FASEB J.
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

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 Email this article to a friend
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
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by TÜRECI, O.
Right arrow Articles by HAMMER, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by TÜRECI, O.
Right arrow Articles by HAMMER, J.
(The FASEB Journal. 2003;17:376-385.)
© 2003 FASEB

Computational dissection of tissue contamination for identification of colon cancer-specific expression profiles

ÖZLEM TÜRECI1, JIAYI DING*, HOLLY HILTON*, HONGJIN BIAN*, HITOMI OHKAWA*, MICHAEL BRAXENTHALER*, GERHARD SEITZ{dagger}, LAURA RADDRIZZANI*, HELMUT FRIESS{ddagger}, MARKUS BUCHLER{ddagger}, UGUR SAHIN2 and JUERGEN HAMMER*,2

III. Medizinische Klinik und Poliklinik, Johannes Gutenberg Universität Mainz, D-55131 Mainz, Germany;
* Roche Genomic and Information Sciences, Hoffmann-La Roche Inc., Nutley, New Jersey, USA;
{dagger} Institut für Pathologie, Klinikum Bamberg, 96049 Bamberg, Germany; and
{ddagger} Department of Surgery, Neuenheimer Feld 110, 69120 Heidelberg, Germany

1Correspondence: III. Medizinische Klinik, Johannes-Gutenberg Universität Mainz, Verfügungsgebäude für Forschung und Entwicklung, Obere Zahlbacher Str. 63, D-55131 Mainz, Germany. E-mail: Tureci{at}mail.uni-mainz.de


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Microarray profiles of bulk tumor tissues reflect gene expression corresponding to malignant cells as well as to many different types of contaminating normal cells. In this report, we assess the feasibility of querying baseline multitissue transcriptome databases to dissect disease-specific genes. Using colon cancer as a model tumor, we show that the application of Boolean operators (AND, OR, BUTNOT) for database searches leads to genes with expression patterns of interest. The BUTNOT operator for example allows the assignment of "expression signatures" to normal tissue specimens. These expression signatures were then used to computationally identify contaminating cells within conventionally dissected tissue specimens. The combination of several logic operators together with an expression database based on multiple human tissue specimens can resolve the problem of tissue contamination, revealing novel cancer-specific gene expression. Several markers, previously not known to be colon cancer associated, are provided.—Türeci, O., Ding, J., Hilton, H., Bian, H., Ohkawa, H., Braxenthaler, M., Seitz, G., Raddrizzani, L., Friess, H., Buchler, M., Sahin, U., Hammer, J. Computational dissection of tissue contamination for identification of colon cancer-specific expression profiles.


Key Words: expression signature • colorectal cancer • profile


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
WITH ITS CAPACITY for simultaneous monitoring of the transcriptional state of thousands of genes, the inauguration of microarray technology has provided investigators with a unique opportunity for high-throughput genetic analysis in physiological and pathological states (1 , 2) . There is a considerable interest to apply microarray technology for human cancers. Several studies resulted in identification of new tumor markers and the determination of molecularly defined subsets of cancers (3 4 5 6 7) . However, there are still many technical and logistic challenges. Each individual microarray profile from bulk tumor tissue reflects gene expression that corresponds to malignant cells as well as to many different types of contaminating normal cells. Resorting to tumor cell lines as surrogates may not accurately reflect the molecular events taking place in situ in the tissue milieu from which they were derived. Microdissection techniques (8) and positive and negative affinity purification protocols (9) might allow a purer sampling of cells from fresh tumor specimen. But it is not apparent yet whether such approaches are feasible for primary screens of larger numbers of samples and whether amplification of mRNA obtained from microdissected tissues conserves the representation of transcripts. With colorectal cancer as an example, we report that such limitations can be circumvented by computational means allowing an exclusion of expression data contributed by contaminating normal cells in heterogeneous tumor samples. We show that the genetic variability among humans does allow for the creation of generic baseline transcriptomes representing genome-wide profiles for normal tissues. We demonstrate the benefits of multitissue expression databases by generating a model database. Interrogating it with logic operators such as AND (intersection), OR (union), and BUTNOT (difference) can help search for sets of genes with expression patterns of interest. Logic operators allowed for the formation of tissue-specific expression signatures, quality control, and computational "dissection" of human transcript profiles. Our results indicate that genome-wide expression analysis can be achieved and is of value in elucidating the genetic events associated with colon cancer progression.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
RNA extraction and labeling
Total RNA was extracted from snap-frozen human cells using Ultraspec RNA isolation kits (Biotecx, Houston, TX, USA), and purified using RNeasy mini kits (Qiagen, Chatsworth, CA, USA). Total RNA (5–20 µg) was converted into double-stranded cDNA by reverse transcription (GIBCO BRL Life Technologies, Grand Island, NY, USA) using the T7-T24 primer [5'-GGC CAG TGA ATT GTA ATA CGA CTC ACT ATA GGG AGG CGG (dT24)] and cleaned up by phenol/chloroform/isoamyl extraction using phase lock gel (5 Prime-3 Prime Inc., Boulder, CO, USA). For conversion into cRNA, the in vitro transcription (IVT) MEGAscriptTM T7 kit (Ambion, Austin, TX, USA) and biotinylated nucleotides were used. The IVT product was purified using RNeasy mini columns (Qiagen) and fragmented.

Hybridization, staining, and image analysis
Hybridization of fragmented IVT product to Affymetrix HuGeneFL (6800 human full-length genes) and Hu35K sub A-D arrays (35000 human ESTs) (http://www.affymetrix.com/products/tech_probe.html) and wash steps were performed as suggested by the manufacturer (Affymetrix, Santa Clara, CA, USA). Each hybridized Affymetrix GeneChip® array was scanned with an argon-ion laser scanner at 570 nm (Agilent/Affymetrix). Initial absolute and comparative analysis on the resulting data images was performed with Affymetrix custom image analysis software, GeneChip® version 3.1.

Computational analysis of data
Primary analysis of array data was performed using Affymetrix GeneChip® software, resulting in parameters (see manufacturer) such as "Absolute Call" (absent, present), Difference Call (increased, decreased), Average Difference intensity), Fold Change, and Sort Score (significance), which were published into our expression database and used for querying. For pairwise analysis of array data, Affymetrix GeneChip® software was used. The program DiseaseExpressionMap was developed to evaluate expression data from profiling experiments over multiple matched pairs of disease/normal samples. For each gene, the program computes two new parameters: the frequency (% patient coverage) of change of expression within the population:

and the average Sort Score g (see manufacturer) determined for tissue pairs showing change of expression:

In both definitions, Nchangeg is the number of disease/normal sample pairs in which gene g was differentially expressed (>=2-fold).

In cases of variable numbers of unmatched normal/disease tissue samples a hierarchical scheme of logical operations was applied. The top level condition for a gene to be of interest then becomes: (fcancerg>=fcancerg,cut) AND (fnormalg<=fnormalg,cut), where fg is the computed frequency of gene g in the cancer and normal tissue group, respectively, and fg,cut is the user-defined cutoff frequency for the cancer and normal tissue group, respectively. In our example the user-defined parameters were fcancerg,cut = 0.5 and fnormalg,cut = 0.0. Computation of a frequency required a definition of the sample set for which the frequency is computed and one or more conditions that are evaluated over the sample set to generate the frequency. The condition to be evaluated for frequency generation was the Absolute Call (present or absent/marginal).

To manage the output from the Affymetrix GeneChip® software, a relational database was created and divided into four parts: description of scanning experiments, description of samples used, numerical experimental results for individual genes, and gene information. The data are effectively queried via ad hoc SQL (Structured Query Language) queries as well as applications in a variety of forms, e.g., on the Web, as a client (both command line and with graphical interface). Applications developed for repeatedly performed data analyses and processing through ad hoc SQL queries turned out to be very useful for answering exploratory questions.

Real-time quantitative PCR
Duplex real-time PCR (target gene and GAPDH as reference gene) on 96-well optical plates was performed with reverse transcribed RNA derived from different tissues using the TaqMan® technology and analyzed on ABI PRISM® PE7700 Sequence Detection System [Perkin-Elmer Applied Biosystems (PtdEtn), Norwalk, CT, USA] according to standard protocols. The data were normalized to internal GAPDH and represented as relative expression (E)

whereas delta Ct is the difference of threshold of cycle number between GAPDH and the target gene. Specific PCR primer pairs (5’, 3’) and fluorogenic probes (P) respectively were used for genes of interest at the following PCR conditions: 50°C for 2 min, then 95°C for 10 min, followed by 40 cycles at 95°C/15 s and 62°C/1 min.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Pairwise comparisons of matched "cancer/normal" specimen
The generation of basic comparison profiles requires a minimum of two parameters: 1) gene identity and 2) difference of expression in disease and normal tissue (e.g., fold change). Using high-density Affymetrix GeneChip® microarrays covering 6800 full-length genes plus ESTs equivalent to 35 000 genes, we performed a pilot study by profiling 19 primary colorectal cancer tissues and 19 matched adjacent tissues. We identified 1350 independent genes differentially expressed over 19 matched tumor/normal pairs with the full-length gene array alone, considering only the two parameters mentioned above. The recruitment of such a large number of genes in downstream approaches for reassessment of their authenticity is not feasible, showing the need for a more stringent selection process. We defined two novel parameters to select genes that correlate more significantly with disease: 1) percent patient coverage (PPC) and 2) average Sort Score. PPC measures the fraction of sample pairs in which a given gene is called differentially expressed (>2-fold). The average Sort Score of a gene is the mean of Sort Score parameters derived from sample pairs in which a gene is called differentially (>2-fold) expressed. We have generated an algorithm that reads expression data and computes both parameters. Each differentially expressed gene was plotted according to both parameters and the resulting disease expression map was divided in multiple bins (Fig. 1 ). We found more genes with both high average Sort Score (>1) and high PPC (>0.75) than genes with a low average Sort Score (<1) and high PPC (>0.75). Hence, genes differentially expressed in a large fraction of tissue pairs are on average more significant, i.e., less likely to be noise. In our colorectal cancer example (Fig. 1) , bins 1–5 were excluded due to a general low reproducibility of data with Sort Score <1, whereas bin 6 and 10 were excluded due to a low PPC. Considering all other bins ([average Sort Score <1 AND PPC >0.5] and [average Sort Score >1 AND PPC >0.25]) reduced the number of genes for further analysis from 1350 to 192 (15%) (Fig. 1) .



View larger version (28K):
[in this window]
[in a new window]
  		Figure 1. Disease expression map for colorectal cancer from 19 patients derived with HuGeneFL array. A) Each differentially expressed gene was plotted according to both average Sort Score and PPC using the program DiseaseExpressionMap. B) A large percentage of genes falls into bins that are less correlated with disease.

The differentially expressed genes of bin 16 (average Sort Score >2 AND PPC >0.75) are most significantly correlating with the disease phenotype (Table 1 ). Deregulation of many of these genes has already been described in the context of colorectal cancer (references provided).


View this table:
[in this window]
[in a new window]
  		Table 1. Genes of ‘bin16’a

Generating baseline transcriptomes for normal tissues
To test the feasibility of genome-wide profiles for normal tissues (baseline transcriptomes) and determine levels of variation in expression patterns for different human tissues, we profiled multiple human tissues using Affymetrix HuGeneFL array and subsequently subjected the data obtained to comparison analyses (Table 2 ). The variation in expression level between tissues derived from different donors was compared for several types of normal human tissues (muscle, colon, pancreas, lymph node, liver). At higher significance levels (Sort Score >1), the inter-human variation among normal tissues was quite low, ranging from 1 to 5% (Table 2) , particularly if compared with inter-tissue variation. Variations may reflect differences in the anatomical complexity of a given tissue coupled with sampling differences in surgery. Similar observations were made by comparing different tissue types by scatter blots (data not shown), indicating a low level of variation, making the generation of tissue type-specific baseline transcriptomes feasible. A relatively small number of normal tissue samples per tissue type might therefore be sufficient for the generation of low-resolution tissue expression maps.


View this table:
[in this window]
[in a new window]
  		Table 2. Tissue expression variationa

Creating expression signatures by using BUTNOT operator queries in a multitissue expression database
A model expression database that consisted of expression profiles of tissues along the clinical path of colorectal cancer was generated. The database contained > 1.5 million expression data points derived from profiling 30 additional colon and primary colorectal cancer tissues, lymph nodes, and local lymph node metastases, as well as liver and remote liver metastases generated using Affymetrix HuGeneFL and Hu35KsubA-D oligonucleotide arrays. The data were published into an Oracle expression database and interrogated by Boolean search operators. To determine whether contaminating tissue was present in our surgical specimens, we generated tissue type-specific expression signatures for normal colon, lymph node, and liver tissues by applying the BUTNOT operator (Table 3 )on expression databases. Thus, genes were selected, for example, that are present in all liver tissues but not in lymph nodes and not in colon tissues, or vice versa. A significant part of the colon expression signature was also found in primary colorectal cancer tissues, local lymph node metastases, and remote liver metastases, confirming that the origin of the malignant tissue are colon mucosa cells (Fig. 2 A). In contrast, liver and lymph node expression signatures were mainly found in metastases located in liver and lymph node (Fig. 2A ), respectively, a pattern most easily explained by contamination with adjacent tissue. The power of the expression signature concept for data analysis is demonstrated by the following example. To identify metastases-specific transcripts, we performed a simple pairwise comparison of liver metastases to primary colorectal cancer tissues using our expression database and the BUTNOT operator, revealing 181 candidate metastasis-specific genes that were found to be expressed in at least 50% of liver metastases but not in primary colon cancer. However, by comparing these results to the liver expression signatures more than two-thirds of these transcripts were identified in normal liver tissues (Fig. 2A ). Thus, a multitissue database approach rather than a simple pairwise comparison disclosed with higher authenticity the panel of genes with expression patterns of interest.


View this table:
[in this window]
[in a new window]
  		Table 3. Tissue expression signatures



View larger version (67K):
[in this window]
[in a new window]
  		Figure 2. Interrogation of a model expression database using logic operators. A) Examples for expression signature genes from normal colon, lymph node, and liver shown across colon cancer and its metastases. Expression signatures contain genes whose expression was turned on (Absolute Call= present) consistently in only one of the three tissues. Black boxes are genes called present, empty boxes are genes called absent. B) Examples of gene expression specific to primary colorectal cancer and metastases using the query "[primary cancer OR local metastases OR remote metastases] BUTNOT [colon OR lymph node OR liver]." The patterns are sorted based on frequency of expression in cancer tissues and are derived from microarray experiments using the HuGeneFL array. C) Genes identified by this query were analyzed in 81 other normal and tumor tissues, revealing tumor-restricted expression spectrum of the endogeneous retroviral protease of RVTLH (ERVP).

Identification of disease-specific genes via multiple operator queries
Based on the expression signatures obtained for our cancer tissues, we designed simple but robust queries to maximize the selection for candidate cancer-specific transcripts. For example, the query "[primary cancer OR local metastases OR remote metastases] BUTNOT [colon OR lymph node OR liver]" selected transcripts that were present in metastases and/or primary cancer tissues but not in any of the host tissues from which the metastatic specimen were dissected (Fig. 2B ). Thus, gene expression determined in crude liver metastasis profiles was compared with the respective tissue signatures to computationally eliminate colon transcripts representing the cell of origin, contaminating lymphatic cell transcripts, and hepatic transcripts contributed by the host organ. Of 29,057 expressed genes in our model database, only 60 were expressed in >= 50% cancer tissues, but in none of the host tissues (Fig. 2B ). Twenty-three were derived from known full-length genes and 37 from ESTs. The number of selected genes collapsed further, since not all ESTs that matched to full-length genes were excluded within Affymetrix Hu35K designs, so that only 34 hits from ESTs remained. For several of these, BLAST searches in databases revealed homology to known genes. Table 4 exemplifies genes identified by this approach. Figure 3 shows confirmation of array data for several of these genes by quantitative RT-PCR in an extended panel of tissues supporting authenticity of transcriptional differences obtained with array technology by an independent assay. Genes like GA733 and fibroblast activation protein, described not only in the context of colon cancer but also used in experimental therapeutical studies, were found, confirming our approach. Others have not been described in the context of colon cancer so far.


View this table:
[in this window]
[in a new window]
  		Table 4. Selection of disease-specific genes using logical operatorsa



View larger version (53K):
[in this window]
[in a new window]
  		Figure 3. Confirmation of query results by high-throughput Real-time PCR. The relative expression levels of selected genes (Table 4) in 78 human tissue specimens. Data are normalized to internal GAPDH, whereby expression of GAPDH is set to 1. Note that EST 4 and 6 show a higher change in relative expression (~1000- to 10,000-fold) in disease tissues than EST 1 and 2, even though array data would suggest the contrary (Table 4) , indicating the need for secondary filters with broader dynamic range.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Although microarrays have now become established tools to study gene expression patterns in human cancer (10) , the process of carrying out array experiments and extracting useful information from data obtained is not standardized and depends on multiple parameters, including the type of specimen used. A particular challenge are tumor specimen, which naturally represent a mixture of tissues. We addressed this by testing two different approaches to analyze data obtained from profiling of colon cancer by high-density Affymetrix GeneChip® microarrays.

On the one hand, pairwise comparisons of matched normal and tumor specimen obtained from the same individuals were performed. Introducing average Sort Score ("significance") and PPC ("frequency") as new parameters, a disease expression map was generated to aid the systematic analysis of differentially expressed genes from large datasets derived from multiple patients. Thresholds could be set depending on the number of patients and the specific questions asked. Since the differentially expressed genes of bin 16 (average Sort Score >2 AND PPC >0.75) are most significantly correlating with the disease phenotype, they were of major interest in our colorectal cancer example. Deregulation of many of the genes identified have been described previously in the context of colorectal cancer, thus confirming our simple strategy of integrating multiple pairwise comparisons. However, this approach has its limitations. Whereas significant decreases are frequently due to an authentic down-regulation of the respective genes during malignant transformation, elevations in expression levels of many genes may be contributed by contaminating leucocytes or, in the case of samples derived from metastases, by stroma of the host organ. They also may reflect persistence of differentiation markers of the cell type of origin, physiologically present at low numbers in the respective tissue, during neoplastic clonal expansion. Indeed, we have observed that many of the gene products identified in other bins represent leukocyte markers (data not shown). To discern patterns of gene expression consistent with the presence of a mixture of cell types in conventionally dissected tissue samples, we resorted to the generation of multitissue expression databases for interrogation with logic operators. We could prove in this report that bulk human tissue and disease profiling is feasible due to low variation of expression in normal tissues and the feasibility of publishing scaled expression data into databases rather than being dependent on simple pairwise comparisons. We also demonstrated that transcriptomes can be determined first. Once a critical mass of expression data has been reached for disease and surrounding host tissues, logic operators can be used to define in silico 1) means for quality control or 2) queries to isolate disease-related genes. The potential of this approach proven here can be fully exploited and increased by making reference databases available from all cellular subtypes constituting a tissue of interest after cell sorting.

The rediscovery of known colon cancer markers validated our approach. The surface antigen GA733 (17A-1) is expressed on the majority of human colorectal carcinoma cells and has been used as a target for monoclonal antibody-based immunotherapeutical trials of colon cancer patients during the last decade (11 , 12) . EST 6 matched to fibroblast activation protein-{alpha}, a cell surface glycoprotein abundantly expressed by reactive stromal fibroblasts of epithelial cancers (13) , also an immune target in colon cancer for clinical trials using monoclonal antibodies (14) . MMP-1 is known from earlier studies to play a significant role in colorectal cancer progression (15) . Whereas MMP-1 expression has been correlated with invasiveness and high DUKES stage (16) , our data reveal expression in primary tumors to be more prominent than in metastasis-derived specimens. Unexpectedly, we revealed increased transcript levels for ESTs 1 and 5 matching to kallikrein-10 and maspin, both reported to be down-regulated in breast cancers and regarded as potential tumor suppressor genes (17 , 18) . Real-time PCR confirmed the up-regulation of these genes in colon cancer. Our observations are in line with the finding of elevated kallikrein-10 serum-levels in 15% of patients with gastrointestinal tumors as well as increase of maspin expression reported for tumors such as hepatocarcinoma and pancreatic cancer. The coincident divergence of expression levels in breast and colon cancer for both transcripts may reflect distinct roles of these transcripts in different tissue types.

Another novel discovery is overexpression of the protease gene of endogenous retrovirus RTVLH. As immune responses against endogenous retrovirus-derived subgenomic particles have been described in other tumors, e.g., testicular malignancies (19) , our finding calls for closer evaluation of this gene product as a potential target for cancer immunotherapy.

The approach described here is one of many ways to interrogate expression databases and was chosen because of our specific interest in immune therapeutic targets where very stringent "specificity filters" are required. Statistical approaches can be used to search for more subtle differences in gene expression, whereas application of powerful clustering algorithms may lead to the identification of co-regulated genes in large expression datasets (20 21 22 23) . The common goal of each of these computational processes is to allow a stringent narrowing down to a selected panel of candidate genes, for which as far as quantity is concerned downstream strategies for further assessment are manageable. Although expression databases in combination with computational tools are important primary filters for large datasets, they cannot circumvent the limits of the underlying technological platform. Both the larger dynamic range and the >1000-fold higher sensitivity of real-time PCR technology did indeed confirm our query results, but revealed relative expression differences between disease and normal tissues rather than on/off patterns, as array data might have suggested. We therefore use routinely high-throughput quantitative Real-time PCR on a larger population of cancer and normal tissues as a secondary filter to select for candidate targets. The strategy for further downstream analysis and the generation of additional information on these gene products depends on the intentions pursued. The strategy to identify immunogenic epitopes for cancer vaccine therapy we proposed elsewhere would be one option (24) . The purpose of disease profiling is the identification of genes whose level of expression correlates with the disease phenotype. Although correlation with disease does not a priori imply causation, it provides a means to filter for candidate therapeutic and diagnostic targets. The application of this approach to clinical cancer specimens may provide a key step to rapid advances in cancer detection, diagnosis, and therapeutics.


   ACKNOWLEDGMENTS
 
We are grateful to Steve Ritland, Chris Harrington, Mitchell Martin, John Hill, and Duncan Walker for helpful discussions and comments on the manuscript. We thank the Cooperative Human Tissue Network for providing access to human tissue samples. We thank Christoph Huber for continuous discussions and support. Ö.T. and U.S. have been sponsored by the Deutsche Forschungsgemeinschaft (TU 115/1–2 and D1 of the Sonderforschungsbereich SFB432).


   FOOTNOTES
 
2 Both authors contributed equally

Received for publication May 21, 2002. Accepted for publication November 25, 2002.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Schena, M., Shalon, D., Davis, R. W., Brown, P. O. (1995) Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270,467-470[Abstract/Free Full Text]
  2. DeRisi, J. L., Iyer, V. R., Brown, P. O. (1997) Exploring the metabolic and genetic control of gene expression on a genomic scale. Science 278,680-686[Abstract/Free Full Text]
  3. Alon, U., Barkai, N., Notterman, D. A., Gish, K., Ybarra, S., Mack, D., Levine, A. J. (1999) Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc. Natl. Acad. Sci. USA 96,6745-6750[Abstract/Free Full Text]
  4. Moch, H., Schraml, P., Bubendorf, L., Mirlacher, M., Kononen, J., Gasser, T., Mihatsch, M. J., Kallioniemi, O. P., Sauter, G. (1999) High-throughput tissue microarray analysis to evaluate genes uncovered by cDNA microarray screening in renal cell carcinoma. Am. J. Pathol. 15,979-980
  5. Khan, J., Simon, R., Bittner, M., Chen, Y., Leighton, S. B., Pohida, T., Smith, P. D., Jiang, Y., Gooden, G. C., Trent, J. M., et al (1998) Gene expression profiling of alveolar rhabdomyosarcoma with cDNA microarrays. Cancer Res. 58,5009-5013[Abstract/Free Full Text]
  6. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P., Coller, H., Loh, M. L., Downing, J. R., Caligiuri, M. A., Bloomfield, C. D., Lander, E. S. (1999) Molecular classification of cancer: class discovery and classprediction by gene expression monitoring. Science 286,531-537[Abstract/Free Full Text]
  7. Wellmann, A., Thieblemont, A., Pittaluga, S., Sakai, A., Jaffe, E. S., Siebert, P., Raffeld, M. (2000) Detection of differentially expressed genes in lymphomas using oligonucleotide microarrays: identification of clusterin as a new diagnostic marker for anaplastic large-cell lymphomas. Blood 96,398-404[Abstract/Free Full Text]
  8. Emmert-Buck, M. R., Bonner, R. F., Smith, P. D., Chuaqui, R. F., Zhuang, Z., Goldstein, S. R., Weiss, R. A., Liotta, L. A. (1996) Laser capture microdissection. Science 274,998-1001[Abstract/Free Full Text]
  9. St. Croix, B., Rago, C., Velculescu, V., Traverso, G., Romans, K. E., Montgomery, E., Lal, A., Riggins, G. J., Lengauer, C., Vogelstein, B., et al (2000) Genes expressed in human tumor endothelium. Science 28,1197-1202
  10. Lockhart, D. J., Dong, H., Byrne, M. C., Follettie, M. T., Gallo, M. V., Chee, M. S., Mittmann, M., Wang, C., Kobayashi, M., Horton, H., et al (1996) Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat. Biotechnol. 1,1675-1680
  11. Zutter, M. M. (1998) Gastrointestinal carcinoma antigen GA733: target for immunodestruction and potential modifier of invasiveness and chemoresponsiveness. J. Natl. Cancer Inst. 90,642-644[Free Full Text]
  12. Riethmuller, G., Holz, E., Schlimok, G., Schmiegel, W., Raab, R., Hoffken, K., Gruber, R., Funke, I., Pichlmaier, H., Hirche, H., et al (1998) Monoclonal antibody therapy for resected Dukes C colorectal cancer: seven-year outcome of a multicenter randomized trial. J. Clin. Oncol. 16,1788-1794[Abstract]
  13. Rettig, W. J., Su, S. L., Fortunato, S. R., Scanlan, M. J., Raj, B. K., Garin-Chesa, P., Healey, J. H., Old, L. J. (1994) Fibroblast activation protein: purification, epitope mapping and induction by growth factors. Int. J. Cancer 58,385-392[Medline]
  14. Welt, S., Divgi, C. R., Kemeny, N., Finn, R. D., Scott, A. M., Graham, M., Germain, J. S., Richards, E. C., Larson, S. M., Oettgen, H. F., et al (1994) Antibody targeting in metastatic colon cancer: a phase I study of monoclonal antibody F19 against a cell-surface protein of reactive tumor stromal fibroblasts. J. Clin. Oncol. 12,1193-1203[Abstract/Free Full Text]
  15. Murray, G. I., Duncan, M. E., O'Neil, P., Melvin, W. T., Fothergill, J. E. (1996) Matrix metalloproteinase-1 is associated with poor prognosis in colorectal cancer. Nat. Med. 2,461-462[CrossRef][Medline]
  16. Shiozawa, J., Ito, M., Nakayama, T., Nakashima, M., Kohno, S., Sekine, I. (2000) Expression of matrix metalloproteinase-1 in human colorectal carcinoma. Mod. Pathol. 13,925-933[CrossRef][Medline]
  17. Zou, Z., Anisowicz, A., Hendrix, M. J., Thor, A., Neveu, M., Sheng, S., Rafidi, K., Seftor, E., Sager, R. (1994) Maspin, a serpin with tumor-suppressing activity in human mammary epithelial cells. Science 26,526-529
  18. Liu, X. L., Wazer, D. E., Watanabe, K., Band, V. (1996) Identification of a novel serine protease-like gene, the expression of which is down-regulated during breast cancer progression. Cancer Res. 56,3371-3379[Abstract/Free Full Text]
  19. Boller, K., Janssen, O., Schuldes, H., Tonjes, R. R., Kurth, R. (1997) Characterization of the antibody response specific for the human endogenous retrovirus HTDV/HERV-K. J. Virol. 71,4581-4588[Abstract]
  20. Alon, U., Barkai, N., Notterman, D. A., Gish, K., Ybarra, S., Mack, D., Levine, A. J. (1999) Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc. Natl. Acad. Sci. USA 96,6745-6750
  21. Werner, T. (2001) Cluster analysis and promoter modelling as bioinformatics tools for the identification of target genes from expression array data. Pharmacogenomics 2,25-36[CrossRef][Medline]
  22. Herrero, J., Valencia, A., Dopazo, J. (2001) A hierarchical unsupervised growing neural network for clustering gene expression patterns. Bioinformatics 17,126-136[Abstract/Free Full Text]
  23. Notterman, D. A., Alon, U., Sierk, A. J., Levine, A. J. (2001) Transcriptional gene expression profiles of colorectal adenoma, adenocarcinoma, and normal tissue examined by oligonucleotide arrays. Cancer Res. 61,3124-3130[Abstract/Free Full Text]
  24. Sturniolo, T., Bono, E., Ding, J., Raddrizzani, L., Türeci, Ö., Sahin, U., Braxenthaler, M., Gallazzi, F., Protti, M. P., Sinigaglia, F., et al (1999) Generation of tissue-specific and promiscuous HLA ligand databases using DNA microarrays and virtual HLA class II matrices. Nat. Biotechnol. 1,555-561
  25. Takahashi, K., Sasano, H., Fukushima, K., Hirasawa, G., Miura, H., Sasaki, I., Matsuno, S., Krozowski, Z. S., Nagura, H. (1998) 11 beta-hydroxysteroid dehydrogenase type II in human colon: a new marker of fetal development and differentiation in neoplasms. Anticancer Res. 18,3381-3388[Medline]
  26. Tschmelitsch, J., Barendswaard, E., Williams, C., Jr, Yao, T. J., Cohen, A. M., Old, L. J., Welt, S. (1997) Enhanced antitumor activity of combination radioimmunotherapy (131I-labeled monoclonal antibody A33) with chemotherapy (fluorouracil). Cancer Res. 57,2181-2186[Abstract/Free Full Text]
  27. Shankavaram, U. T., DeWitt, D. L., Funk, S. E., Sage, E. H., Wahl, L. M. (1997) Regulation of human monocyte matrix metalloproteinases by SPARC. J. Cell. Physiol. 173,327-334[CrossRef][Medline]
  28. Melis, R., White, R. (1999) Characterization of colonic polyps by two-dimensional gel electrophoresis. Electrophoresis 20,1055-1064[CrossRef][Medline]
  29. Rechreche, H., Mallo, G. V., Montalto, G., Dagorn, J. C., Iovanna, J. L. (1997) Cloning and expression of the mRNA of human galectin-4, an S-type lectin down-regulated in colorectal cancer. Eur. J. Biochem. 248,225-230[Medline]
  30. Hanamura, N., Yoshida, T., Matsumoto, E., Kawarada, Y., Sakakura, T. (1997) Expression of fibronectin and tenascin-C mRNA by myofibroblasts, vascular cells and epithelial cells in human colon adenomas and carcinomas. Int. J. Cancer 73,10-15[CrossRef][Medline]
  31. Bekku, S., Mochizuki, H., Takayama, E., Shinomiya, N., Fukamachi, H., Ichinose, M., Tadakuma, T., Yamamoto, T. (1998) Carbonic anhydrase I and II as a differentiation marker of human and rat colonic enterocytes. Res. Exp. Med. 198,175-185[CrossRef][Medline]
  32. Hanski, C., Riede, E., Gratchev, A., Foss, H. D., Bohm, C., Klussmann, E., Hummel, M., Mann, B., Buhr, H. J., Stein, H., et al (1997) MUC2 gene suppression in human colorectal carcinomas and their metastases: in vitro evidence of the modulatory role of DNA methylation.. Lab. Invest. 77,685-695[Medline]
  33. Fleming, R. E., Parkkila, S., Parkkila, A. K., Rajaniemi, H., Waheed, A., Sly, W. S. (1995) Carbonic anhydrase IV expression in rat and human gastrointestinal tract regional, cellular, and subcellular localization. J. Clin. Invest. 96,2907-2913
  34. Mulder, T. P., Verspaget, H. W., Janssens, A. R., de Bruin, P. A., Griffioen, G., Lamers, C. B. (1990) Neoplasia-related changes of two copper (Cu)/zinc (Zn) proteins in the human colon. Free Radic. Biol. Med. 9,501-506[CrossRef][Medline]
  35. Cohen, M. B., Hawkins, J. A., Witte, D. P. (1998) Guanylin mRNA expression in human intestine and colorectal adenocarcinoma. Lab. Invest. 78,101-108[Medline]
  36. Broitman, S. A., Cerda, S., Wilkinson, J. T. (1993) Cholesterol metabolism and colon cancer. Prog. Food Nutr. Sci. 17,1-40[Medline]
  37. Kinugasa, T., Kuroki, M., Takeo, H., Matsuo, Y., Ohshima, K., Yamashita, Y., Shirakusa, T., Matsuoka, Y. (1998) Expression of four CEA family antigens (CEA, NCA, BGP and CGM2) in normal and cancerous gastric epithelial cells: up-regulation of BGP and CGM2 in carcinomas. Int. J. Cancer 76,148-153[CrossRef][Medline]
  38. Schweinfest, C. W., Henderson, K. W., Suster, S., Kondoh, N., Papas, T. S. (1993) Identification of a colon mucosa gene that is down-regulated in colon adenomas and adenocarcinomas. Proc. Natl. Acad. Sci. USA 90,4166-4170[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Ann. Surg. Oncol.Home page
D. Z. J. Chu, G. Gibson, D. David, and Y. Yen
The Surgeon's Role in Cancer Prevention. The Model in Colorectal Carcinoma
Ann. Surg. Oncol., November 1, 2007; 14(11): 3054 - 3069.
[Abstract] [Full Text] [PDF]

Home page
 Proc. Natl. Acad. Sci. USAHome page
A. Abdollahi, C. Schwager, J. Kleeff, I. Esposito, S. Domhan, P. Peschke, K. Hauser, P. Hahnfeldt, L. Hlatky, J. Debus, et al.
Transcriptional network governing the angiogenic switch in human pancreatic cancer
PNAS, July 31, 2007; 104(31): 12890 - 12895.
[Abstract] [Full Text] [PDF]

Home page
 Clin. Cancer Res.Home page
C.-Y. Hao, D. H. Moore, P. Wong, J. L. Bennington, N. M. Lee, and L.-C. Chen
Alteration of Gene Expression in Macroscopically Normal Colonic Mucosa from Individuals with a Family History of Sporadic Colon Cancer
Clin. Cancer Res., February 15, 2005; 11(4): 1400 - 1407.
[Abstract] [Full Text] [PDF]

Home page
 CarcinogenesisHome page
A. Guweidhi, J. Kleeff, N. Giese, J. E. Fitori, K. Ketterer, T. Giese, M. W. Buchler, M. Korc, and H. Friess
Enhanced expression of 14-3-3sigma in pancreatic cancer and its role in cell cycle regulation and apoptosis
Carcinogenesis, September 1, 2004; 25(9): 1575 - 1585.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
L.-C. Chen, C.-Y. Hao, Y. S. Y. Chiu, P. Wong, J. S. Melnick, M. Brotman, J. Moretto, F. Mendes, A. P. Smith, J. L. Bennington, et al.
Alteration of Gene Expression in Normal-Appearing Colon Mucosa of APCmin Mice and Human Cancer Patients
Cancer Res., May 15, 2004; 64(10): 3694 - 3700.
[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 Email this article to a friend
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
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by TÜRECI, O.
Right arrow Articles by HAMMER, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by TÜRECI, O.
Right arrow Articles by HAMMER, J.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS