| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Short Communications |
From the Departments of Pathology,*
Statistics,
Internal
Medicine,
Pediatrics,
Biostatistics,¶
Epidemiology,||
and
Surgery,**
The University of Michigan, Ann Arbor,
Michigan; and the National Cancer Control
Center,
Clalit Health Services
and Technion University, Haifa University, Haifa, Israel
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Another major anticipated benefit of these technologies is the establishment of organ- and tumor-specific profiles that, among other potential benefits, might assist with the diagnostic work-up of patients with metastatic cancer of unknown origin at the time of initial diagnosis. A comprehensive library of unique gene expression profiles of all of the major tumor types would permit a definitive diagnosis in the absence of pertinent clinical history, imaging studies, and/or surgical exploration, thus simplifying the diagnostic evaluation. For example, comparing the gene expression profile of a patient’s brain lesion to the gene expression library may be sufficient to establish a diagnosis of primary lung adenocarcinoma in the absence of thoracic imaging studies. Furthermore, these profiles might assist in the diagnosis of histologically similar primary tumor types, such as in distinguishing poorly differentiated lung carcinoma from malignant mesothelioma. However, at this point, it has not been established if it will be possible to define gene expression profiles that will discriminate the major tumor types in an organ-specific manner. A crucial step in establishing the diagnostic relevance of gene expression profiles is to compare profiles of histologically similar tumors, such as adenocarcinomas, from different organs. In this study, we compared the gene expression profiles of 154 primary adenocarcinomas of lung, colon, and ovary and demonstrated these profiles could discriminate the tumors in an organ-specific manner. In addition, we identified genes that are potentially useful as diagnostic markers for these tumors.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The primary tumors analyzed in this study were derived from several sources. The lung adenocarcinomas were procured from the University of Michigan Health System between 1994 and 1999. The colon adenocarcinomas were procured from five Israeli hospitals as part of the Molecular Epidemiology of Colorectal Cancer Study, a collaborative project between the University of Michigan and the National Cancer Control Center, Haifa, Israel (NIH CA81488). The ovarian tumors were procured from several sources, including the University of Michigan Health System, the Cooperative Human Tissue Network, and Cornell New York Hospital. All procedures were approved by the University of Michigan Institutional Review Board (IRB-Medicine).
All tumors were processed in a similar manner. Frozen tumor samples were embedded in OCT freezing media (Miles Scientific, Naperville, IL), cryotome sectioned (5 µm), and evaluated by routine hematoxylin and eosin (H&E) stains by one of three surgical pathologists. Whenever possible, the corresponding H&E sections from paraffin blocks were also evaluated. Areas of relatively pure tumor (at least 70% tumor cells) were selected for RNA isolation. All grades of differentiation were exhibited by the tumors.
RNA Isolation
Single isolates of tumor samples were homogenized in the presence of Trizol reagent (Life Technologies, Gaithersburg, MD) and total cellular RNA was purified according to manufacturer’s procedures. RNA samples were further purified using RNeasy spin columns (Qiagen, Valencia, CA) and used to prepare cRNA probes. RNA quality of the lung and ovary tumors was assessed by 1% agarose gel electrophoresis in the presence of ethidium bromide. Samples that did not reveal intact and approximately equal 18S and 28S ribosomal bands were excluded from further study (5% of the lung and 17% of the ovary cases).
cRNA Synthesis and Gene Expression Profiling
This study used commercially available high-density microarrays (Affymetrix, Santa Clara, CA) that produce gene expression levels on 7129 known genes and expressed sequence tags (HuGeneFL Array). Preparation of cRNA, hybridization, and scanning of the arrays were performed according to manufacturer’s protocols. Briefly, 5 µg of total RNA was used to generate double-stranded cDNA by reverse transcription using a cDNA synthesis kit (Superscript Choice System; Life Technologies, Inc., Rockville, MD) that uses an oligo(dT)24 primer containing a T7 RNA polymerase promoter 3' to the poly T (Geneset, La Jolla, CA), followed by second-strand synthesis. Labeled cRNA was prepared from the double-stranded cDNA by in vitro transcription by T7 RNA polymerase in the presence of biotin-11-CTP and biotin-16-UTP (Enzo, Farmington, NY). The labeled cRNA was purified over RNeasy columns. Fifteen µg of cRNA was fragmented at 94°C for 35 minutes in 40 mmol/L of Tris-acetate, pH 8.1, 100 mmol/L of potassium acetate, and 30 mmol/L of magnesium acetate. The cRNA was then used to prepare 300 µl of hybridization cocktail (100 mmol/L MES, 1 mol/L NaCl, 20 mmol/L ethylenediaminetetraacetic acid, 0.01% Tween 20) containing 0.1 mg/ml of herring sperm DNA (Promega, Madison, WI) and 500 µg/ml of acetylated bovine serum albumin (Life Technologies, Inc.). Before hybridization, the cocktails were heated to 94°C for 5 minutes, equilibrated at 45°C for 5 minutes, and then clarified by centrifugation (16,000 x g) at room temperature for 5 minutes. Aliquots of this hybridization cocktail containing 10 µg of fragmented cRNA were hybridized to HuGeneFL arrays at 45°C for 16 hours in a rotisserie oven at 60 rpm. The arrays were washed using nonstringent buffer (6x SSPE) at 25°C, followed by stringent buffer (100 mmol/L MES, pH 6.7, 0.1 mol/L NaCl, 0.01% Tween 20) at 50°C. The arrays were stained with streptavidin-phycoerythrin (Molecular Probes, Eugene, OR), washed with 6x sodium chloride, sodium phosphate, EDTA (SSPE buffer), incubated with biotinylated anti-streptavidin IgG, stained again with streptavidin-phycoerythrin, and washed again with 6x SSPE. The arrays were scanned using the GeneArray scanner (Affymetrix). Image analysis was performed with GeneChip software (Affymetrix).
Statistical Analysis
The HuGeneFL chip consists of 7129 probe sets, each representing a transcript. Each probe set typically consists of 20 perfectly complementary 25 base long probes as well as 20 mismatch probes that are identical except for an altered central base. We subtract the mismatch probe values from the perfect match values and average the middle 50% of these differences as the expression measure for that probe set.
A quantile normalization procedure was used to adjust for differences in the probe intensity distribution across different chips. We applied a monotone linear spline to each chip that mapped quantiles 0.02 up to 0.98 (in increments of 0.02) exactly to the corresponding median quantiles for all 154 samples. Then, the transform log(100 + max(X + 100; 0)) was applied to the data from each chip.
We built a classifier out of our 154 training samples as follows. We selected N markers from each of the three tumor classes, giving 3N markers in all (a range of values for N was considered, as discussed in the Results section). To classify a new sample of unknown tumor type, we compute the correlation coefficient between the 3N markers on the unidentified sample and the same markers on each of the 154 training samples. The class identities of the five training samples having the greatest correlation with the unclassified sample are then considered. If three or more of these samples belong to a common class, then this is the predicted class for the unclassified sample. Otherwise, the prediction is considered to be indeterminate. This strategy is known in the classification literature as "five-nearest neighbors with majority voting."3
We used a cross-validation procedure to estimate the error rate of our classifier. In the generic procedure, we set aside a single validation sample, leaving 153 samples to train a classifier, as described above. Note that compared to the classifier that would be used in practice, this new classifier will use a slightly different set of markers as well as having one fewer training sample. Moreover, this classifier does not have access to the expression values or tissue type of the single held-out sample. We then use this classifier to predict the class of the held-out sample, and record whether this prediction is correct. This process is repeated 154 times, with each sample being held out exactly once. The aggregate error rate across the 154 predictions is used as an estimate for the error rate that would be expected to occur in practice.
Immunohistochemistry
Routine immunohistochemistry was performed using formalin-fixed,
paraffin-embedded sections using the avidin-biotin complex
method.4
The following antibodies, dilutions, and
pretreatment conditions were used: anti-keratin (CAM 5.2, 1:10, trypsin
pretreatment; Becton-Dickinson, San Jose, CA), anti-human epithelial
keratins (AE1:AE3, 1:800, no pretreatment; Roche Diagnostics,
Indianapolis, IN), anti-human melanoma (HMB45, 1:25, no pretreatment;
DAKO, Carpinteria, CA), anti-cow S-100 (1:500, no pretreatment; DAKO),
anti-vimentin (1:800, no pretreatment; DAKO), anti-human cytokeratin 20
(CK20) (1:25, DAKO Protease 1 pretreatment; DAKO), anti-human CK7
(1:25, DAKO Protease 2 pretreatment; DAKO), anti-CEA (monoclonal D-14,
1:8, DAKO Protease 1 pretreatment: E-Z-EM, Westbury, NY), anti-
smooth muscle actin (1:1600, no pretreatment; Sigma Chemical Co., St.
Louis, MO), anti-human muscle actin (HHF35, 1:100, DAKO Protease 2
pretreatment; DAKO) and anti-c-KIT (1:100, citrate buffer pretreatment;
DAKO).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Comprehensive gene expression profiles of 57 lung, 51 colon, and
46 ovary primary carcinomas were generated using high-density
oligonucleotide arrays with 7129 probe sets, which in total
interrogated some 6800 genes. To provide a visual assessment of
relationships between the tumors based on gene expression, we
considered each sample to be represented by a point in a 7129
multidimensional space, with each coordinate given by a gene expression
level. Several views of this set of points were generated using
principal component analysis (PCA), which locates the two-dimensional
views that capture the greatest amount of variability in the data. We
note that these views were based solely on aggregate expression
variation and no references to the tissue classifications were made. We
generated four views in all, by stratifying the genes into four
quarters of equal size, and then applying PCA to the measurements in
each quarter separately (Figure 1)
. The
strata were formed based on the average expression level of each
transcript across the 154 samples; the first stratum contained the 25%
of transcripts with the least average abundance, the second stratum
contained the next 25% of the genes, and so on.
|
indicated substantial
differences in gene expression between the three tumor types throughout
the range of expression measures. Thus, not only did genes with high
average expression levels allow the three tumor types to be
distinguished from each other (Figure 1
, top quarter), but genes that
were expressed at low average levels were informative as well (Figure 1
, first quarter). Particularly notable was the wide margin that
separated lung and colon tumors (Figure 1
, top quarter). It was also
evident that substantial heterogeneity occurred within each tumor type,
with the ovarian tumors showing the greatest heterogeneity, and the
colon tumors showing the least. Much of the substructure in the various
panels of Figure 1
was attributed to identified differences in tumor
histopathology. The ovarian tumors could be divided into two subsets. A
minor subset highlighted in Figure 1
(top quarter) clustered separately
from the major ovarian set, overlapped with colon tumors in the second
and third quarterlies (Figure 1)
, and consisted exclusively of mucinous
ovarian tumors. However, some mucinous tumors also clustered with the
major group. The colon tumor that was substantially separated from the
remainder was a sarcoma (see below), in contrast to the remaining colon
tumors, all of which were adenocarcinomas. The ovarian tumor most
completely embedded within the colon cohort was subsequently determined
to have a number of features (see below) indicating it was a colonic
adenocarcinoma metastatic to the ovary.
Next we applied a gene selection procedure to identify a set of markers
for each tumor type (lung, colon, ovary). Specifically, we sought to
identify the genes that had much greater expression in one tumor type
compared to either of the other two types. The difference between the
average expression of each gene within a given tumor type and the
larger of the two average expression values for the same gene in the
other two types was computed. Because the data were log-transformed,
this value can be interpreted as the logarithm of the ratio between the
geometric mean expression in the more highly expressing of the other
two types. The result is that for each tumor class, we obtained a
ranking of the genes, with the genes having high rank being the
strongest markers for the tumor type. The top 20 markers for each type
are shown in Table 1
. Three of the 60
genes were identified twice and one three times, as they are
represented more than once in the arrays with distinct probe sets.
|
We then built a classifier out of our 154 tumor samples as described in Materials and Methods. Values of N = 3, 5, 7, 10, 15, and 20 for each tumor type were considered for the number of markers per tumor class that were made available to the classifier. Using fewer than seven markers (six unique genes) led to a degradation in performance, whereas more than 10 markers did not provide any improvement. When 10 markers (nine unique colon and ovary genes and seven unique lung genes) were considered, 152 of 154 samples were correctly classified. This is considered to be the best possible result, because the probable colonic metastasis diagnosed as an ovarian primary represented an apparent erroneous diagnosis and the colonic sarcoma could not have been correctly classified as it represented the only sarcoma in the study.
Additionally, to provide another visual assessment of relationships
between the tumors based on the most differentially expressed genes, we
used PCA as above using the gene expression data for 60 genes (55
unique), the top 20 from each tumor type identified by the classifier.
The PCA view derived from these genes is shown (Figure 2)
and was very similar to the view
derived from the top quarter of expressed genes (Figure 1)
, but showed
more distinct separation of the tumor types. The two absolute outlier
tumors identified by the classifier were clearly seen as outliers and
the subset of mucinous ovarian tumors identified previously were again
ascertained as colon-like.
|
The two absolute outlier tumors, one colonic and one ovarian, were
further investigated by routine immunohistochemical methods to further
define their histopathological classification. This approach is
typically used by practicing surgical pathologists to determine the
likely origin of tumors of uncertain primary site. The colonic tumor, a
pleomorphic spindle-cell neoplasm on routine H&E (Figure 3A)
, was correctly diagnosed at the
contributing hospital as a nonepithelial neoplasm, probably
sarcoma, but was included in the colon cancer cohort because all of
these tumors were derived from a population-based study of incident,
invasive colorectal cancers, almost all of which are
adenocarcinomas. Immunostains for low- and high-molecular weight
cytokeratins, CK7 and CK20, S-100, HMB-45, vimentin, muscle-specific
actin, and smooth muscle actin showed the neoplastic cells to be
negative for cytokeratins (Figure 3B)
, S-100 (not shown), and HMB-45
(not shown), strongly positive for vimentin (Figure 3C)
, and focally
positive for both actins (not shown). This immunohistochemical profile,
together with the histopathology, is diagnostic of high-grade
leiomyosarcoma. Gastrointestinal stromal tumor was not an appropriate
diagnosis based on histopathology, a negative KIT immunostain (not
shown), and no detectable expression of the c-kit gene on
the array (not shown).
|
. Review of the microarray
expression data for CK7 and CK20 and CEA showed low levels of
expression for CK7 and high levels of expression for CK20, in sharp
contrast to other tumors in the ovarian cohort (high CK7 and low CK20).
The CEA data were not informative. Immunohistochemical stains for CK7,
CK20, and CEA, performed to validate the expression data and to
investigate the possibility that this tumor was metastatic to ovary,
showed strong and diffuse tumor immunoreactivity for CK20 (Figure 3E)
,
whereas CK7 showed no immunoreactivity (Figure 3F)
. CEA showed strong
and diffuse immunoreactivity of the tumor cells and associated mucin
(Figure 3G)
. These results offered strong support for the view that
this tumor was a colonic adenocarcinoma metastatic to ovary rather than
a primary ovarian adenocarcinoma. Classification Using Known Diagnostic Markers
In practice, it is possible to discriminate most colon, ovary, and lung adenocarcinomas with histopathology and a limited immunohistochemical profile that includes markers CK7, CK20, and thyroid transcription factor (TTF)-1, in which colon tumors are CK7 (-), CK20 (+), and TTF-1 (-), ovary tumors are CK7 (+), CK20 (-), and TTF-1 (-), and lung tumors are CK7 (+), CK20 (-), and TTF-1 (+). In our cohort of tumors, using the array expression data for these three markers and using the same statistical approach, these markers correctly classified 138 of 154 samples, with 152 of 152 being the best possible outcome (see above).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A previous proof of principle gene expression-profiling study reported molecular discrimination of acute myelogenous leukemia from acute lymphoblastic leukemia.5 Although these leukemias can be separated on morphological grounds in the majority of cases, there are tumors that require adjuvant methodologies, such as flow cytometry, for accurate diagnosis. Similarly, the current study compares gene expression profiles of histologically similar tumors, ie, adenocarcinomas, from three common sites that usually can be morphologically separated but sometimes pose diagnostic difficulties. For instance, the histopathological separation of primary ovarian carcinomas from metastases from a gastrointestinal source is a common diagnostic dilemma, as is illustrated by the outlier ovarian tumor in this study. This tumor, diagnosed as primary ovarian mucinous adenocarcinoma, was shown to be of probable colonic origin by immunohistochemical evaluation of low-molecular weight cytokeratin proteins 7 and 20 and CEA, markers with demonstrated diagnostic utility.6 The ability of the expression profiles to identify this tumor as not belonging to the ovarian cohort further validates this approach. Thus, it would be interesting to determine the utility of this approach to the evaluation of mucinous tumors of the abdominal cavity of unknown origin.
The other nonconforming case, a primary invasive colorectal cancer, was
similarly investigated and shown to be a colonic sarcoma by
histopathology and immunohistochemical profile. This case also provides
further validation of the gene expression profiles. Comparison of the
expression data from the sarcoma to the colonic adenocarcinoma cohort
illustrates some of the differences in specific gene expression.
Examples of such differentially expressed genes include vimentin and
enteric smooth muscle-
2 actin, genes previously shown to be
expressed in sarcomas and useful diagnostic immunohistochemical
markers,7,8
and other genes not previously shown to be
expressed in sarcomas (for example, C-type lectin superfamily member 2;
CLECSF2).
Many of the potential differentially expressed genes identified by this method have previously been shown to be either diagnostically useful and/or expressed in an organ-specific manner. TTF-1 is a nuclear protein expressed in pulmonary and thyroid epithelium,9 plays a role in lung and thyroid development,10 and is an immunohistochemical marker used to distinguish primary from metastatic lung adenocarcinoma.11,12 TTF-1, as well as several surfactant-related genes, was identified as one of the genes whose expression is primarily restricted to lung adenocarcinomas. Similarly, the CDX1 and CDX2 genes, which encode intestine-specific transcription factors,13,14 were identified as preferentially expressed in the colon tumors, as was CK20, a known marker of colonic adenocarcinoma.6 Finally, the estrogen receptor 1 gene, known as a marker of breast and gynecological malignancies,15 was identified as preferentially expressed in the ovarian tumors. The independent identification by this study of these known organ-specific markers provides strong validation of the utility of gene expression profiling as a highly effective gene discovery tool.
Using expression data for three genes commonly used as immunohistochemical markers in clinical practice, the molecular approach was able to correctly classify the large majority (91%) of tumors. However, the success of this relatively limited diagnostic work-up when compared with the analysis of thousands of genes by microarrays should not diminish the merits of global gene expression profiling. As molecular profiling is extended to a larger number of tumor types, it will be necessary to use additional numbers of genes to define organ-specific profiles. Furthermore, as shown by this study and others,1,2,16 this approach is quite fruitful as a discovery tool to identify additional diagnostic markers.
The implications of this study are broad, suggesting that the establishment of gene expression profiles can be used to classify neoplasms in an organ-specific manner, one of the charges of the National Cancer Institute’s "Director’s Challenge" program for the molecular classification of cancer. The success of this study of three common tumor types suggests it will be feasible to extend this approach to a comprehensive cohort of tumors. The availability of such a comprehensive cancer map will profoundly impact clinical cancer care through improved diagnosis, prognosis, and treatment.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Supported by funds from the National Cancer Institute (NIH CA84952 and CA81488) and in part by the Tissue Core of the University of Michigan Comprehensive Cancer Center (NIH CA46952).
Accepted for publication July 6, 2001.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  Y. Zhai, R. Kuick, B. Nan, I. Ota, S. J. Weiss, C. L. Trimble, E. R. Fearon, and K. R. Cho Gene Expression Analysis of Preinvasive and Invasive Cervical Squamous Cell Carcinomas Identifies HOXC10 as a Key Mediator of Invasion Cancer Res., November 1, 2007; 67(21): 10163 - 10172. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Vachani, M. Nebozhyn, S. Singhal, L. Alila, E. Wakeam, R. Muschel, C. A. Powell, P. Gaffney, B. Singh, M. S. Brose, et al. A 10-Gene Classifier for Distinguishing Head and Neck Squamous Cell Carcinoma and Lung Squamous Cell Carcinoma Clin. Cancer Res., May 15, 2007; 13(10): 2905 - 2915. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. R. Schwartz, K. Moin, B. Yao, L. M. Matrisian, L. M. Coussens, T. H. Bugge, B. Fingleton, H. B. Acuff, M. Sinnamon, H. Nassar, et al. Hu/Mu ProtIn Oligonucleotide Microarray: Dual-Species Array for Profiling Protease and Protease Inhibitor Gene Expression in Tumors and Their Microenvironment Mol. Cancer Res., May 1, 2007; 5(5): 443 - 454. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. A. Irizarry, Z. Wu, and H. A. Jaffee Comparison of Affymetrix GeneChip expression measures Bioinformatics, April 1, 2006; 22(7): 789 - 794. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. J. Giordano, A. Y.M. Au, R. Kuick, D. G. Thomas, D. R. Rhodes, K. G. Wilhelm Jr., M. Vinco, D. E. Misek, D. Sanders, Z. Zhu, et al. Delineation, Functional Validation, and Bioinformatic Evaluation of Gene Expression in Thyroid Follicular Carcinomas with the PAX8-PPARG Translocation. Clin. Cancer Res., April 1, 2006; 12(7): 1983 - 1993. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Wang, R. Wu, K. R. Cho, K. A. Shedden, T. J. Barder, and D. M. Lubman Classification of Cancer Cell Lines Using an Automated Two-dimensional Liquid Mapping Method with Hierarchical Clustering Techniques Mol. Cell. Proteomics, January 1, 2006; 5(1): 43 - 52. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. E. Wiggins, M. Goyal, S. K. Sanden, B. L. Wharram, K. A. Shedden, D. E. Misek, R. D. Kuick, and R. C. Wiggins Podocyte Hypertrophy, "Adaptation," and "Decompensation" Associated with Glomerular Enlargement and Glomerulosclerosis in the Aging Rat: Prevention by Calorie Restriction J. Am. Soc. Nephrol., October 1, 2005; 16(10): 2953 - 2966. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Chen, M. S. Bhojani, A. C. Heaford, D. C. Chang, B. Laxman, D. G. Thomas, L. B. Griffin, J. Yu, J. M. Coppola, T. J. Giordano, et al. Phosphorylated FADD induces NF-{kappa}B, perturbs cell cycle, and is associated with poor outcome in lung adenocarcinomas PNAS, August 30, 2005; 102(35): 12507 - 12512. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. S. Rozek, S. M. Lipkin, E. R. Fearon, S. Hanash, T. J. Giordano, J. K. Greenson, R. Kuick, D. E. Misek, J. M.G. Taylor, J. A. Douglas, et al. CDX2 Polymorphisms, RNA Expression, and Risk of Colorectal Cancer Cancer Res., July 1, 2005; 65(13): 5488 - 5492. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. W. Tothill, A. Kowalczyk, D. Rischin, A. Bousioutas, I. Haviv, R. K. van Laar, P. M. Waring, J. Zalcberg, R. Ward, A. V. Biankin, et al. An Expression-Based Site of Origin Diagnostic Method Designed for Clinical Application to Cancer of Unknown Origin Cancer Res., May 15, 2005; 65(10): 4031 - 4040. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. A. Shedden, M. P. Kshirsagar, D. R. Schwartz, R. Wu, H. Yu, D. E. Misek, S. Hanash, H. Katabuchi, L. H. Ellenson, E. R. Fearon, et al. Histologic Type, Organ of Origin, and Wnt Pathway Status: Effect on Gene Expression in Ovarian and Uterine Carcinomas Clin. Cancer Res., March 15, 2005; 11(6): 2123 - 2131. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. A. Granville and P. A. Dennis An Overview of Lung Cancer Genomics and Proteomics Am. J. Respir. Cell Mol. Biol., March 1, 2005; 32(3): 169 - 176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
|
|
Z. Zhang, R. C. Bast Jr., Y. Yu, J. Li, L. J. Sokoll, A. J. Rai, J. M. Rosenzweig, B. Cameron, Y. Y. Wang, X.-Y. Meng, et al. Three Biomarkers Identified from Serum Proteomic Analysis for the Detection of Early Stage Ovarian Cancer Cancer Res., August 15, 2004; 64(16): 5882 - 5890. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S.-H. Hong, D. E. Misek, H. Wang, E. Puravs, T. J. Giordano, J. K. Greenson, D. E. Brenner, D. M. Simeone, C. D. Logsdon, and S. M. Hanash An Autoantibody-Mediated Immune Response to Calreticulin Isoforms in Pancreatic Cancer Cancer Res., August 1, 2004; 64(15): 5504 - 5510. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Hibbs, K. M. Skubitz, S. E. Pambuccian, R. C. Casey, K. M. Burleson, T. R. Oegema Jr, J. J. Thiele, S. M. Grindle, R. L. Bliss, and A. P.N. Skubitz Differential Gene Expression in Ovarian Carcinoma: Identification of Potential Biomarkers Am. J. Pathol., August 1, 2004; 165(2): 397 - 414. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. M. Mintzer, M. Warhol, A.-M. Martin, and G. Greene Cancer of Unknown Primary: Changing Approaches. A Multidisciplinary Case Presentation from the Joan Karnell Cancer Center of Pennsylvania Hospital Oncologist, June 1, 2004; 9(3): 330 - 338. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
|
|
R. D. Petty, M. C. Nicolson, K. M. Kerr, E. Collie-Duguid, and G. I. Murray Gene Expression Profiling in Non-Small Cell Lung Cancer: From Molecular Mechanisms to Clinical Application Clin. Cancer Res., May 15, 2004; 10(10): 3237 - 3248. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. E. Schmalbach, D. B. Chepeha, T. J. Giordano, M. A. Rubin, T. N. Teknos, C. R. Bradford, G. T. Wolf, R. Kuick, D. E. Misek, D. K. Trask, et al. Molecular Profiling and the Identification of Genes Associated With Metastatic Oral Cavity/Pharynx Squamous Cell Carcinoma Arch Otolaryngol Head Neck Surg, March 1, 2004; 130(3): 295 - 302. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. H. Lim, A. Aggarwal, T. Agasthian, P.-S. Wong, C. Tan, E. Sim, L. Tan, P.-S. Goh, S.-C. Wang, K.-L. Khoo, et al. Feasibility of Using Low-Volume Tissue Samples for Gene Expression Profiling of Advanced Non-Small Cell Lung Cancers Clin. Cancer Res., December 1, 2003; 9(16): 5980 - 5987. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Bardelli, S. Saha, J. A. Sager, K. E. Romans, B. Xin, S. D. Markowitz, C. Lengauer, V. E. Velculescu, K. W. Kinzler, and B. Vogelstein PRL-3 Expression in Metastatic Cancers Clin. Cancer Res., November 15, 2003; 9(15): 5607 - 5615. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. A. Shedden, J. M. G. Taylor, T. J. Giordano, R. Kuick, D. E. Misek, G. Rennert, D. R. Schwartz, S. B. Gruber, C. Logsdon, D. Simeone, et al. Accurate Molecular Classification of Human Cancers Based on Gene Expression Using a Simple Classifier with a Pathological Tree-Based Framework Am. J. Pathol., November 1, 2003; 163(5): 1985 - 1995. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Nishizuka, S.-T. Chen, F. G. Gwadry, J. Alexander, S. M. Major, U. Scherf, W. C. Reinhold, M. Waltham, L. Charboneau, L. Young, et al. Diagnostic Markers That Distinguish Colon and Ovarian Adenocarcinomas: Identification by Genomic, Proteomic, and Tissue Array Profiling Cancer Res., September 1, 2003; 63(17): 5243 - 5250. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. A. Powell, A. Spira, A. Derti, C. DeLisi, G. Liu, A. Borczuk, S. Busch, S. Sahasrabudhe, Y. Chen, D. Sugarbaker, et al. Gene Expression in Lung Adenocarcinomas of Smokers and Nonsmokers Am. J. Respir. Cell Mol. Biol., August 1, 2003; 29(2): 157 - 162. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. K. Vanaja, J. C. Cheville, S. J. Iturria, and C. Y. F. Young Transcriptional Silencing of Zinc Finger Protein 185 Identified by Expression Profiling Is Associated with Prostate Cancer Progression Cancer Res., July 15, 2003; 63(14): 3877 - 3882. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. T. Miller, S. Aggarwal, T. K. Lin, S. L. Dagenais, J. I. Contreras, M. B. Orringer, T. W. Glover, D. G. Beer, and L. Lin Amplification and Overexpression of the Dual-Specificity Tyrosine-(Y)-Phosphorylation Regulated Kinase 2 (DYRK2) Gene in Esophageal and Lung Adenocarcinomas Cancer Res., July 15, 2003; 63(14): 4136 - 4143. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. R. Schwartz, R. Wu, S. L. R. Kardia, A. M. Levin, C.-C. Huang, K. A. Shedden, R. Kuick, D. E. Misek, S. M. Hanash, J. M. G. Taylor, et al. Novel Candidate Targets of {beta}-Catenin/T-cell Factor Signaling Identified by Gene Expression Profiling of Ovarian Endometrioid Adenocarcinomas Cancer Res., June 1, 2003; 63(11): 2913 - 2922. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. D. Logsdon, D. M. Simeone, C. Binkley, T. Arumugam, J. K. Greenson, T. J. Giordano, D. E. Misek, and S. Hanash Molecular Profiling of Pancreatic Adenocarcinoma and Chronic Pancreatitis Identifies Multiple Genes Differentially Regulated in Pancreatic Cancer Cancer Res., May 15, 2003; 63(10): 2649 - 2657. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. L. Neuhouser, R. E. Patterson, M. D. Thornquist, G. S. Omenn, I. B. King, and G. E. Goodman Fruits and Vegetables Are Associated with Lower Lung Cancer Risk Only in the Placebo Arm of the {beta}-Carotene and Retinol Efficacy Trial (CARET) Cancer Epidemiol. Biomarkers Prev., April 1, 2003; 12(4): 350 - 358. [Abstract] [Full Text] [PDF]
|
|
|
|
