai, A. J.; Rosenzweig, J. M.; Cameron, B.; Wang, Y. Y.; Meng, X. Y.; Berchuck, A.; Haaften-Day, C.; Hacker, N. F.; de Bruijn, H. W.; van der Zee, A. G.; Jacobs, I. J.; Fung, E. T.; Chan, D. W. Cancer Res. 2004, 64, 5882-5890. (30) Davignon, J.; Gregg, R. E.; Sing, C. F. Arteriosclerosis 1988, 8, 1-21. (31) Lee, S. H.; Williams, M. V.; DuBois, R. N.; Blair, I. A. J. Biol. Chem. 2005, Jun 17; [Epub ahead of print]. (32) Oe, T.; Lee, S. H.; Silva Elipe, M. V.; Arison, B. H.; Blair, I. A. Chem. Res. Toxicol. 2003, 16, 1598-1605. (33) Corder, E. H.; Saunders, A. M.; Strittmatter, W. J.; Schmechel, D. E.; Gaskell, P. C.; Small, G. W.; Roses, A. D.; Haines, J. L.; PericakVance, M. A. Science 1993, 261, 921-923. (34) Roses, A. D. Annu. Rev. Med. 1996, 47, 387-400. (35) Li, Y. J.; Hauser, M. A.; Scott, W. K.; Martin, E. R.; Booze, M. W.; Qin, X. J.; Walter, J. W.; Nance, M. A.; Hubble, J. P.; Koller, W. C.; Pahwa, R.; Stern, M. B.; Hiner, B. C.; Jankovic, J.; Goetz, C. G.; Small, G. W.; Mastaglia, F.; Haines, J. L.; Pericak-Vance, M. A.; Vance, J. M. Neurology 2004, 62, 2005-2009. (36) Pham, T.; Kodvawala, A.; Hui, D. Y. Biochemistry 2005, 44, 75777582. (37) Lin, C. T.; Xu, Y. F.; Wu, J. Y.; Chan, L. J. Clin. Invest. 1986, 78, 947-958. (38) Bach-Ngohou, K.; Ouguerram, K.; Nazih, H.; Maugere, P.; RipollesPiquer, B.; Zair, Y.; Frenais, R.; Krempf, M.; Bard, J. M. Int. J. Obes. Relat. Metab. Disord. 2002, 26, 1451-1458. (39) Exocrine pancreas, In American Joint Committee on Cancer: AJCC Cancer Staging Manual, 6th ed.; Greene, F. L., Page, D. L., Fleming, I. D., Eds.; Springer: New York, 2002; pp 157-164.

PR050174L

Journal of Proteome Research • Vol. 4, No. 5, 2005 1751