Annals of Surgical Oncology Cite Track
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
10.1245/s10434-006-9090-0
Annals of Surgical Oncology 14:1058-1064 (2007)
© 2007 Society of Surgical Oncology
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Marchet, A.
Right arrow Articles by Nitti, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Marchet, A.
Right arrow Articles by Nitti, D.

Original Article

Gene Expression Profile of Primary Gastric Cancer: Towards the Prediction of Lymph Node Status

Alberto Marchet, MD1, Simone Mocellin, MD1, Claudio Belluco, MD2, Alessandro Ambrosi, PhD1,3, Francesco DeMarchi, MD2, Enzo Mammano, MD1, Maura Digito, PhD1, Alberta Leon, PhD4, Antonello D’Arrigo, PhD4, Mario Lise, MD1,2 and Donato Nitti, MD1

1 Clinica Chirurgica II, Dipartimento di Scienze Oncologiche e Chirurgiche, Istituto Oncologico Veneto IRCCS and University of Padova, Padova, Italy
2 Surgical Oncology, Centro di Riferimento Oncologico IRCCS, Aviano, Italy
3 University Centre of Statistics for the Biomedical Sciences, Vita-Salute San Raffaele University, Milan, Italy
4 Research & Innovation (R&I) Company, Padova, Italy

Correspondence: Address correspondence and reprint requests to: Donato Nitti, MD; E-mail: donato.nitti{at}unipd.it


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Background: The identification of gastric tumors associated with a higher risk of lymph node metastasis could help surgeons select patients who may benefit from extended lymph node dissection. The aim of this study was to screen the genome in the search of primary gastric cancer gene expression profiles that might predict lymph node status.

Methods: The gene expression profile was evaluated in frozen tumor samples obtained from 32 patients with primary gastric adenocarcinomas. The array consisted of a duplicated spot panel of 5,541 human genes. To classify node-positive (N+) and node-negative (N–) cases, a logistic regression model was fitted optimizing the Akaike Information Criteria after a stepwise gene selection. The accuracy was evaluated by means of leave-one-out cross validation.

Results: All patients underwent radical gastrectomy and extended lymphadenectomy. Of all the cases, 21 were N+ and 11 demonstrated no lymph node involvement (N–). After quality filtering, the analysis of variance selected a set of 136 genes potentially correlated with nodal involvement (P value <.05). Of these 136 genes, 5 were differentially expressed (adjusted P value <.05). After a stepwise gene selection, only three genes (Bik, aurora kinase B, eIF5A2) were retained in the logistic model, which could correctly predict lymph node status in 30 of 32 cases.

Conclusions: If our findings were confirmed, the identified gene pattern might be used to tailor the extent of lymph node dissection on a single patient basis.

Key Words: Gastric cancer • Gene expression profile • Lymph node status • Prognostic markers


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Although the incidence of gastric cancer has significantly decreased over the past few decades, this neoplasia still represents the fourth most common cause of tumor death worldwide1,2. The penetration of the tumor into the gastric wall (T stage), the number of metastatic lymph nodes (N stage) and the recently proposed percentage of metastatic lymph nodes (N ratio)3 are currently considered the most reliable prognostic factors for patients with resectable gastric cancer1,2. At present, there is no consensus on the type of lymphadenectomy for accurate staging and adequate surgical treatment of patients with gastric carcinoma, as randomized trials have failed to demonstrate a survival advantage and different morbidity rates have been reported4,5. In addition, preoperative imaging techniques (e.g., ultrasound endoscopy, CT scan) and intraoperative procedures (e.g., sentinel node technique) are not sufficiently reliable in predicting lymph node spread6–10. As a consequence, the choice between extended (D2) and limited (D1) lymphadenectomy usually depends on the beliefs prevailing in each single institution regarding the role of lymph node dissection in the management of these patients. With these considerations in mind, molecular indicators of disease metastatic potential are being eagerly investigated11.

In recent years, investigators have been proficient in elucidating the cascade of molecular events leading to cancer progression12, and some authors have reported a significant association between the altered expression of single genes/proteins and gastric cancer prognosis13–15 and lymph node status16–19. However, as tumor progression is a process involving the dys-regulation of several genes, it is unlikely that the abnormal expression of single genes or proteins might sustain tumor aggressiveness and—in particular—predict tumor spread to lymph nodes.

Unlike the traditional molecular analyses, which support a reductionist approach to research, high-throughput technologies (e.g., SAGE, DNA micro-array) allow one to test in the same experiment not only multiple hypotheses but also multiple combinations of hypotheses20. Among these techniques, DNA microarrays have become prominent because they are easier to use and do not require large-scale DNA sequencing. Using this genome-wide approach, investigators have recently identified a set of genes that significantly correlate with pathological features of tumors and accurately predict the risk of both disease recurrence and tumor-specific death21,22. Moreover, specific gene clusters sorted out of micro-array experiments have been shown to correlate with lymphatic vessel invasion and lymph node status22,23.

In this study, we analyzed the gene expression profile of the primary tumor from 32 patients who underwent radical surgery for gastric carcinoma. The genetic signature of these tumors was then correlated with the lymph node status, and a statistical model was fitted to predict lymph node metastasis based on the molecular profile of the primary tumors.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Patients and tissue samples
The study population consisted of 32 fully informed patients who had undergone radical resection for gastric adenocarcinoma at our Institution between April 1998 and December 2003 (Table 1Go). Cases were selected from our prospective gastric cancer database on the basis of their pathological characteristics (consecutive cases radically resected staged T2a, T2b and T3 regardless of lymph node involvement) and the availability of frozen tissue. Tumors were staged according to the International Union Against Cancer and the American Joint Committee on Cancer, UICC/AJCC (2002 edition) system (TNM). The histological grade was assessed according to the College of American Pathologists criteria.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Patients and tumor characteristics
 
Of the 32 patients, 17 were male and 15 female. The mean age was 67 years (range: 47–91 years). Tumors were located in the esophago-gastric junction and in the upper, middle and lower thirds of the stomach in 3 (9.4%), 5 (15.6%), 6 (18.7%) and 18 (56.3%) patients, respectively.

According to the TNM classification, 8 cases were T2a (25%), 13 were T2b (40.6%) and 11 were T3 (34.4%). Moreover, at pathological examination, no lymph node metastasis was found in 11 (34.4%) cases (N–); whereas, in 21 cases (65.6%) lymph node involvement was present (N+). In the latter group, 6 were N1 (18.7%) and 15 were N2 (46.9%). According to the TNM staging system, 9 patients were classified in stage Ib (28.1%), 11 in stage II (34.4%), 5 in stage IIIa (15.6%), 6 in stage IIIB (18.8%), and 1 in stage IV (3.1%; T3N2).

With regard to tumor grading, 1 (3.1%) tumor was well differentiated, 15 (46.9%) were moderately differentiated, 13 (40.6%) were poorly differentiated, and 3 (9.4%) were undifferentiated. For gene profiling purposes, one bulk tumor tissue sample of about 5x 5 mm was obtained from each surgical specimen. Biopsies were snap-frozen in liquid nitrogen immediately after excision using RNase-free vials without other protective solutions. Samples were stored in liquid nitrogen until use.

Array construction
Duplicated cDNA arrays comprising 5,541 human genes were assembled onto mirrored aminosilane type-7 STAR slides (Amersham-Pharmacia Biotech, Little Chalfont, UK) by a Lucidea Spotter (Amersham). The DNAs were purchased in the form of purified, sequence-verified PCR products from RZPD (Berlin, Germany) and were diluted in 50% DMSO. The final concentrations of the DNA samples ranged between 100 fmol/µl and 400 fmol/µl. The gene panel was selected with particular regard to the relevance for molecular events related to tumor progression, and included genes controlling apoptosis and cell cycle as well as genes encoding for adhesion molecules and factors involved in cell differentiation. After deposition, slides were briefly exposed to a brief pulse of UVC light to stabilize DNA attachment to the slide surface.

RNA extraction, cDNA synthesis, fluorescent labeling and image visualization
These methods have been described by us in detail previously24. Briefly, total RNA was extracted from tumor biopsies using standard methods (TRIzol, Invitrogen). RNA integrity was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies). A dendrimer-based labeling system (Array50 version 2, Genisphere Inc.) was used for the preparation of fluorescent cDNA probes from tumor samples and slide hybridization. Briefly, 10 µg of total RNA was reverse transcribed into cDNA using 5' tagged oligo (dT) primers. The tagged cDNAs were hybridized overnight at 50°C with the arrays in a humidified chamber, and the slides were sequentially washed as follows: 2X SSC+0.2% SDS at 50°C, 2X SSC at room temperature and 0.2X SSC at room temperature (10 min per washing step). The slides were then incubated with DNA dendrimers containing cyanine dyes and including sequences complementary to the cDNA tags. As a reference sample, a pool of total RNAs from ten different cell lines was used (Universal Human Reference RNA, Stratagene). Cy3 and Cy5 dyes were used for tumor and reference sample labeling, respectively. After washing and drying, fluorescent signals were generated by laser excitation with a Gen III Laser Scanner (Amersham-Pharmacia/ Molecular Dynamics). Images were visualized and signals quantified by Array Vision (Imaging Research Inc.).

Quantitative real-time PCR
The amount of starting RNA was normalized using 18S ribosomal RNA as a control transcript. To this end, a QuantumRNA 18S internal standard kit (Ambion) was utilized, followed by quantification of the electrophoretic bands by ImageQuant (Molecular Dynamics). Primers for quantitative real-time PCR were designed with Primer Express 2.0 (Applied Biosystems). Primer and probe sequences were as follows: Bik, 5'-CCT GGA ACC CCC GAC CAT-3' (forward), 5'-CAC TGC CCT CCA TGC ATT C-3' (reverse), 5'-AGG ACC TGG ACC CTA TGG AGG ACT TCG-3' (probe); aurora kinase B, 5'-ACG CGG CAC TTC ACA ATT GA-3' (forward), 5'-GAG CGC CAC GAT GAA ATG G-3' (reverse) and 5'-TTG GAA ACG TGT ACT TGG CTC GGG A-3' (probe); eIF5A2, 5'-AAG CAG GCC ATT TCA GCA T-3' (forward), 5'-TCA TTA ACC CCA GTT TAT TGA ATC-3' (reverse) and 5'-AGG CAA GTG GCT GGA TGG TAT TCG AA-3' (probe).

Probes were labeled with 6-FAM at the 5'-end and TAMRA at the 3'-end. For the amplification, the qPCR core kit was utilized (Applied Biosystem). PCR conditions were as specified by the manufacturer. A threshold level of fluorescence within the log phase was chosen, and the relative levels of RNA were calculated as a function of the number of amplification cycles required to reach the threshold.

Data analysis
After nonparametric normalization of the logarithmic transformation of raw data, only spots with intensity above the background in both duplicate arrays of each slide and in all samples in each group were considered for further analysis. Denoting the transformed relative intensity as y, we considered the linear model ygijk = µg + Ni + Tj + {varepsilon}gijk, where µg is the mean level of intensity for gene g (g = 1,...,G), and Ni(i = –,+) and Tj(j = 2a, 2b, 3) capture the variations among patients grouped by lymph node status (N– versus N+) and tumor depth (T2a versus T2b versus T3). The error term {varepsilon}gijk was used to describe the residual variability associated to the k-th subject.

Statistical tests were based on the F statistic of the ANOVA relative to the above linear model. P values were obtained by means of permutations and then adjusted using the procedure described by Reiner et al.25. Genes were ranked based on their adjusted P values. We defined as informative any genes with a raw P value of less than .05; those with an adjusted P value of less than .05 were defined as differentially expressed genes.

In order to predict lymph node status, we fitted a logistic model given a subset of genes (x1··· xGs) : p(N + |x1, ··· xGs ) = exp ({sum}Gsi=1{alpha}ixi)/ {lfloor}1 + exp({sum}Gsi=1{alpha}ixi){rfloor}. Since, a number of genes contained in the microarray slides might be irrelevant or redundant with respect to the T and the N stages, the performance of the statistical model might be impaired if a high number of non-informative genes were included in the analysis. To address this issue, we identified a subset of Gs genes by performing a stepwise selection that optimized the Akaike Information Criterion. We, thus, considered as the starting reference set all the genes resulted to be informative in the ANOVA.

Finally, we compared the prediction effciency of the logistic regression model with that obtained with the logitboost model based on the same set of selected genes26.

The accuracy of the fitted models was computed by means of leave-one-out cross validation (LOOCV).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
cDNA microarray analysis
In order to identify gene expression changes associated with lymph node status, cDNA arrays of 5,541 human genes were used to analyze gene expression profile. Valid spots were defined as those with fluorescence intensity above the background in at least 90% of samples in each group. After this initial step, 1,587 genes were sorted out. To identify those genes up-or downregulated in the groups of interest (i.e., N+ versus N–, T2a versus T2b versus T3), we queried this data set by selecting genes differentially expressed according to a paired sample t-test. When the P value was not adjusted, 136 genes were shown to be informative regarding the N stage. Cluster analysis confirmed that these genes were able to correctly identify the two groups of patients (N– versus N+, Fig. 1Go). After adjustment, however, only 5 genes in N+ cases (diacylglycerol [DAG] kinase alpha, hyaluran-mediated motility receptor [HMMR], ADP-ribosylation factor-like 1 [ARL1], cartilage associated protein [CRTAP], EST moderately similar to ZRF1) displayed differential expression when compared with N cases.


Figure 1
View larger version (100K):
[in this window]
[in a new window]

  		FIG. 1. Cluster analysis of 136 informative genes (not adjusted P value <.05) expressed in 32 primary gastric carcinomas. 27 cases (8 node negative and 19 node positive) were correctly classified. Rows genes; columns cases; orange bars node-positive cases; blue bars node-negative cases; heat map red color indicates gene over-expression and green color gene downregulation in primary gastric cancer biopsy relative to a reference sample (a pool of total RNAs from ten different cell lines).

 
Analysis of variance also identified 77 genes informative regarding the T stage, 16 of those being in common with the 136 genes informative for N stage. However, none of these 77 genes was demonstrated to be differentially expressed after adjustment for multiplicity.

Lymph node status prediction model
Considering all the informative genes (i.e., those related to both N and T stages) as a starting set, after a stepwise gene selection the logistic regression model identified three genes (Bcl2-interacting killer [Bik], aurora kinase B/serine-threonine kinase-12 [AIK2/ STK12], and eukaryotic translation initiation factor 5A2 [eIF5A2]) that were strongly associated with lymph node status (Table 2Go).


View this table:
[in this window]
[in a new window]

  		TABLE 2. Selection of genes predictive of lymph node status in 32 patients with gastric carcinoma. Using all informative (not adjusted P value <.05) genes sorted out of microarray experiments, logistic regression analysis (stepwise mode) identified three genes (Bik, aurora kinase B, eIF5A2) strongly associated with lymph node status
 
These genes—which belonged to the set informative for N stage—were not among those differentially expressed after adjustment for multiplicity. Nevertheless, the LOOCV method showed that the model based on these three genes could correctly predict lymph node status in 30 of 32 cases, with an accuracy, sensitivity, specificity, positive predictive value and negative predictive value of 93.7, 90.9, 95.2, 90.9 and 95.2%, respectively (Table 3Go, upper panel).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Prediction of lymph node status (node positive [N+] versus node negative [N–]) in 32 patients with gastric carcinoma using logistic regression and logitboost analysis. CI 95% confidence interval, PPV positive predictive value, NPV negative predictive value
 
To further explore the predictive potential of this set of genes, we used the logitboost model, which could correctly predict lymph node status in 31 of 32 cases, with an accuracy, sensitivity, specificity, positive predictive value and negative predictive value of 96.8, 90.9, 100, 100 and 95.4%, respectively (Table 3Go, lower panel).

Quantitative real-time PCR
Quantitative real-time PCR results are shown in Table 4Go. Of the three genes (Bik, aurora kinase B, eIF5A2) we tested, only Bik was differentially expressed in the two study groups (N+ versus N–). In particular, the transcriptional levels of this pro-apoptotic gene were higher in node-negative than in node-positive patients.


View this table:
[in this window]
[in a new window]

  		TABLE 4. Quantitative real-time PCR analysis of the three genes utilized for lymph node status prediction in 32 patients with gastric cancer. Standard deviation was below 10% of cycle number for all genes considered. N+/N0 ratio = 2cycle difference
 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
The appropriate extension of lymphadenectomy in gastric cancer patients is still debated. Surgeons who advocate D2 extended lymph node dissection claim that this approach guarantees an adequate staging evaluation and might lead to a survival advantage27–29; thus, they recommend D2 lymphadenecto-my as the standard treatment regardless of lymph node status, feeling that extended lymph node dissection may find positive lymph nodes not detected by D1 dissection4,5,27–29. However, detractors of D2 lymphadenectomy are supported by the negative results of randomized trials and underscore the higher morbidity rates associated with this surgical procedure than with more limited (D1) lymphadenectomy5.

D2 lymphadenectomy might be better accepted if performed in subgroups of patients at higher risk of lymph node metastasis. By focusing on this subset of patients, investigators might definitively assess whether or not this type of lymph node dissection has a therapeutic impact on the management of patients with gastric cancer. Although pathological features (e.g., T stage) and single molecular markers (e.g., VEGF, c-Met) are associated with different risks of lymph node metastasis16–19, none of them is reliable enough to be implemented in the clinical setting to predict lymph node status on a single patient basis.

In the search for novel prognostic markers or combinations of them, high-throughput gene micro-array provides investigators with a powerful tool to screen the whole genome20. Using this approach, Weiss et al. reported that gene expression profiles of primary tumors (n=35) can identify groups of patients with statistically different risks of lymph node metastasis23. However, the accuracy (i.e., the percentage of cases correctly classified) and the negative predictive value (i.e., the probability that a patient has no lymph node metastasis when he/she is classified as node negative) were not encouraging (74.2% and 40%, respectively). Moreover, no data were reported on the list of informative genes, precluding any consideration on their biological significance with respect to lymph node status. By utilizing a k-nearest neighbor classifier, in their series (n=54) Teramoto et al. obtained a better accuracy (91.4%)22, but no data were reported on the predictive values of their model.

In our exploratory study, we confirmed that high-throughput gene microarray is an effective method to screen the genome in search of a gene set correlated with lymph node status in patients with gastric cancer. After adjustment for multiplicity, only five genes in N+ cases (DAG kinase alpha, HMMR, ARL1, CRTAP and EST moderately similar to ZRF1) displayed differential expression when compared with N– cases. For two of these five, data exist on their potential role in cancer development and progression. DAG kinase alpha—which belongs to a family of nine kinases (alpha to jota)—phosphorylates the lipid second messenger DAG to produce phosphatidic acid; thus, by influencing the intracellular levels of DAG, DAG kinase alpha can contribute to the regulation of activity of target proteins that are activated by DAG and have an established role in cancer biology (e.g., protein kinase C)30. As regards HMMR, preclinical models indicate that this receptor plays a key role in cell migration and, thus, might contribute to the metastatic potential of malignant cells; moreover, recent evidence suggests that this gene is involved in Ras and Hedgehog signaling pathways31, which are well known to be of relevance in cancer development and progression32,33.

Interestingly, when we used microarray-generated data to build a predictive model, the three genes most "informative" to predict lymph node status (Bik, aurora kinase B, eIF5A2) were not among those differentially expressed after adjustment for multiplicity. These genes code for proteins with a demonstrated role in tumor biology. Bik is a BH3-only member of the Bcl-2 intracellular protein family, which includes Bim, Bmf, Bik, Bad, Bid, Puma, Noxa and Hrk34. These proteins mediate many developmentally programmed and induced cytotoxic signals, and compounds mimicking them are promising anti-cancer agents. When activated, these death ligands engage anti-apoptotic Bcl-2-like proteins via the BH3 domain, inactivating their function and promoting apoptosis. Remarkably, although Bik was not among the genes identified by Teramoto et al., in that work, most genes selected for optimal prediction of lymph node status were apoptosis-correlated (e.g., survivin, clusterin, caspase-8, DPP4)22. Aurora kinases (A to C) are closely related kinases that have been implicated in tumorigenesis as they are important regulators of diverse cell cycle events, ranging from the entry into mitosis, centrosome function, mitotic spindle formation, chromosome biorientation and segregation, and cytokinesis35. Finally, eIF5A2—which functions in the initiation of ribosome-mediated translation of mRNA into a polypeptide—has been found overexpressed in certain human cancer cells (in contrast to its weak normal expression limited to human testis and brain), suggesting a potential role as an oncogene36.

Although quantitative real-time PCR showed that among these three genes only Bik was differentially expressed in node positive when compared with node negative patients, the high accuracy (93.7–96.8%) and negative predictive value (95.2–95.4%) of our prediction models suggest that the combination of the expression levels of these three genes might be biologically more important than the average transcriptional abundance of each single gene. However, the functional interplay among these three genes is purely hypothetical and further research is warranted to biologically substantiate this microarray-generated finding.

Taken together, our results support the strategy of using high-throughput technologies coupled with appropriate statistical models for predicting lymph node status in patients with gastric cancer. However, larger series of patients need to be evaluated before the analysis of the molecular profile of primary tumors might be implemented in the clinical setting to guide surgeons in the decision-making process for the therapeutic management of gastric carcinoma.


   ACKNOWLEDGMENTS
 
This work was in part supported by the AIRC Regional Grant 2005 and by the grant PRIN (MIUR) 2005.


   FOOTNOTES
 
Presented at the 59th Annual Cancer Symposium of the Society of Surgical Oncology, San Diego, California, 23–26 March 2006.

Received for publication June 5, 2006. Accepted for publication June 5, 2006.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Dicken BJ, Bigam DL, Cass C, et al. Gastric adenocarcinoma: review and considerations for future directions. Ann Surg 2005; 241(1):27–39.[Medline]
  2. Hohenberger P, Gretschel S. Gastric cancer. Lancet 2003; 362(9380):305–15.[CrossRef][Medline]
  3. Nitti D, Marchet A, Olivieri M, et al. Ratio between metastatic and examined lymph nodes is an independent prognostic factor after D2 resection for gastric cancer: analysis of a large European monoinstitutional experience. Ann Surg Oncol 2003; 10(9):1077–85.[Abstract/Free Full Text]
  4. Jansen EP, Boot H, Verheij M, van de Velde CJ. Optimal locoregional treatment in gastric cancer. J Clin Oncol 2005; 23(20):4509–17.[Abstract/Free Full Text]
  5. McCulloch P, Niita ME, Kazi H, Gama-Rodrigues JJ. Gastrectomy with extended lymphadenectomy for primary treatment of gastric cancer. Br J Surg 2005; 92(1):5–13.[CrossRef][Medline]
  6. Javaid G, Shah OJ, Dar MA, et al. Role of endoscopic ultra-sonography in preoperative staging of gastric carcinoma. ANZ J Surg 2004; 74(3):108–11.[CrossRef][Medline]
  7. Kim HJ, Kim AY, Oh ST, et al. Gastric cancer staging at multi-detector row CT gastrography: comparison of transverse and volumetric CT scanning. Radiology 2005; 236(3):879–85.[Abstract/Free Full Text]
  8. Kim MC, Kim HH, Jung GJ, et al. Lymphatic mapping and sentinel node biopsy using 99mTc tin colloid in gastric cancer. Ann Surg 2004; 239(3):383–7.[CrossRef][Medline]
  9. Ajisaka H, Miwa K. Micrometastases in sentinel nodes of gastric cancer. Br J Cancer 2003; 89(4):676–80.[CrossRef][Medline]
  10. Ryu KW, Lee JH, Kim HS, et al. Prediction of lymph nodes metastasis by sentinel node biopsy in gastric cancer. Eur J Surg Oncol 2003; 29(10):895–99.[CrossRef][Medline]
  11. Ludwig JA, Weinstein JN. Biomarkers in cancer staging, prognosis and treatment selection. Nat Rev Cancer 2005; 5(11):845–56.[CrossRef][Medline]
  12. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med 2004; 10(8):789–99.[CrossRef][Medline]
  13. Songun I, Litvinov SV, van de Velde CJ, et al. Loss of Ep-CAM (CO17–1A) expression predicts survival in patients with gastric cancer. Br J Cancer 2005; 92(9):1767–72.[CrossRef][Medline]
  14. Scartozzi M, Galizia E, Freddari F, et al. Molecular biology of sporadic gastric cancer: prognostic indicators and novel therapeutic approaches. Cancer Treat Rev 2004; 30(5):451–9.[CrossRef][Medline]
  15. Nitti D, Belluco C, Mammano E, et al. Low level of p27(Kip1) protein expression in gastric adenocarcinoma is associated with disease progression and poor outcome. J Surg Oncol 2002; 81(4):167–75; discussion 175–6.[CrossRef][Medline]
  16. Wang CS, Lin KH, Chen SL, et al. Overexpression of SPARC gene in human gastric carcinoma and its clinic-pathologic significance. Br J Cancer 2004; 91(11):1924–30.[CrossRef][Medline]
  17. Fan Z, Li J, Dong B, Huang X. Expression of Cdx2 and hepatocyte antigen in gastric carcinoma: correlation with histologic type and implications for prognosis. Clin Cancer Res 2005; 11(17):6162–70.[Abstract/Free Full Text]
  18. Juttner S, Wissmann C, Jons T, et al. Vascular endothelial growth factor-D and its receptor VEGFR-3: two novel independent prognostic markers in gastric adenocarcinoma. J Clin Oncol 2006; 24(2):228–40.[Abstract/Free Full Text]
  19. Wang JY, Hsieh JS, Chen CC, et al. Alterations of APC, c-met, and p53 genes in tumor tissue and serum of patients with gastric cancers. J Surg Res 2004; 120(2):242–8.[CrossRef][Medline]
  20. Mocellin S, Provenzano M, Rossi CR, et al. DNA array-based gene profiling: from surgical specimen to the molecular portrait of cancer. Ann Surg 2005; 241(1):16–26.[Medline]
  21. Chen CN, Lin JJ, Chen JJ, et al. Gene expression profile predicts patient survival of gastric cancer after surgical resection. J Clin Oncol 2005; 23(29):7286–95.[Abstract/Free Full Text]
  22. Teramoto K, Tada M, Tamoto E, et al. Prediction of lymphatic invasion/lymph node metastasis, recurrence, and survival in patients with gastric cancer by cDNA array-based expression profiling. J Surg Res 2005; 124(2):225–36.[CrossRef][Medline]
  23. Weiss MM, Kuipers EJ, Postma C, et al. Genomic profiling of gastric cancer predicts lymph node status and survival. Oncogene 2003; 22(12):1872–79.[CrossRef][Medline]
  24. D’Arrigo A, Belluco C, Ambrosi A, et al. Metastatic transcriptional pattern revealed by gene expression profiling in primary colorectal carcinoma. Int J Cancer 2005; 115(2):256–62.[CrossRef][Medline]
  25. Reiner A, Yekutieli D, Benjamini Y. Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics 2003; 19(3):368–75.[Abstract/Free Full Text]
  26. Dettling M, Buhlmann P. Boosting for tumor classification with gene expression data. Bioinformatics 2003; 19(9):1061–69.[Abstract/Free Full Text]
  27. Hartgrink HH, van de Velde CJ. Status of extended lymph node dissection: locoregional control is the only way to survive gastric cancer. J Surg Oncol 2005; 90(3):153–65.[CrossRef][Medline]
  28. Brennan MF. Current status of surgery for gastric cancer: a review. Gastric Cancer 2005; 8(2):64–70.[CrossRef][Medline]
  29. Wu CW, Hsiung CA, Lo SS, et al. Stage migration influences on stage-specific survival comparison between D1 and D3 gastric cancer surgeries. Eur J Surg Oncol 2005; 31(2):153–7.[CrossRef][Medline]
  30. Kazanietz MG. Targeting protein kinase C and "non-kinase" phorbol ester receptors: emerging concepts and therapeutic implications. Biochim Biophys Acta 2005; 1754(1–2):296–304.[Medline]
  31. Levy P, Vidaud D, Leroy K, et al. Molecular profiling of malignant peripheral nerve sheath tumors associated with neurofibromatosis type 1, based on large-scale real-time RT-PCR. Mol Cancer 2004; 3:20.[CrossRef][Medline]
  32. Nottage M, Siu LL. Rationale for Ras and rafkinase as a target for cancer therapeutics. Curr Pharm Des 2002; 8(25):2231–42.[CrossRef][Medline]
  33. Kasper M, Regl G, Frischauf AM, Aberger F. GLI transcription factors: Mediators of oncogenic Hedgehog signalling. Eur J Cancer 2006; 42(4):437–45.[CrossRef][Medline]
  34. Willis SN, Adams JM. Life in the balance: how BH3-only proteins induce apoptosis. Curr Opin Cell Biol 2005; 17(6):617–25.[CrossRef][Medline]
  35. Keen N, Taylor S. Aurorakinase inhibitors as anticancer agents. Nat Rev Cancer 2004; 4(12):927–936.[CrossRef][Medline]
  36. Clement PM, Henderson CA, Jenkins ZA, et al. Identification and characterization of eukaryotic initiation factor 5A-2. Eur J Biochem 2003; 270(21):4254–63.[Medline]



This article has been cited by other articles:


Home page
 Cancer Epidemiol. Biomarkers Prev.Home page
W. Chen, J.-H. Luo, W.-F. Hua, F.-J. Zhou, M. C. Lin, H.-F. Kung, Y.-X. Zeng, X.-Y. Guan, and D. Xie
Overexpression of EIF-5A2 Is an Independent Predictor of Outcome in Patients of Urothelial Carcinoma of the Bladder Treated with Radical Cystectomy
Cancer Epidemiol. Biomarkers Prev., February 1, 2009; 18(2): 400 - 408.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Marchet, A.
Right arrow Articles by Nitti, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Marchet, A.
Right arrow Articles by Nitti, D.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS