摘要
Background: Proteomics-based approaches complement the genome initiatives and may be the next step in attempts to understand the biology of cancer. We used matrix-assisted laser desorption/ionisation mass spectrometry directly from 1-mm regions of single frozen tissue sections for profiling of protein expression from surgically resected tissues to classify lung tumours. Methods: Proteomic spectra were obtained and aligned from 79 lung tumours and 14 normal lung tissues. We built a class-prediction model with the proteomic patterns in a training cohort of 42 lung tumours and eight normal lung samples, and assessed their statistical significance. We then applied this model to a blinded test cohort, including 37 lung tumours and six normal lung samples, to estimate the misclassification rate. Findings: We obtained more than 1600 protein peaks from histologically selected 1 mm diameter regions of single frozen sections from each tissue. Class-prediction models based on differentially expressed peaks enabled us to perfectly classify lung cancer histologies, distinguish primary tumours from metastases to the lung from other sites, and classify nodal involvement with 85% accuracy in the training cohort. This model nearly perfectly classified samples in the independent blinded test cohort. We also obtained a proteomic pattern comprised of 15 distinct mass spectrometry peaks that distinguished between patients with resected non-small-cell lung cancer who had poor prognosis (median survival 6 months, n=25) and those who had good prognosis (median survival 33 months, n=41, p<0.0001). Interpretation: Proteomic patterns obtained directly from small amounts of fresh frozen lung-tumour tissue could be used to accurately classify and predict histological groups as well as nodal involvement and survival in resected non-small-cell lung cancer.
| 原文 | English |
|---|---|
| 頁（從 - 到） | 433-439 |
| 頁數 | 7 |
| 期刊 | Lancet |
| 卷 | 362 |
| 發行號 | 9382 |
| DOIs | |
| 出版狀態 | Published - 2003 8月 9 |
All Science Journal Classification (ASJC) codes
- 一般醫學
引用此
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver
}
於: Lancet, 卷 362, 編號 9382, 09.08.2003, p. 433-439.
研究成果: Article › 同行評審
TY - JOUR
T1 - Proteomic patterns of tumour subsets in non-small-cell lung cancer
AU - Yanagisawa, Kiyoshi
AU - Shyr, Yu
AU - Xu, Baogang J.
AU - Massion, Pierre P.
AU - Larsen, Paul H.
AU - White, Bill C.
AU - Roberts, John R.
AU - Edgerton, Mary
AU - Gonzalez, Adriana
AU - Nadaf, Sorena
AU - Moore, Jason H.
AU - Caprioli, Richard M.
AU - Carbone, David P.
N1 - Funding Information: In this study, we directly profiled protein expression from tumour tissue with MALDI-TOF MS, and defined profiles that enabled classification of surgically resected lung tumours into biologically meaningful groups. Previous studies have defined subclasses and prognostic subsets of lung adenocarcinomas based on gene expression patterns. 3–5 However, mRNA expression cannot always indicate which proteins are expressed or how their activity might be modulated after translation. 6,7 Accordingly, analysis of the proteome in serum, tumour tissues, and other clinical materials might better reflect the underlying pathological state of cancers than gene expression patterns. A few reports have shown that proteomic patterns obtained from serum with surface enhanced laser description/ionisation-TOF can distinguish cancer patients from non-cancer patients, 23–26 but these studies have been limited to cancer detection and were unable to further classify tumours or predict tumour behaviour. Mass spectrometry profiles obtained from complex protein mixtures can contain thousands of data points, derived from real protein signatures but contaminated with electronic and chemical noise, variability in instrumentation, and variable crystallisation of matrix, necessitating careful analytical techniques. 25 Accordingly, we developed artificial intelligence software to align and bin thousands of peaks from hundreds of protein profiles and used advanced statistical methods combining the most significant proteins associated with the biological variable of interest from several analytical methods. We also used hierarchical clustering to examine similarities among lung cancers in their patterns of protein expression. These technologies enabled us to detect proteomic patterns that could be used to accurately classify surgically resected primary NSCLC, according to classical clinicopathological variables, from very small amounts of fresh frozen tissue. Our class-prediction models were found to classify lung tumour versus normal lung (with 82 MS signals) with 100% accuracy in the training and the blinded cohort. Similarly, primary NSCLC could be distinguished from normal lung (91 MS signals), and primary NSCLC from cancer metastatic to the lung (23 MS signals), adenocarcinoma from squamous-cell carcinoma (20 MS signals), and squamous-cell from large-cell carcinoma (12 MS signals) with 100% accuracy in both sample sets. Our profile distinguishing adenocarcinoma from large-cell carcinoma (20 MS signals) had one apparent misclassification of a large-cell carcinoma as an adenocarcinoma in the blinded cohort; however, it is possible that this tumour was actually an adenocarcinoma that was too poorly differentiated to be identified as such by light microscopy. Some statistical limitations need to be addressed. First, since the study sample size was small, a larger scale study to confirm our findings is necessary. Second, the number of peaks reported in this paper was not based on the smallest number of peaks that could discriminate the classes but based on the statistical evidence. The possibility of achieving similar misclassification rates based on different subsets of peaks does exist. Additionally, the near perfect discrimination obtained with the agglomerative hierarchical clustering is not surprising, since it uses covariates that were themselves chosen to have maximal discriminating power. Future studies will undoubtedly refine these models. Identities of most of the proteins that make up these profiles are largely unknown at present; however, we identified three specific tumour markers, including an activated form of SUMO-2. SUMO represents a class of ubiquitin-like proteins that is conjugated, like ubiquitin, by a set of enzymes to cellular regulatory proteins, including oncogenes and tumour suppressor genes, and that may have key roles in the control of cell growth, differentiation, apoptosis, cell cycle, DNA repair, stress response, and nuclear transport. 27–29 SUMO conjugation affects the substrates' subcellular localisation and stability as well as transcriptional activities. In view of the substrates involved, SUMO conjugation might be of previously unsuspected importance in lung tumorigenesis. Although evidence to support this notion is still scarce, the role of SUMO in the interference of the ubiquitinylation of MDM2 and prevention of subsequent degradation, which leads to inactivation of p53 and deregulation of the cell-cycle, 30 suggests its potential involvement in NSCLC. Our identification in these studies of another protein, thymosin- 4, is in accord with previous findings that showed increased expression of this protein in proliferating glioblastoma and neoplastic lesions compared with benign tissue. 11,31,32 This protein is able to sequester cytoplasmic monomeric actin, and its expression in tumour cells is correlated with tumorigenicity and metastatic potential in malignant fibrosarcoma cell lines. 33 There is a discrepancy between the expression level of this protein in our study and that of mRNA previously reported and analysed by microarray. 3 Thus, the proteins identified by these studies might be useful as novel biomarkers for detection and classification of lung cancers and lead us to a better understanding of lung cancer biology than might be achieved by RNA analysis alone. Further investigations are warranted to identify the discriminating proteins that define our subsets and acsertain their functional significance in NSCLC. Proteomic patterns predictive of features accurately identified by light microscopy, while potentially useful to identify novel biological targets, have little clinical use. However, in this study we also obtained a proteomic pattern from the initial resected primary tumour comprised of 15 distinct MS peaks that divided these NSCLC patients into groups with poor prognosis and good prognosis. If this pattern is confirmed in larger studies, its prognostic power could exceed that of almost any previously published standard molecular marker. Perhaps more strikingly, we selected a proteomic pattern that might be associated with nodal involvement at the time of surgical lymph node dissection. Although there were several apparently misclassified samples, the relative insensitivity of standard histological assessment of lymph nodes can lead to an overestimation of false positives. The accuracy of this pattern is, in fact, comparable to standard staging techniques, and better than any other molecular markers identified to date. Since nodal involvement is one of the most important factors in determining therapeutic strategies, if this accuracy could be confirmed and improved through analysis of a larger cohort, the potential clinical usefulness of this profile could be great. Thus, our data show that using nanogram amounts of fresh frozen tumour readily collectable in a clinical setting, protein profiles obtained from unprocessed tissue samples can be used to accurately classify tumours. The obtained profiles might also be used to allocate patients into good and poor prognostic groups, and predict risk for nodal involvement. Since such small tissue samples can be used, it would be of great interest to analyse protein expression patterns of tissue samples from needle aspirates, or from the different cell subtypes within the lung, or attempt to derive patterns associated with response to specific treatments, with smoking exposure or preneoplasia, and with the risk of progression to cancer. If these data are confirmed in larger numbers of patients, this technology could have great implications for the clinical management of patients with NSCLC. Contributors K Yanagisawa, R Caprioli, and D Carbone conceived the study, participated in modelling procedures and analysis, and wrote the report. PLarsen and Y Shyr developed the statistical methods and did the data analysis. K Yanagisawa, B Xu, and R Caprioli participated in generation, analysis, and presentation of the mass spectrometry data. B White and JMoore conceived and developed the software for alignment of mass spectrometry data. P Massion, J Roberts, A Gonzalez, and S Nadaf provided or reviewed the tissue samples in the study, and assisted in the preparation of the paper. M Edgerton developed the biomedical informatics systems. Conflict of interest statement None declared. Acknowledgments We thank Ben Garcia, Darienne Adkins, and Pierre Chaurand at the Vanderbilt University Medical Center for helpful support and thoughtful suggestions. Our research was funded by Lung Cancer Special Program of Research Excellence grants 1P50CA90949, P50CA70907, GM 58008, and CA 86243, the Mathers Foundation, and the Robert A and Helen CKleberg Foundation.
PY - 2003/8/9
Y1 - 2003/8/9
N2 - Background: Proteomics-based approaches complement the genome initiatives and may be the next step in attempts to understand the biology of cancer. We used matrix-assisted laser desorption/ionisation mass spectrometry directly from 1-mm regions of single frozen tissue sections for profiling of protein expression from surgically resected tissues to classify lung tumours. Methods: Proteomic spectra were obtained and aligned from 79 lung tumours and 14 normal lung tissues. We built a class-prediction model with the proteomic patterns in a training cohort of 42 lung tumours and eight normal lung samples, and assessed their statistical significance. We then applied this model to a blinded test cohort, including 37 lung tumours and six normal lung samples, to estimate the misclassification rate. Findings: We obtained more than 1600 protein peaks from histologically selected 1 mm diameter regions of single frozen sections from each tissue. Class-prediction models based on differentially expressed peaks enabled us to perfectly classify lung cancer histologies, distinguish primary tumours from metastases to the lung from other sites, and classify nodal involvement with 85% accuracy in the training cohort. This model nearly perfectly classified samples in the independent blinded test cohort. We also obtained a proteomic pattern comprised of 15 distinct mass spectrometry peaks that distinguished between patients with resected non-small-cell lung cancer who had poor prognosis (median survival 6 months, n=25) and those who had good prognosis (median survival 33 months, n=41, p<0.0001). Interpretation: Proteomic patterns obtained directly from small amounts of fresh frozen lung-tumour tissue could be used to accurately classify and predict histological groups as well as nodal involvement and survival in resected non-small-cell lung cancer.
AB - Background: Proteomics-based approaches complement the genome initiatives and may be the next step in attempts to understand the biology of cancer. We used matrix-assisted laser desorption/ionisation mass spectrometry directly from 1-mm regions of single frozen tissue sections for profiling of protein expression from surgically resected tissues to classify lung tumours. Methods: Proteomic spectra were obtained and aligned from 79 lung tumours and 14 normal lung tissues. We built a class-prediction model with the proteomic patterns in a training cohort of 42 lung tumours and eight normal lung samples, and assessed their statistical significance. We then applied this model to a blinded test cohort, including 37 lung tumours and six normal lung samples, to estimate the misclassification rate. Findings: We obtained more than 1600 protein peaks from histologically selected 1 mm diameter regions of single frozen sections from each tissue. Class-prediction models based on differentially expressed peaks enabled us to perfectly classify lung cancer histologies, distinguish primary tumours from metastases to the lung from other sites, and classify nodal involvement with 85% accuracy in the training cohort. This model nearly perfectly classified samples in the independent blinded test cohort. We also obtained a proteomic pattern comprised of 15 distinct mass spectrometry peaks that distinguished between patients with resected non-small-cell lung cancer who had poor prognosis (median survival 6 months, n=25) and those who had good prognosis (median survival 33 months, n=41, p<0.0001). Interpretation: Proteomic patterns obtained directly from small amounts of fresh frozen lung-tumour tissue could be used to accurately classify and predict histological groups as well as nodal involvement and survival in resected non-small-cell lung cancer.
UR - http://www.scopus.com/inward/record.url?scp=0041735992&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0041735992&partnerID=8YFLogxK
U2 - 10.1016/S0140-6736(03)14068-8
DO - 10.1016/S0140-6736(03)14068-8
M3 - Article
C2 - 12927430
AN - SCOPUS:0041735992
SN - 0140-6736
VL - 362
SP - 433
EP - 439
JO - Lancet
JF - Lancet
IS - 9382
ER -