Diagnosis of Esophageal Adenocarcinoma by Serum Proteomic Pattern

GENERAL THORACIC

Diagnosis of Esophageal Adenocarcinoma by Serum Proteomic Pattern Zane T. Hammoud, MD, Lacey Dobrolecki, BS, Kenneth A. Kesler, MD, Emad Rahmani, MD, Karen Rieger, MD, Linda H. Malkas, PhD, and Robert J. Hickey, PhD Department of Surgery, Cardiothoracic Surgery Division, and Department of Medicine, Indiana University School of Medicine, Indianapolis, Indiana

Background. Currently, endoscopic biopsy is the only method used to diagnose esophageal adenocarcinoma. Using surface-enhanced laser desorption/ionization (SELDI) ProteinChip technology, we sought to identify a potentially diagnostic serum protein pattern that can serve as a reliable blood test for the diagnosis of esophageal adenocarcinoma. In addition, we sought to identify potential biomarkers for esophageal adenocarcinoma. Methods. Whole serum was collected using standard techniques from subjects with a known diagnosis of esophageal adenocarcinoma as well as from subjects without any known esophageal disease. The samples were spotted onto a hydrophobic (H50) and immobilized metal affinity (IMAC30) chip surface and allowed to incubate. All samples were run in duplicate. After several washes, matrix was added and a mass range of 1500

to 30000 daltons was analyzed by SELDI–Time-of-Flight mass spectroscopy. Statistical analysis was performed using Biomarker Pattern Software (Bio-Rad Laboratories, Hercules, CA). Results. For the H50 analysis, 3 peaks were identified that correctly diagnosed 42 of 43 cancers and 10 of 11 normals. For the IMAC30, 4 peaks were identified that correctly diagnosed 50 of 50 cancers and 10 of 10 normals. Conclusions. Serum proteomic pattern shows great promise in the diagnosis of esophageal adenocarcinoma. This technology may lead to the development of a noninvasive screening test as well as to the identification of potential novel biomarkers for esophageal adenocarcinoma. (Ann Thorac Surg 2007;84:384 –92) © 2007 by The Society of Thoracic Surgeons

C

a wide variety of proteins that may play a role in the cellular response to various alterations. Surface-enhanced laser desorption/ionization (SELDI) is an affinity-based mass spectrometric method in which proteins of interest are selectively adsorbed to a chemically modified surface on a biochip [1]. The proteins are captured, concentrated, and purified on the chemical surface of the SELDI chip. The molecular weight (mass/ charge or m/z) and relative intensity of each protein adhesed to the chip are measured with time-of-flight mass spectrometry (TOF-MS). A small quantity of serum can be used to generate a comprehensive proteomic profile that, when combined with bioinformatic analysis, may reliably distinguish between individual groups. Studies designed to “pan” for biomarkers in the sera of patients with various cancers, including ovarian, prostate, and breast, using SELDI technology have been described [2– 4]. These studies have indicated that distinctive polypeptide profiles can be obtained that appear to discriminate between patients with disease and those without disease, thereby supporting the use of this technology in cancer biomarker research. SELDI can be used to investigate a variety of protein molecular mass ranges generate profiles. Although the patterns generated can serve as a reliable means to diagnose disease, the individual peaks may point to novel biomarkers that may serve as a reliable means to screen for disease.

urrently, there are no reliable genetic biomarkers of esophageal adenocarcinoma. Efforts to use a variety of genetic alterations as surrogate markers of disease have not been successful. In theory, DNA, RNA, and proteins can be used as potential biomarkers; however, alterations in DNA or RNA, or both, may not lead to alterations in protein expression. Individual protein expression may thus be an attractive means by which to identify potential biomarkers of disease. Two-dimensional polyacrylamide gel electrophoresis has been the principal tool used for the separation and simultaneous analysis of multiple proteins. Although the resolving power of this methodology remains unchallenged, the high sensitivity, speed, and reproducibility of approaches based on mass spectrometry have led to the increased use of such approaches in all aspects of protein analysis, including profiling, discovery, and identification, creating what is now called proteomics. Serum protein pattern analysis, or proteomics, allows for the possible identification of Accepted for publication March 26, 2007. Presented at the Forty-third Annual Meeting of The Society of Thoracic Surgeon, San Diego, CA, Jan 29 –31, 2007. Address correspondence to Dr Hammoud, Cardiothoracic Surgery, Indiana University School of Medicine, 545 Barnhill Dr, EH 215, Indianapolis, IN 46202; e-mail: [email protected].

© 2007 by The Society of Thoracic Surgeons Published by Elsevier Inc

0003-4975/07/$32.00 doi:10.1016/j.athoracsur.2007.03.088

We performed SELDI analysis of serum from patients with esophageal adenocarcinoma in an attempt to identify a signature pattern that can accurately identify the disease. We also attempted to identify protein peaks that may serve as potential novel biomarkers for esophageal adenocarcinoma.

Material and Methods The study was approved by the Indiana University Institutional Review Board. All subjects included in the study provided informed consent according to institutional guidelines. Standard laboratory procedures were used to draw blood samples were obtained from 50 patients with biopsy-proven esophageal adenocarcinoma who had not received any therapy. All patients were staged clinically by endoscopic ultrasound (n ⫽ 53) or a combination of computed tomography (CT) and positron emission tomography, or both. As a control, blood samples were obtained from 11 individuals (normals) who had no known esophageal disease or medical condition. The blood was allowed to clot for 45 minutes and was then centrifuged at 2000 rpm for 10 minutes. The serum was aspirated, placed in a separate vial, and stored at ⫺80°C. On the day of analysis, the serum was placed on ice and thawed completely. Each sample was centrifuged at 13,200 rpm for 10 minutes at 4°C. Any samples that were hemolyzed or greatly lipemic were not used for analysis. For each study, all samples had been through the same number of freeze-thaw cycles. Each sample was randomly applied in duplicate to the appropriate chip type. Two types of protein chips were used, hydrophobic (H50 ProteinChip) and immobilized metal affinity (IMAC30 ProteinChip) both from Bio-Rad Labs, Hercules, California. Both chip types were processed on a Biomek 2000 (Beckman Coulter, Fullerton, CA) liquid-handling robot by using a ProteinChip bioprocessor (Bio-Rad Laboratories).

IMAC30 ProteinChip Processing Samples that were run on IMAC30 chips were pretreated with urea. The centrifuged serum (20 ␮L) was mixed with 30 ␮L U9 buffer (9 M urea, 2% 3[(3-cholamidopropyl) dimethylammonio]-propane sulfonte, 50 mM Trishydrochloride, pH 9.0) and placed on ice for 20 minutes. After the urea incubation, the samples were centrifuged again at 10,000 rpm for 5 minutes. The IMAC30 chips were placed in a ProteinChip bioprocessor and 50 ␮L of 0.1M copper sulfate was incubated on the chips for 5 minutes with vigorous shaking. The chips were then rinsed with deionized water and neutralized for 5 minutes with 50 ␮L of 0.1M sodium acetate, pH 4.0. After another rinse with deionized water, the chips were equilibrated by washing twice with 150 ␮L of binding buffer (1X phosphate-buffered saline and 0.3M sodium chloride, pH 7.4). Next, the chips were loaded with 90 ␮L of binding buffer to which 10 ␮L of the urea processed serum was

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

385

added. The serum was incubated with the chips for 1 hour with vigorous shaking. The serum was removed after an hour, and the chips were washed three times with 150 ␮L of binding buffer. A final rinse with deionized water was performed, the bioprocessor was removed, and the chips were allowed to air dry. Once the chips were dry, 0.7 ␮L of a 50% solution of ␣-cyano-5-hydroxycinnamic acid (CHCA) diluted in 50% (v/v) acetonitrile and 0.5% trifluoroacetic acid (v/v) was applied to each spot. The CHCA solution was allowed to dry completely before it was added for a second time.

H50 ProteinChip Processing The H50 chips were placed in a ProteinChip bioprocessor and pre-rinsed twice with 50 ␮L of 50% acetonitrile. The chips were then equilibrated by washing twice with 150 ␮L of binding buffer (10% acetonitrile, 0.1% trifluoroacetic acid). Next, the chips were loaded with 90 ␮L of binding buffer to which 10 ␮L of serum was added. The serum was incubated with the chips for 1 hour with vigorous shaking. After an hour, the serum was removed and the chips were washed three times with 150 ␮L of binding buffer. A final rinse with deionized water was performed, the bioprocessor was removed, and the chips were allowed to air dry. Once the chips were dry, 0.7 ␮L of a 50% solution of CHCA diluted in 50% (v/v) acetonitrile and 0.5% trifluoroacetic acid (v/v) was applied to each spot. The CHCA solution was allowed to dry completely before it was added for a second time.

Surface-Enhanced Laser Desorption/Ionization Time-of-Flight Mass Spectrometric Analysis The chips were analyzed in a Protein Biological System IIc TOF MS (PBS-IIc, Bio-Rad Laboratories, Hercules, CA). The mass spectra were obtained using the following parameters: 160 laser shots/spectra collected in the positive mode; detector sensitivity of 9; and a detector

Table 1. Clinical Tumor-Node-Metastasis Stage of 50 Patients With Esophageal Adenocarcinoma Stage

No. of Patients

T1 N0 T2 N0 T2 N1 T2 Nx T3 N0 T3 N1 T3 Nx T4 N1 M1a M1b Unknown (endoscopy, no EUS) PET only (no distant disease) EUS ⫽ endoscopic ultrasound; M ⫽ metastasis; PET ⫽ positron emission tomography; T ⫽ tumor.

7 5 2 1 6 14 1 1 4 5 2 2 N ⫽ node;

GENERAL THORACIC

Ann Thorac Surg 2007;84:384 –92

GENERAL THORACIC

386

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

Ann Thorac Surg 2007;84:384 –92

Fig 1. Raw spectral data identifies 3 peaks using the H50 ProteinChip (Bio-Rad Laboratories, Hercules, CA). The top 2 specimens in each row are cancer, and the bottom 2 are normal.

voltage of 2950V. The laser intensity was 140 for the H50 analysis and 160 for the IMAC30 analysis. The PBS-IIc was externally calibrated using the All-in-One Peptide Standard mass standard (Bio-Rad Laboratories).

Data Processing The spectra from both experiments were processed using the CiphergenExpress Software package (BioRad Laboratories). The baseline subtraction was optimized, and the spectra were normalized by total ion current. Any spectrum that had a normalization factor greater than two standard deviations from the mean

was discarded from the analysis. External mass calibration was performed using a mixture of known protein standards that correspond to the molecular weight that was measured. A calibration equation was created from these standards and was applied to each spectrum. Finally, peaks were detected using a signal/ noise cutoff of two. Decision tree analyses were performed using Biomarker Patterns Software (BPS, Bio-Rad Laboratories). This software program has the ability to combine multiple biomarkers to distinguish between independent groups, thereby increasing sensitivity and specificity compared

Fig 2. Cluster plots generated using H50 ProteinChip (Bio-Rad Laboratories, Hercules, CA) for the 3 peaks of interest.

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

387

Fig 3. Decision tree analysis for H50 Protein Chip (Bio-Rad Laboratories, Hercules, CA) using the 3317, 4393, and 8632 peaks to identify cancer and normal specimens. The analysis correctly identified 42 of 43 cancers and 10 of 11 normals. The number after the peaks (eg, 2.827 for 8632) represents the intensity of the peak. Class 0 ⫽ cancer; Class 1 ⫽ normal.

with single biomarker predictors. BPS was also used to perform a 10-fold cross-validation because the size of the data set was too small for an independent validation set. This process allows the sensitivity and specificity to be predicted for future data and provides a more accurate estimate for the predictive accuracy of the selected decision tree.

Results There were 44 men and 6 women with a mean age at diagnosis of 64.5 years (range, 39 to 91 years). Table 1 lists the clinical stage of the 50 patients with esophageal adenocarcinoma. For the H50 analysis, 11 normal sera and 43 adenocarcinoma sera were analyzed. Figure 1 shows the spectral data obtained using the H50 ProteinChip, identifying 3 peaks (8632, 4393, and 3317) that appear to discriminate between the normal samples and those of the adenocarcinoma samples. The pattern generated by the cancer specimens is similar to other cancer specimens, whereas the pattern for normals is similar to

other normals. Figure 2 represents the same data in cluster format. Figure 3 summarizes the decision tree analysis for the H50 chip using the 3 peaks identified. The peaks and the relative intensity distinguished cancer from normal. The 8632 peak, at an intensity of 2.827 or less, correctly identified 10 of 11 normals and 33 of 43 cancers. The remaining 10 cancer samples were analyzed with the 4393 peak, at an intensity of 0.492 or less, and seven additional cancers were correctly identified. Finally, the remaining three cancer samples were analyzed with the 3317 peak at an intensity of 5.528 or less. Two of the three remaining cancer samples were correctly identified. By performing this analysis, 10 of 11 normal samples were identified as normal and 42 of 43 cancer samples were correctly identified. For the IMAC30 analysis, 10 normal and 50 adenocarcinoma sera were analyzed. The raw spectral data (Fig 4) identified 4 peaks (2743, 2861, 3217, 3421) that appear to discriminate normal from cancer. These peaks also showed relative homogeneity among the cancer and

GENERAL THORACIC

Ann Thorac Surg 2007;84:384 –92

GENERAL THORACIC

388

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

Ann Thorac Surg 2007;84:384 –92

Fig 4. Raw spectral data generated with IMAC30 ProteinChip (Bio-Rad Laboratories, Hercules, CA), showing 4 peaks (2743, 2861, 3217, 3421) that distinguish cancer from normal. The top 2 spectra in each graph are cancer, and the bottom 2 are normal.

normal specimens, respectively. Figure 5 demonstrates the cluster data for all 4 peaks. Decision tree analysis using the peaks at 3217 and 3421 correctly identified all 10 normals and all 50 cancers (Fig 6). Using the peak at 3217 and an intensity of 6.11 or less, all normals were correctly identified and 43 of 50 cancers were correctly identified. Analyzing the remaining seven cancer samples using the 3421 peak at an intensity of 0.157 led to the correct identification of all of these specimens. Decision tree analysis using the peaks at 3217 and 2743 correctly identified 9 of 10 normal samples and all 50 cancers (Fig 7). Analysis using the peak at 2861 alone correctly identified 9 of 10 normal samples and 47 of 50 cancers (Fig 8). A separate analysis of the 12 early stage tumors (T1 N0 and T2 N0) was performed and revealed two peaks of interest at 2861 and 6434. Figure 9 shows the decision tree analysis for these tumors at the 2 peaks. Using this analysis, all 10 normal samples were correctly identified, and 11 of 12 cancers were correctly identified.

Comment The prognosis of patients with esophageal adenocarcinoma remains poor. As with other types of cancer, early

detection offers the possibility of significant improvement in long-term survival. Currently, esophageal endoscopy with biopsy is the standard modality used to establish the diagnosis of esophageal adenocarcinoma. Endoscopy is also the only current method by which to screen patients at risk of developing esophageal adenocarcinoma (eg, patients with Barrett esophagus). However, endoscopy is invasive, expensive, and usually performed only when patients present with symptoms, which is an indication of advanced disease; therefore, alternate methods of diagnosis are desired. Once discovered, these methods may then prove effective in the routine screening of patients at risk for developing esophageal adenocarcinoma. Proteomic pattern analysis, or proteomics, has generated recent excitement as a diagnostic tool because it can, in theory, lead to a rapid and accurate blood test for the detection of cancer. This approach relies on mass spectrometry to generate a pattern of signals and to identify differentially abundant peaks within disease and normal samples, thus distinguishing between the two. The use of serum makes this approach particularly attractive, for access to patient serum is easy and inexpensive. Because pathologic and cancerous changes can be reflected as protein changes in serum, the detection of

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

389

Fig 5. Cluster plots of the IMAC30 ProteinChip (Bio-Rad Laboratories, Hercules, CA) at the 4 peaks identified.

such changes may lead to earlier diagnosis and, potentially, to the discovery of novel biomarkers for the detection of a given pathologic or cancerous condition in patients who are at risk for the development of that condition. The unique mass/charge value of each protein enables unambiguous detection and quantification in the patient’s sera, even without knowing the exact identity of the individual protein. Proteomics has been used to generate signature patterns that can accurately and reliably distinguish cancer from noncancer. In their landmark study, Petricoin and colleagues [3] used SELDI and an artificial intelligence algorithm to identify a mass spectrometric pattern that distinguished ovarian cancer from normal with 100% sensitivity and 95% specificity. This study demonstrated the rapidity and ease of the technique for the analysis of serum proteins. Other reports have also documented the efficacy of proteomics. Petricoin and colleagues [4] also identified a proteomic pattern that correctly predicted 36 of 38 patients with prostate cancer and 177 of 228 patients with benign conditions. Honda and colleagues [5] used SELDI to identify 4 peaks that accurately discriminated 69 of 71 patients with pancreatic cancer and 67 of 71 patients without disease. Other reports have further used this technology in the accurate diagnosis of colorectal adenocarcinoma, gastric cancer, lung cancer, and breast cancer [6 –9]. We used SELDI-TOF MS to analyze serum from patients with esophageal adenocarcinoma and have shown that this method accurately diagnosed patients with disease. A signature pattern for patients with esophageal adenocarcinoma was identified and protein peaks were generated. These peaks may serve as potential biomarkers of disease, although the identities

of the individual proteins have yet to be elucidated. We are currently addressing this issue by isolating the peaks and identifying the proteins represented by the peaks. However, the pattern itself may be useful as a diagnostic tool independent of the protein identities. It may be possible to use these patterns to diagnose patients with esophageal adenocarcinoma at an early stage. It may also be possible to monitor serial patterns to monitor disease (eg, response to chemotherapy or recurrence after treatment). Validation of our preliminary data by analysis of many more sera from patients with esophageal adenocarcinoma will be required to explore the potential clinical utility of this technology. Analysis of cancer specimens at various individual stage of disease will be required to determine the potential utility of SELDI serum analysis in the diagnosis of early stage disease. We also plan to perform SELDI analysis on sera from patients with Barrett esophagus as well as Barrett esophagus with high-grade dysplasia in an attempt to identify signature patterns or biomarkers, or both, that may be used to distinguish these disease states from invasive adenocarcinoma. Although the Barrett esophagus metaplasia-dysplasia-carcinoma sequence in esophageal adenocarcinoma is well established, currently the only method of identifying disease progression is by pathologic examination. Furthermore, there is controversy about the optimal management of patients with highgrade dysplasia, for not all patients with high-grade dysplasia will progress to invasive carcinoma. Using proteomic patterns may lead to the identification of patterns that may risk stratify these patients. The patterns may also lead to the discovery of novel biomarkers that can serve as markers of disease.

GENERAL THORACIC

Ann Thorac Surg 2007;84:384 –92

GENERAL THORACIC

390

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

Ann Thorac Surg 2007;84:384 –92

A further limitation of our study is the lack of adequate controls. Ideally, a control population of subjects matched for age and sex is used when one group is compared with another. Our experience, however, has shown it to be extremely difficult to identify such a control population for this study. The mean age of the adenocarcinoma subjects is older than 60 years. The prevalence of various comorbid conditions in patients in this age group makes it nearly impossible to identify a large number of these patients in whom the presence of such comorbidities would not lead to potentially confounding serum patterns. As shown in Figures 1 and 4, there is a high degree of consistency in the patterns observed in our control population, thereby leading us to conclude that this population was suitable for analysis. We also used the sera from patients with various stages of disease (Table 1). Again, however, we observed a consistency in the patterns generated regardless of stage. Although analysis of each individual stage of disease would be desirable, the small number of patients in each stage makes such an analysis statistically suspect. A more fundamental objection to the mass spectrometry method is that it cannot detect the necessary discriminatory proteins, for these cancer markers may only be present in low amounts in the blood (serum). For

Fig 6. Decision tree analysis for IMAC 30 Protein Chip (Bio-Rad Laboratories, Hercules, CA) at peaks 3217 and 3421. The analysis correctly identified 50 of 50 cancers and 10 of 10 normals. Class 0 ⫽ cancer; Class 1 ⫽ normal.

Our study has several limitations. Most of these relate to the technology applied—mass spectrometry—and the resultant proteomic patterns generated. The peaks generated and the proteins they represent have not been identified. One issue is that the patterns may be difficult to reproduce by different investigators. However, one must ensure that the same methods are used to generate the patterns. We also do not know what effect, if any, the number of freeze-thaw cycles will have on the proteins contained in any given serum. It is also unknown if the fluctuations in an individual patient’s protein pattern, such as those that may occur with an upper respiratory infection or even the consumption of a given meal, may upset the diagnostic accuracy of the analysis of the serum. Such limitations of serum analysis have led others to perform proteomic analysis on plasma. Our validation studies will address all of these issues and should lead to a reproducible method by which to consistently generate the patterns. These validation studies will also address the issue of the potential effect of cancer stage on the patterns generated; that is, differences in pattern or peaks, or both, between early and later stage esophageal adenocarcinoma.

Fig 7. Decision tree analysis using the IMAC30 ProteinChip (BioRad Laboratories, Hercules, CA) and the peaks at 3217 and 2743. This analysis correctly identified 50 of 50 cancers and 9 of 10 normals. Class 0 ⫽ cancer; Class 1 ⫽ normal.

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

391

instance, prostate-specific antigen, a well-established biomarker of prostate cancer, has yet to be identified in any of the studies that have generated proteomic patterns that can accurately diagnose prostate cancer. This may be because the abundance of prostate-specific antigen in serum is below the level of the ability of mass spectrometry to detect the molecule. This is a legitimate concern, but the patterns may nonetheless identify other proteins that, when combined with other known markers, may prove clinically useful or that may lead to the discovery of low abundance molecules by pointing to a protein that is of interest. Such a strategy has been shown to be successful in pancreatic cancer and in colon cancer [10, 11]. An additional potential limitation is that the peaks discovered may not represent proteins but may simply represent artifacts. These artifacts may be due to sample preparation, contamination, or to proteins such as digested albumin that may be irrelevant to tumor. Such limitations will only be overcome through validation trials that will generate patterns in a blinded fashion and that will test cancer against a host of other conditions other than normal; for example, inflammation and infection. In summary, we used SELDI-TOF MS to generate proteomic patterns from sera of patients with esopha-

Fig 9. Decision tree analysis using IMAC 30 Protein Chip (Bio-Rad Laboratories, Hercules, CA) for 12 early stage tumors (T1 N0 and T2 N0) at peaks 2861 and 6434. This analysis correctly identified 11 of 12 cancers and 10 of 10 normals. Class 0 ⫽ cancer; Class 1 ⫽ normal.

geal adenocarcinoma. These patterns, combined with statistical analysis, led to the discovery of certain peaks that may represent a unique pattern that can distinguish esophageal adenocarcinoma from normal. These preliminary data hold promise for the clinical application of this technique for the diagnosis of esophageal adenocarcinoma from patient serum. In addition, the technology may lead to future identification of novel biomarkers for esophageal adenocarcinoma. Such biomarkers may prove useful in screening patients at risk of this disease as well as serve as indicators of disease prognosis.

References

Fig 8. Decision tree analysis using the IMAC30 ProteinChip (BioRad Laboratories, Hercules, CA) and the peak at 2861. This analysis correctly identified 47 of 50 cancers and 9 of 10 normals. Class 0 ⫽ cancer; Class 1 ⫽ normal.

1. Zhang LJ, Rozenzweig Z, Wang YY, Chan DW. Proteomics and bioinformatics approached for identification of serum biomarkers to detecting breast cancer. Clin Chem J 2002;48:1296 –304. 2. Adam B, Qu Y, Davis JW, et al. Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men. Can Res 2002;62:3609 –14. 3. Petricoin EF, Ardekani AM, Hitt BA, et al. Use of proteomic patterns in serum to identify ovarian cancer. Lancet 2002;359:572–7. 4. Petricoin EF, Ornstein DK, Paweletz CP, et al. Serum proteomic patterns for the detection of prostate cancer. J Natl Cancer Inst 2002;94:1576 – 8. 5. Honda K, Hayashida Y, Umaki T, et al. Possible detection of pancreatic cancer by plasma protein profiling. Cancer Res 2005;65:10613–22.

GENERAL THORACIC

Ann Thorac Surg 2007;84:384 –92

GENERAL THORACIC

392

HAMMOUD ET AL PROTEOMICS IN ESOPHAGEAL ADENOCARCINOMA

6. Liu X, Shen J, Li Z, Yan L, Gu J. A serum proteomic pattern for the detection of colorectal adenocarcinoma using surface enhanced laser desorption and ionization mass spectrometry. Cancer Inves 2006;24:747–53. 7. Poon TCW, Sung JJY, Chow SM, et al. Diagnosis of gastric cancer by serum proteomic fingerprinting. Gastroenterology 2006;130;1858 – 64. 8. Yang S, Xiao X, Zhang W, et al. Application of serum SELDI proteomic patterns in diagnosis of lung cancer. BMC Cancer 2005;5:1–7.

Ann Thorac Surg 2007;84:384 –92

9. Hudelist G, Singer CF, Pischinger KID, et al. Proteomic analysis in human breast cancer: identification of a characteristic protein expression profile of malignant breast epithelium. Proteomics 2006;6:1989 –2002. 10. Chen R, Pan S, Brentall TA, Aebersold R. Proteomic profiling of pancreatic cancer for biomarker discovery. Mol Cell Proteomics 2005;4:523–33. 11. Ward DG, Suggett N, Cheng Y, et al. Identification of serum biomarkers for colon cancer by proteomic analysis. Br J Cancer 2006;94:1898 –905.

DISCUSSION DR MARK J. KRASNA (Baltimore, MD): Just a very quick comment and question. My colleague, Steve Meltzer, who has recently moved to Hopkins, and I had for many years done both IHC and then genomic studies looking at loss of heterozygosity. One of the things that stuck out very obviously is that when we identified potential genes to look at as markers, it was not only looking at the cancers, but in most of these cases, the adjacent tissue which may have Barrett’s was also involved. I wonder within the current study, did you in fact have any pathologic analysis of those cancers, and of those cancers that had Barrett’s as well as adenocarcinoma, could you actually see the signature in those patients? Was there a difference in intensity or a difference in representation? Obviously it’s very intriguing and a lot easier than getting it from tissue biopsies like we had to do in genomics. DR HAMMOUD: Thank you, Dr Krasna. The short answer is no, we have not done a retrospective look at the types of cancers in terms of whether or not they included Barrett’s. In the manuscript, we did go back and identify the stage, T versus N and T and N, of the specimens that we looked at. Like most esophageal cancers, it is not surprising to know that a lot of these are T3 N0 or T3 N1. I’d like to emphasize that this is preliminary data, and we really need many more specimens. Although the data looks great after statistically looking at it and kind of plotting it out in this commercially-available program, I think as the data matures and as we analyze more specimens, it probably won’t look as good. The key thing here is going to be whether or not these proteins that are identified or represented by these peaks represent novel proteins. SELDI itself as a technique is currently the subject of much criticism, and some people have discounted it in the field of proteomics. But I think that my biggest take on this is that the pattern is far more important than what we’re actually looking at, because if we can identify a pattern from basically drawing serum from a patient’s blood and saying, “Well, you’ve got esophageal cancer,” I think that would be extremely valuable. DR JOSEPH LOCICERO (Brooklyn, NY): Just to expand a little bit on what you’ve already said, can you give us any more information on your demographics? Were these normal controls, age-matched, or matched in any way? What sort of a spread do you have on the cancers? How many patients had metastatic disease? Did you have early cancers? Just a little more information that will help us out here.

DR HAMMOUD: With respect to the normals, they were not age-matched. They were simply every normal that I could grasp because of the strict criteria that we had for “normal.” So it would have been difficult to age-match that population. In the manuscript, I go into a little bit more detail, which hopefully once it is published you will be able to read, in terms of the average age of the patients and the staging. We did have some early cancers within them, but the vast majority were T3 N0 or T3 N1. DR JONATHAN E. RHOADS (York, PA): I just wondered if you had proteomic studies on tumors in addition to esophageal and whether or not you find the same tumors in cancers arising from various different tissues or if these are tissue-specific. DR HAMMOUD: I think that the question is whether or not one can obtain signature patterns of other cancers? DR RHOADS: Yes. DR HAMMOUD: Yes, and that is what has been published for prostate, for breast, as well as for ovarian cancers. People have reported, quote/unquote, “signature patterns” that identify cancer reliably. DR JOE B. PUTNAM, JR (Nashville, TN): This information is pretty exciting, and I am sure that the next question is of interest to us all: Have you evaluated the protein patterns in patients preresection and postresection to see if the protein patterns indeed change? DR HAMMOUD: We’re doing it. DR PUTNAM: You’ve not snuck one in sort of like under the radar screen or something? DR HAMMOUD: No. The numbers are too small. Since most of these patients are receiving neoadjuvant therapy, we are currently trying to collect specimens for analysis at diagnosis, after completion of therapy, and after resection. The aim is to try to identify patterns that will be predictive of response to therapy, especially complete pathologic response. DR PUTNAM: Thank you very much. We look forward to hearing more about this in the future.