Proteomic Classification of Pancreatic Adenocarcinoma Tissue Using Protein Chip Technology
2006; Elsevier BV; Volume: 130; Issue: 6 Linguagem: Inglês
10.1053/j.gastro.2006.02.036
ISSN1528-0012
AutoresChristopher J. Scarlett, Ross C. Smith, Alex Saxby, Aiqun Nielsen, Jaswinder S. Samra, Susan R. Wilson, Robert C. Baxter,
Tópico(s)Advanced Biosensing Techniques and Applications
ResumoBackground & Aims: Pancreatic adenocarcinoma is a most devastating cancer that presents late and is rapidly progressive. This study aimed to identify unique, tissue-specific protein biomarkers capable of differentiating pancreatic adenocarcinoma (PC) from adjacent uninvolved pancreatic tissue (AP), benign pancreatic disease (B), and nonmalignant tumor tissue (NM). Methods: Tissue samples representing PC (n = 31), AP (n = 44), and B (n = 19) tissue were analyzed on hydrophobic protein chip arrays by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry. Training models were developed using logistic regression and validated using the 10-fold cross-validation approach. Results: The hydrophobic protein chip array revealed 13 protein peaks differentially expressed between PC and AP (receiver operating characteristic [ROC] area under the curve [AUC], 0.64–0.85), 8 between PC and B (ROC AUC, 0.67–0.78), and 12 between PC and NM tissue (ROC AUC, 0.63–0.81). Logistic regression and cross-validation identified overlapping panels of peaks to develop a training model that distinguished PC from AP (77.4% sensitivity, 84.1% specificity), B (83.9% sensitivity, 78.9% specificity), and NM tissue (58.1% sensitivity, 90.5% specificity). The final panels selected correctly classified 80.6% of PC and 88.6% of AP samples (ROC AUC, 0.92), 93.5% of PC and 89.5% of B samples (ROC AUC, 0.99), and 71.0% of PC and 92.1% of NM samples (ROC AUC, 0.91). Conclusions: This study used surface-enhanced laser desorption/ionization time-of-flight mass spectrometry to discover a number of protein panels that can distinguish effectively between pancreatic adenocarcinoma, benign, and adjacent pancreatic tissue. Identification of these proteins will add to our understanding of the biology of pancreatic cancer. Furthermore, these protein panels may have important diagnostic implications. Background & Aims: Pancreatic adenocarcinoma is a most devastating cancer that presents late and is rapidly progressive. This study aimed to identify unique, tissue-specific protein biomarkers capable of differentiating pancreatic adenocarcinoma (PC) from adjacent uninvolved pancreatic tissue (AP), benign pancreatic disease (B), and nonmalignant tumor tissue (NM). Methods: Tissue samples representing PC (n = 31), AP (n = 44), and B (n = 19) tissue were analyzed on hydrophobic protein chip arrays by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry. Training models were developed using logistic regression and validated using the 10-fold cross-validation approach. Results: The hydrophobic protein chip array revealed 13 protein peaks differentially expressed between PC and AP (receiver operating characteristic [ROC] area under the curve [AUC], 0.64–0.85), 8 between PC and B (ROC AUC, 0.67–0.78), and 12 between PC and NM tissue (ROC AUC, 0.63–0.81). Logistic regression and cross-validation identified overlapping panels of peaks to develop a training model that distinguished PC from AP (77.4% sensitivity, 84.1% specificity), B (83.9% sensitivity, 78.9% specificity), and NM tissue (58.1% sensitivity, 90.5% specificity). The final panels selected correctly classified 80.6% of PC and 88.6% of AP samples (ROC AUC, 0.92), 93.5% of PC and 89.5% of B samples (ROC AUC, 0.99), and 71.0% of PC and 92.1% of NM samples (ROC AUC, 0.91). Conclusions: This study used surface-enhanced laser desorption/ionization time-of-flight mass spectrometry to discover a number of protein panels that can distinguish effectively between pancreatic adenocarcinoma, benign, and adjacent pancreatic tissue. Identification of these proteins will add to our understanding of the biology of pancreatic cancer. Furthermore, these protein panels may have important diagnostic implications. Pancreatic cancer is one of the most devastating cancers because it presents relatively late and is rapidly progressive. In 2004, there were an estimated 32,000 new pancreatic cancer cases in the United States, with an approximately equal number of deaths.1Jemal A. Tiwari R.C. Murray T. Ghafoor A. Samuels A. Ward E. Feuer E.J. Thun M.J. Cancer statistics, 2004.CA Cancer J Clin. 2004; 54: 8-29Crossref PubMed Scopus (3943) Google Scholar Only about 20% of patients who present with clinical symptoms have a resectable tumor, leading to a 5-year survival of less than 5%.2Jemal A. Murray T. Samuels A. Ghafoor A. Ward E. Thun M.J. Cancer statistics, 2003.CA Cancer J Clin. 2003; 53: 5-26Crossref PubMed Scopus (3394) Google Scholar Consequently early detection and targeted therapy is paramount for improving a patient's chance of survival. Early detection of small resectable cancers via an imaging test along with identification of tumor markers of benign precancerous lesions or early localized cancers may be one way to improve outcome.3Yeo T.P. Hruban R.H. Leach S.D. Wilentz R.E. Sohn T.A. Kern S.E. Iacobuzio-Donahue C.A. Maitra A. Goggins M. Canto M.I. Abrams R.A. Laheru D. Jaffee E.M. Hidalgo M. Yeo C.J. Pancreatic cancer.Curr Probl Cancer. 2002; 26: 176-275Abstract Full Text Full Text PDF PubMed Scopus (251) Google Scholar Endoscopic ultrasonography (EUS) is an emerging imaging modality used in conjunction with spiral computed tomography scanning for screening and early detection, thereby improving the detection of smaller pancreatic lesions.4Legmann P. Vignaux O. Dousset B. Baraza A.J. Palazzo L. Dumontier I. Coste J. Louvel A. Roseau G. Couturier D. Bonnin A. Pancreatic tumors comparison of dual-phase helical CT and endoscopic sonography.AJR Am J Roentgenol. 1998; 170: 1315-1322Crossref PubMed Scopus (403) Google Scholar, 5Rosch T. Braig C. Gain T. Feuerbach S. Siewert J.R. Schusdziarra V. Classen M. Staging of pancreatic and ampullary carcinoma by endoscopic ultrasonography. Comparison with conventional sonography, computed tomography, and angiography.Gastroenterology. 1992; 102: 188-199PubMed Google Scholar, 6Bluemke D.A. Cameron J.L. Hruban R.H. Pitt H.A. Siegelman S.S. Soyer P. Fishman E.K. Potentially resectable pancreatic adenocarcinoma spiral CT assessment with surgical and pathologic correlation.Radiology. 1995; 197: 381-385PubMed Google Scholar EUS also can discriminate readily between solid and cystic lesions. Fine-needle aspiration (FNA) can be performed during an EUS procedure and a cytologic diagnosis of small lesions can be provided. The sensitivity of EUS FNA for the diagnosis of pancreatic cancer has varied from 70% to 90%.7Gress F. Gottlieb K. Sherman S. Lehman G. Endoscopic ultrasonography-guided fine-needle aspiration biopsy of suspected pancreatic cancer.Ann Intern Med. 2001; 134: 459-464Crossref PubMed Scopus (305) Google Scholar, 8Harewood G.C. Wiersema M.J. Endosonography-guided fine needle aspiration biopsy in the evaluation of pancreatic masses.Am J Gastroenterol. 2002; 97: 1386-1391Crossref PubMed Google Scholar, 9Suits J. Frazee R. Erickson R.A. Endoscopic ultrasound and fine needle aspiration for the evaluation of pancreatic masses.Arch Surg. 1999; 134: 639-643Crossref PubMed Scopus (128) Google Scholar, 10Varadarajulu S. Wallace M.B. Applications of endoscopic ultrasonography in pancreatic cancer.Cancer Control. 2004; 11: 15-22PubMed Google Scholar, 11Chhieng D.C. Jhala D. Jhala N. Eltoum I. Chen V.K. Vickers S. Heslin M.J. Wilcox C.M. Eloubeidi M.A. Endoscopic ultrasound-guided fine-needle aspiration biopsy a study of 103 cases.Cancer. 2002; 96: 232-239Crossref PubMed Scopus (124) Google Scholar, 12Ylagan L.R. Edmundowicz S. Kasal K. Walsh D. Lu D.W. Endoscopic ultrasound guided fine-needle aspiration cytology of pancreatic carcinoma a 3-year experience and review of the literature.Cancer. 2002; 96: 362-369Crossref PubMed Scopus (110) Google Scholar The current lack of effective treatment with surgery, radiotherapy, and chemotherapy is stimulating new diagnostic applications and therapeutic modalities in cancer therapy.13Clark J.W. Glicksman A.S. Wanebo H.J. Systemic and adjuvant therapy for patients with pancreatic carcinoma.Cancer. 1996; 78: 688-693Crossref PubMed Google Scholar, 14Gunzburg W.H. Salmons B. Novel clinical strategies for the treatment of pancreatic carcinoma.Trends Mol Med. 2001; 7: 30-37Abstract Full Text Full Text PDF PubMed Scopus (39) Google Scholar Proteomics is emerging as a powerful new tool for the diagnosis of hidden tumors. Because pancreatic cancer presents late as a result of its protected position in the body, proteomics promises to have an important place in the diagnosis and management of such a concealed tumor when used as an adjunct to other diagnostic methods. Protein chip technology coupled with surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) allows for high-throughput screening of protein expression within tissue, serum, and other complex biological mixtures in such a way that rapid comparisons can be made between cancer and normal samples. This technology has been used for differentiating a number of cancers such as prostate, breast, ovarian, head and neck, colorectal, hepatocellular, gastric, cervical, lung, and bladder carcinoma from controls,15Wright Jr, G.L. Cazares L.H. Leung S.M. Nasim S. Adam B.L. Yip T.T. Schellhammer P.F. Gong L. Vlahou A. Proteinchip(R) surface enhanced laser desorption/ionization (SELDI) mass spectrometry a novel protein biochip technology for detection of prostate cancer biomarkers in complex protein mixtures.Prostate Cancer Prostatic Dis. 1999; 2: 264-276Crossref PubMed Scopus (262) Google Scholar, 16Kozak K.R. Amneus M.W. Pusey S.M. Su F. Luong M.N. Luong S.A. Reddy S.T. Farias-Eisner R. Identification of biomarkers for ovarian cancer using strong anion-exchange ProteinChips potential use in diagnosis and prognosis.Proc Natl Acad Sci U S A. 2003; 100: 12343-12348Crossref PubMed Scopus (240) Google Scholar, 17Melle C. Ernst G. Schimmel B. Bleul A. Koscielny S. Wiesner A. Bogumil R. Moller U. Osterloh D. Halbhuber K.J. Von Eggeling F. Biomarker discovery and identification in laser microdissected head and neck squamous cell carcinoma with ProteinChip(R) technology, two-dimensional gel electrophoresis, tandem mass spectrometry, and immunohistochemistry.Mol Cell Proteomics. 2003; 2: 443-452PubMed Scopus (85) Google Scholar, 18Petricoin E.F. Ardekani A.M. Hitt B.A. Levine P.J. Fusaro V.A. Steinberg S.M. Mills G.B. Simone C. Fishman D.A. Kohn E.C. Liotta L.A. Use of proteomic patterns in serum to identify ovarian cancer.Lancet. 2002; 359: 572-577Abstract Full Text Full Text PDF PubMed Scopus (2902) Google Scholar, 19Paweletz C.P. Trock B. Pennanen M. Tsangaris T. Magnant C. Liotta L.A. Petricoin 3rd, E.F. Proteomic patterns of nipple aspirate fluids obtained by SELDI-TOF potential for new biomarkers to aid in the diagnosis of breast cancer.Dis Markers. 2001; 17: 301-307Crossref PubMed Scopus (266) Google Scholar, 20Shiwa M. Nishimura Y. Wakatabe R. Fukawa A. Arikuni H. Ota H. Kato Y. Yamori T. Rapid discovery and identification of a tissue-specific tumor biomarker from 39 human cancer cell lines using the SELDI ProteinChip platform.Biochem Biophys Res Commun. 2003; 309: 18-25Crossref PubMed Scopus (110) Google Scholar, 21Vlahou A. Schellhammer P.F. Mendrinos S. Patel K. Kondylis F.I. Gong L. Nasim S. Wright Jr, G.L. Development of a novel proteomic approach for the detection of transitional cell carcinoma of the bladder in urine.Am J Pathol. 2001; 158: 1491-1502Abstract Full Text Full Text PDF PubMed Scopus (398) Google Scholar, 22Xiao X. Liu D. Tang Y. Guo F. Xia L. Liu J. He D. Development of proteomic patterns for detecting lung cancer.Dis Markers. 2003; 19: 33-39Crossref PubMed Scopus (90) Google Scholar, 23Zhukov T.A. Johanson R.A. Cantor A.B. Clark R.A. Tockman M.S. Discovery of distinct protein profiles specific for lung tumors and pre-malignant lung lesions by SELDI mass spectrometry.Lung Cancer. 2003; 40: 267-279Abstract Full Text Full Text PDF PubMed Scopus (182) Google Scholar, 24Wadsworth J.T. Somers K.D. Stack Jr, B.C. Cazares L. Malik G. Adam B.L. Wright Jr, G.L. Semmes O.J. Identification of patients with head and neck cancer using serum protein profiles.Arch Otolaryngol Head Neck Surg. 2004; 130: 98-104Crossref PubMed Scopus (109) Google Scholar, 25Wellmann A. Wollscheid V. Lu H. Ma Z.L. Albers P. Schutze K. Rohde V. Behrens P. Dreschers S. Ko Y. Wernert N. Analysis of microdissected prostate tissue with ProteinChip arrays—a way to new insights into carcinogenesis and to diagnostic tools.Int J Mol Med. 2002; 9: 341-347PubMed Google Scholar, 26Wong Y.F. Cheung T.H. Lo K.W. Wang V.W. Chan C.S. Ng T.B. Chung T.K. Mok S.C. Protein profiling of cervical cancer by protein-biochips proteomic scoring to discriminate cervical cancer from normal cervix.Cancer Lett. 2004; 211: 227-234Abstract Full Text Full Text PDF PubMed Scopus (67) Google Scholar, 27Zhang Y.F. Wu D.L. Guan M. Liu W.W. Wu Z. Chen Y.M. Zhang W.Z. Lu Y. Tree analysis of mass spectral urine profiles discriminates transitional cell carcinoma of the bladder from noncancer patient.Clin Biochem. 2004; 37: 772-779Crossref PubMed Scopus (73) Google Scholar, 28Paradis V. Degos F. Dargere D. Pham N. Belghiti J. Degott C. Janeau J.L. Bezeaud A. Delforge D. Cubizolles M. Laurendeau I. Bedossa P. Identification of a new marker of hepatocellular carcinoma by serum protein profiling of patients with chronic liver diseases.Hepatology. 2005; 41: 40-47Crossref PubMed Scopus (225) Google Scholar, 29Melle C. Kaufmann R. Hommann M. Bleul A. Driesch D. Ernst G. von Eggeling F. Proteomic profiling in microdissected hepatocellular carcinoma tissue using ProteinChip technology.Int J Oncol. 2004; 24: 885-891PubMed Google Scholar, 30Pawlik T.M. Fritsche H. Coombes K.R. Xiao L. Krishnamurthy S. Hunt K.K. Pusztai L. Chen J.N. Clarke C.H. Arun B. Hung M.C. Kuerer H.M. Significant differences in nipple aspirate fluid protein expression between healthy women and those with breast cancer demonstrated by time-of-flight mass spectrometry.Breast Cancer Res Treat. 2005; 89: 149-157Crossref PubMed Scopus (95) Google Scholar, 31Sauter E.R. Shan S. Hewett J.E. Speckman P. Du Bois G.C. Proteomic analysis of nipple aspirate fluid using SELDI-TOF-MS.Int J Cancer. 2004; 114: 791-796Crossref Scopus (95) Google Scholar, 32Laronga C. Becker S. Watson P. Gregory B. Cazares L. Lynch H. Perry R.R. Wright Jr, G.L. Drake R.R. Semmes O.J. SELDI-TOF serum profiling for prognostic and diagnostic classification of breast cancers.Dis Markers. 2003; 19: 229-238Crossref PubMed Scopus (83) Google Scholar, 33Becker S. Cazares L.H. Watson P. Lynch H. Semmes O.J. Drake R.R. Laronga C. Surfaced-enhanced laser desorption/ionization time-of-flight (SELDI-TOF) differentiation of serum protein profiles of BRCA-1 and sporadic breast cancer.Ann Surg Oncol. 2004; 11: 907-914Crossref PubMed Scopus (81) Google Scholar, 34Ebert M.P. Meuer J. Wiemer J.C. Schulz H.U. Reymond M.A. Traugott U. Malfertheiner P. Rocken C. Identification of gastric cancer patients by serum protein profiling.J Proteome Res. 2004; 3: 1261-1266Crossref PubMed Scopus (87) Google Scholar, 35Grizzle W.E. Adam B.L. Bigbee W.L. Conrads T.P. Carroll C. Feng Z. Izbicka E. Jendoubi M. Johnsey D. Kagan J. Leach R.J. McCarthy D.B. Semmes O.J. Srivastava S. Thompson I.M. Thornquist M.D. Verma M. Zhang Z. Zou Z. Serum protein expression profiling for cancer detection validation of a SELDI-based approach for prostate cancer.Dis Markers. 2003; 19: 185-195Crossref PubMed Scopus (57) Google Scholar, 36Liu A.Y. Zhang H. Sorensen C.M. Diamond D.L. Analysis of prostate cancer by proteomics using tissue specimens.J Urol. 2005; 173: 73-78Abstract Full Text Full Text PDF PubMed Scopus (81) Google Scholar, 37Krieg R.C. Fogt F. Braunschweig T. Herrmann P.C. Wollscheidt V. Wellmann A. ProteinChip array analysis of microdissected colorectal carcinoma and associated tumor stroma shows specific protein bands in the 3.4 to 3.6 kDa range.Anticancer Res. 2004; 24: 1791-1796PubMed Google Scholar, 38Albrethsen J. Bogebo R. Gammeltoft S. Olsen J. Winther B. Raskov H. Upregulated expression of human neutrophil peptides 1, 2 and 3 (HNP 1-3) in colon cancer serum and tumours a biomarker study.BMC Cancer. 2005; 5: 8Crossref PubMed Scopus (140) Google Scholar while markers for pancreatic cancer also have been documented in both human serum39Koopmann J. Zhang Z. White N. Rosenzweig J. Fedarko N. Jagannath S. Canto M.I. Yeo C.J. Chan D.W. Goggins M. Serum diagnosis of pancreatic adenocarcinoma using surface-enhanced laser desorption and ionization mass spectrometry.Clin Cancer Res. 2004; 10: 860-868Crossref PubMed Scopus (266) Google Scholar, 40Yu Y. Chen S. Wang L.S. Chen W.L. Guo W.J. Yan H. Zhang W.H. Peng C.H. Zhang S.D. Li H.W. Chen G.Q. Prediction of pancreatic cancer by serum biomarkers using surface-enhanced laser desorption/ionization-based decision tree classification.Oncology. 2005; 68: 79-86Crossref PubMed Scopus (92) Google Scholar and pancreatic juice.41Rosty C. Christa L. Kuzdzal S. Baldwin W.M. Zahurak M.L. Carnot F. Chan D.W. Canto M. Lillemoe K.D. Cameron J.L. Yeo C.J. Hruban R.H. Goggins M. Identification of hepatocarcinoma-intestine-pancreas/pancreatitis-associated protein I as a biomarker for pancreatic ductal adenocarcinoma by protein biochip technology.Cancer Res. 2002; 62: 1868-1875PubMed Google Scholar Recently, Chen et al42Chen R. Pan S. Brentnall T.A. Aebersold R. Proteomic profiling of pancreatic cancer for biomarker discovery.Mol Cell Proteomics. 2005; 4: 523-533Crossref PubMed Scopus (143) Google Scholar reviewed the progress and challenges for applying proteomics approaches for biomarker discovery in pancreatic cancer, and suggested that the most direct approach is to identify biomarkers in tissue. Although tissue is not as clinically accessible as blood or other body fluids, candidate biomarkers identified through proteomic profiling of tissue can provide an integral base for biomarker identification in serum. SELDI-TOF MS has the potential to provide a means of detection and identification of known proteins associated with cancer progression, and provide an opportunity for identification of novel targets associated with pancreatic cancer. This study was designed to identify unique, tissue-specific protein biomarkers capable of differentiating pancreatic adenocarcinoma from benign pancreatic and adjacent pancreatic tissue. Informed consent was obtained from a consecutive group of 50 patients, undergoing pancreatic resection for a pancreatic mass or cyst, to obtain tissue samples for a pancreatic cancer tissue bank. This protocol was approved by the Northern Sydney Health Human Research Ethics Committee (Sydney, Australia). For this study, 50 pancreatic tumor samples and 44 adjacent uninvolved pancreatic tissue (AP) samples were collected from both male (n = 22) and female (n = 28) patients. The pancreatic tumors were classified as invasive ductal pancreatic adenocarcinoma (n = 31), although none of the benign group had histologic evidence of malignant conversion (n = 19). The final histologic diagnosis of the benign group was mucinous cystadenomas (n = 7), serous cystadenomas (n = 5), solid pseudopapillary tumors (n = 2), islet cell tumors (n = 2), intraductal mucinous papillary neoplasm (n = 1), pancreatic intraepithelial neoplasia 1-A (n = 1), and chronic pancreatitis (n = 1) (Table 1).Table 1Details of Patients' Clinical Information for Pancreatic Tumor and Adjacent Pancreatic Tissue SamplesnMean age, yRangePancreatic tumor samples5066.232–80 Male2266.347–80 Female2866.132–80Pancreatic ductal adenocarcinoma3168.439–80 Stage I1366.639–79 Stage II374.765–80 Stage III971.849–79 Stage IV661.547–79Adjacent pancreatic tissue4466.932–80Benign pancreatic tissue1962.732–80 Mucinous cystadenoma758.632–75 Serous cystadenoma568.448–80 Solid pseudopapillary tumor256.550–63 Islet cell tumor264.049–74 Intraductal mucinous papillary neoplasm170.0 Pancreatic intraepithelial neoplasia I-A165.0 Chronic pancreatitis163.0 Open table in a new tab Resected specimens were taken immediately to the pathologist where they were painted for assessment of margins, opened for selection of a representative sample of tumor and adjacent pancreatic tissue, and then snap frozen in liquid nitrogen. The time from resection to sample freezing was less than 10 minutes. Tissues were stored at −80°C until analyzed. Pancreatic adenocarcinomas were classified according to the Union Internationale Contre le Cancer TNM classification for pancreatic carcinoma.43Tsunoda T. Ura K. Eto T. Matsumoto T. Tsuchiya R. UICC and Japanese stage classifications for carcinoma of the pancreas.Int J Pancreatol. 1991; 8: 205-214PubMed Google Scholar The 31 pancreatic adenocarcinoma samples were classified into pathologic stages comprising 16 early-stage (stage I, n = 13; stage II, n = 3) and 15 late-stage (stage III, n = 9; stage IV, n = 6) pancreatic adenocarcinoma samples (Table 1). Stage IV pancreatic adenocarcinoma samples were collected as biopsy specimens from unresectable malignant tumors. Pancreatic tissue (∼50 mg) was ground to a fine powder in liquid nitrogen, solubilized by pestle homogenization in 0.6 mL of 9.5 mol/L urea/2% 3-[(3-cholamidopropyl) dimethylammonio]-1-propane-sulphate/1% dithiothreitol, then added to a QiaShredder (Qiagen, Hilden, Germany) spin column and centrifuged (12,000 rpm; 3 min) to remove insoluble material. The tissue homogenates then were reacted with a hydrophobic (H50) protein chip array surface (ProteinChip; Ciphergen Biosystems, Fremont, CA). The protein chip spots were pre-equilibrated with binding buffer (50% acetonitrile/0.5% trifluoroacetic acid) at room temperature in a humidified chamber. The samples were diluted 1:5 in binding buffer (10% acetonitrile/0.1% trifluoroacetic acid). Diluted samples (5 μL) then were applied randomly to the pre-equilibrated protein chip array spots (to assess spot-to-spot and chip-to-chip reproducibility and to reduce systematic bias) and incubated in a humidified chamber for 30 minutes at room temperature. The arrays then were washed in binding buffer (3 × 5 min), rinsed twice in Milli-Q water, and air dried. Each spot then was treated with 1 μL of a 50% saturated solution of sinapinic acid in 50% acetonitrile/0.5% trifluoroacetic acid, allowed to air-dry, then the process was repeated. The arrays then were analyzed using the Ciphergen Protein Biological System IIc ProteinChip Reader (Ciphergen Biosystems). The mass spectra of proteins were generated in the mass/charge (m/z) range of 2000–50,000 by using a laser intensity of 208–218 arbitrary units. Data were averaged from 65 spectra evenly distributed across the array spot. The laser was optimized for data collection between 5000 and 30,000 m/z, the detector sensitivity was set at 7, and peaks less than 1000 m/z were deflected away from the detector. The m/z value for each of the peaks was determined using external calibration with known standards (Sigma, St Louis, MO): bovine insulin (5734.51 +1H), equine cytochrome c (12,361.96 +1H), equine apomyoglobin (16,952.27 +1H), and rabbit muscle aldolase (39,212.28 +1H). All data were analyzed using the Ciphergen ProteinChip Software version 3.1 (Ciphergen Biosystems). The raw peak intensity data for comparison of each spectrum was normalized using the total ion current between 2500 and 20,000 m/z. The detection of peaks differentially expressed between each group was performed using the Biomarker Wizard utility (version 3.1; Ciphergen Biosystems) and sample group statistics were performed for profiles of adjacent pancreatic tissue vs adenocarcinoma tissue, adjacent pancreatic tissue vs benign tissue, and adenocarcinoma vs benign tissue. Univariate analysis was performed between groups using the Mann–Whitney U test and results were considered significantly different when the P value was less than .05. For each putative marker, receiver operating characteristic (ROC) curves were generated to evaluate their discriminatory power (SPSS Software version 13.0; SPSS, Chicago, IL). Training models were developed using multivariate binary logistic regression to determine which peaks were able to best predict pancreatic adenocarcinoma from adjacent pancreatic tissue, benign pancreatic (B) tissue, and nonmalignant (NM) tissue (combination of B and AP tissue). For each comparison analyzed (ie, pancreatic adenocarcinoma [PC] vs AP, PC vs B, and PC vs NM), the models were validated using the 10-fold cross-validation approach as described by Ambroise and McLachlan.44Ambroise C. McLachlan G.J. Selection bias in gene extraction on the basis of microarray gene-expression data.Proc Natl Acad Sci U S A. 2002; 99: 6562-6566Crossref PubMed Scopus (1115) Google Scholar This repeated random sampling procedure allows for the use of all samples within the dataset to be tested and for the correction of any selection bias. Briefly, the 10-fold cross-validation approach divides the entire randomized dataset into 10 nonoverlapping datasets of roughly equal size. The model is trained on 9 of these subsets and then tested on the remaining subset to obtain prediction values. This process is repeated in turn for each of the remaining subsets, each of which leads to a different model. So, for each subset, the different model is tested on an independent sample that was not included in the development of the training model. This enabled the calculation of unbiased estimates of sensitivity and specificity, overall accuracy, and ROC area under the curve (AUC) values of the candidate tumor biomarker panels. As Ambroise and McLachlan44Ambroise C. McLachlan G.J. Selection bias in gene extraction on the basis of microarray gene-expression data.Proc Natl Acad Sci U S A. 2002; 99: 6562-6566Crossref PubMed Scopus (1115) Google Scholar stressed, in obtaining unbiased estimates it is important to avoid selection bias, namely to allow the cross-validation of the prediction rule to be external to the selection process. The effect of accommodating the selection bias is that different peaks may be selected in each training model. This is a direct consequence of the test set playing no part in the selection of the peaks. To assess whether the data within the logistic model were a good fit, the Hosmer and Lemeshow Goodness of Fit test was applied (SPSS software version 13.0). This external cross-validation approach ensures that for each subset the test samples were not included in the development of the training models. Logistic regression also was used to assess whether a panel of proteins could subclassify pancreatic adenocarcinoma successfully into early (stage I/II) vs late (stage III/IV) stage disease. Samples were analyzed in duplicate to allow for calculation of the coefficient of variation for peak intensity and mass accuracy. Replicate peak intensity values generated by SELDI-TOF MS then were averaged. Twelve peaks across the spectrum were used for the estimation of coefficient of variations. These peaks were selected randomly from all common peaks and included individual peaks that were expressed differentially by univariate analysis and peaks not significantly different between groups. The peaks included in the estimation of coefficient of variations possessed a wide distribution of peak intensities. This was performed across the repeated spectra and displayed as an overall mean. The coefficient of variation values in our study (peak intensity, 12%–20%; mass accuracy, 0.01%–0.03%) compare favorably with coefficient of variations reported in previous SELDI studies,16Kozak K.R. Amneus M.W. Pusey S.M. Su F. Luong M.N. Luong S.A. Reddy S.T. Farias-Eisner R. Identification of biomarkers for ovarian cancer using strong anion-exchange ProteinChips potential use in diagnosis and prognosis.Proc Natl Acad Sci U S A. 2003; 100: 12343-12348Crossref PubMed Scopus (240) Google Scholar, 18Petricoin E.F. Ardekani A.M. Hitt B.A. Levine P.J. Fusaro V.A. Steinberg S.M. Mills G.B. Simone C. Fishman D.A. Kohn E.C. Liotta L.A. Use of proteomic patterns in serum to identify ovarian cancer.Lancet. 2002; 359: 572-577Abstract Full Text Full Text PDF PubMed Scopus (2902) Google Scholar, 39Koopmann J. Zhang Z. White N. Rosenzweig J. Fedarko N. Jagannath S. Canto M.I. Yeo C.J. Chan D.W. Goggins M. Serum diagnosis of pancreatic adenocarcinoma using surface-enhanced laser desorption and ionization mass spectrometry.Clin Cancer Res. 2004; 10: 860-868Crossref PubMed Scopus (266) Google Scholar, 45Wadsworth J.T. Somers K.D. Cazares L.H. Malik G. Adam B.L. Stack Jr, B.C. Wright Jr, G.L. Semmes O.J. Serum protein profiles to identify head and neck cancer.Clin Cancer Res. 2004; 10: 1625-1632Crossref PubMed Scopus (106) Google Scholar, 46Li J. Zhang Z. Rosenzweig J. Wang Y.Y. Chan D.W. Proteomics and bioinformatics approaches for identification of serum biomarkers to detect breast cancer.Clin Chem. 2002; 48: 1296-1304PubMed Google Scholar and accuracy might be improved with the use of automation. Protein chip technology coupled with SELDI-TOF MS showed clear differences in protein profile expressions between invasive pancreatic adenocarcinoma, benign, and adjacent pancreatic tissue using the H50 array. An example from a segment of the protein mass profile between 9700 and 12,400 m/z is shown in Figure 1, accompanied by a spectral overlay of these markers. The H50 array revealed 13 protein peaks that were expressed differentially between invasive pancreatic adenocarcinoma vs adjacent pancreatic tissue. These individual putative tumor markers had ROC AUC values ranging from 0.64 to 0.85 (Table 2). Of these 13 protein peaks, 8 were up-regulated in the pancreatic adenocarcinoma group.Table 2Protein Peaks Expressed Differentially Between Pancreatic Adenocarcinoma Tissue Vs Adjacent Pancreatic Tissue, Benign Pancreatic Tissue, and Nonmalignant TissuePancreatic adenocarcinoma
Referência(s)