Cancer Research Grants Frontiers in Basic Cancer Research
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
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
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 Chen, Y.
Right arrow Articles by Schlegel, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chen, Y.
Right arrow Articles by Schlegel, R.
[Cancer Research 63, 1927-1935, April 15, 2003]
© 2003 American Association for Cancer Research

Tumor Biology

Identification of Cervical Cancer Markers by cDNA and Tissue Microarrays

Yan Chen, Christine Miller, Rebecca Mosher, Xumei Zhao, Jim Deeds, Mike Morrissey, Barb Bryant, David Yang, Ron Meyer, Frank Cronin, Bobbie S. Gostout, Karen Smith-McCune and Robert Schlegel1

Departments of Molecular and Cell Biology [Y. C., C. M., R. Mo., X. Z., J. D., D. Y., R. Me., F. C., R. S.] and Applied Bioinformatics [M. M., B. B.], Millennium Pharmaceuticals, Inc., Cambridge, Massachusetts 02139; Gynecologic Surgery, Mayo Clinic, Rochester, Minnesota 55905 [B. S. G.]; and Department of Obstetrics, Gynecology and Reproductive Sciences and Cancer Research Institute, University of California at San Francisco Comprehensive Cancer Center, University of California at San Francisco, San Francisco, California 94143-0128 [K. S-M.]


   ABSTRACT
Top
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The Pap test has effectively reduced the incidence and mortality of cervical cancer. However, because of the morphological basis of this test, sensitivity and specificity are less than ideal, a situation that complicates the clinical management of women diagnosed with low-grade cervical abnormalities. In an attempt to understand the molecular basis of cervical tumorigenesis and to discover molecular markers for accurate cervical cancer screening, we used cDNA microarrays containing >30,000 Unigene clones to examine the gene expression patterns of 34 cervical tissues from different clinically defined stages. It was found that global gene expression patterns separated normal cervical tissues and low-grade squamous intraepithelial lesions from cervical cancers and most of the high-grade squamous intraepithelial lesions (HSILs). Among the top 62 genes/(expressed sequence tags) that were overexpressed in tumors and HSIL tissues, 35 were confirmed using in situ hybridization on cervical tissue micorarrays. Many of these genes were overexpressed in high-grade dysplastic and malignant cervical epithelium or in stroma adjacent to the diseased tissues, with cellular proliferation and extracellular matrix-associated genes being the most common. In general, the extent of gene overexpression increased as the lesions progressed from low-grade squamous intraepithelial lesions to HSILs and finally to cancer. It is hoped that with additional development, some of these markers will improve the interpretation of cervical screening tests and provide useful information for patient management decisions.


   INTRODUCTION
Top
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The Pap test is considered to be the most cost-effective cancer screening test developed to date (1) . It has dramatically decreased the incidence and mortality rates of cervical cancer by >70% since it was introduced in the United States and many other countries of the world (2) . The abnormal morphological changes of Pap tests described by The Bethesda System include ASCUS,2 AGUS, LSIL, HSIL, SCC and ACA (3) . The success of Pap tests is attributed mostly to the diagnosis and treatment of precancerous lesions.

Currently, management of patients with HSIL and more advanced diseases is relatively standard. Most women with such lesions undergo colposcopy and appropriately directed biopsies. If the histological diagnosis is confirmed, ablative or excisional treatment such as electrosurgical loop excision procedure, cryosurgery, or conization is performed. However, management of ambiguous or low-grade cytological results (ASCUS and LSIL) is very controversial. This is mainly because of the nature of this morphology-based test, which inevitably leads to interobserver variability and some Pap test discordance with histological follow-up. It was reported that the mean sensitivity of primary Pap tests is ~58%, and the accuracy of a repeat test is only ~66% (4) . The low sensitivity and poor reproducibility have complicated the management of ASCUS and LSIL patients. If an accelerated repeat Pap test is recommended for the follow-up of women with primary diagnosis of ASCUS or LSIL, patients will risk delay in diagnosis of potential high-grade lesions. However, if these patients are universally referred to colposcopy, the vast majority of women will be over treated. Only 5–10% of women with ASCUS have high-grade disease upon colposcopy, and >80% of LSIL will regress to normal or stay in their current state (5 , 6) .

AGUS represents a much greater risk than ASCUS or LSIL because cytology is less sensitive for this condition, and the disease progresses more rapidly (7) . It was found that 9–54% of women with AGUS have biopsy-confirmed cervical intraepithelial neoplasias, 0–8% have biopsy-confirmed ACA in situ and <1–9% have invasive carcinoma (8) . Because of the greater risk, all patients with AGUS are referred to colposcopy (8) .

The subjectivity of cervical cytology could be reduced by objective markers that determine the presence and severity of dysplastic cells. Because high-risk HPV infection is strongly associated with cervical cancer development (9) , HPV testing using methods like Hybrid Capture II (Digene Diagnostics, Silver Spring, MD) or PCR appears to provide an objective measurement (10) . However, because the vast majority of HPV infections and the resulting squamous intraepithelial lesions regress spontaneously, especially in young women, HPV testing cannot specifically identify patients whose lesions will persist or progress to invasive carcinoma (11 , 12) . As reported in the ASCUS-LSIL Triage Study, 83% of woman with LSIL Pap results test positive for high-risk HPV types, a level too high to be useful for triage (13) . Although triage using HPV testing significantly improved the sensitivity for detecting HSIL in women with ASCUS Pap results, the specificity was comparable with using conventional cytology (14) . A more desirable cervical screening marker would identify all cervical cancers, the majority of HSIL, and the small percentage of true precancers among patients with LSIL and ASCUS on Pap.

It is now well accepted that cervical carcinogenesis occurs in a stepwise fashion (15) . The transition of normal epithelium to preneoplastic lesions and invasive carcinoma occurs sequentially. The morphologically defined steps of dysplastic and malignant abnormalities are a reflection of cellular gene alterations during tumorigenesis. In this study, we used cDNA microarray and ISH technologies to identify and evaluate 62 genes/ESTs that are differentially up-regulated in premalignant cervical lesions and invasive cancers. Many of these genes are expressed in cervical epithelium or in stroma adjacent to dysplastic/malignant tissues, with cellular proliferation and ECM-associated genes as the most common. Identification of such biomarkers may help improve the specificity of cervical cancer screening.


   MATERIALS AND METHODS
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Sample Collection and RNA Preparation.
Cervical tissues were collected and snap frozen in liquid nitrogen at the Department of Laboratory Medicine and Pathology of the Mayo Clinic and Foundation, the Department of Pathology and Laboratory Medicine of Albany Medical Center, and the Department of Obstetrics, Gynecology and Reproductive Sciences and Cancer Research Institute of the University of California at San Francisco. The histology and cellular composition of tissues were confirmed by the scientists at Millennium Pharmaceuticals, Inc. (Cambridge, MA) before RNA extraction was performed. Total RNA was extracted from the frozen tissues using Trizol Reagent (Life Technologies, Inc.) followed by a secondary clean-up step with Qiagen’s RNeasy kit to increase RNA probe labeling efficiency (Qiagen, Valencia, CA). Only RNA with a 28S/18S ribosomal RNA ratio of at least 1.0, calculated using Agilent Technologies 2100 Bioanalyzer (Palo Alto, CA), was used in this study.

cDNA Microarray Hybridization.
cDNA microarrays containing 30,732 Unigene clones from Research Genetics (Huntsville, AL) were generated on nylon filters at Millennium Pharmaceuticals, Inc. A total of 4–6 µg of total RNA was used as template to generate radioactively labeled cDNA by reverse transcription with 33P-dCTP, oligo dT-30 primer, and Superscript II Reverse Transcriptase (Life Technologies, Inc.). 33P-labeled first strand cDNA was preannealed with cot-1 DNA and poly-dA 40–60 (Pharmacia, Peapack, NJ) to reduce nonspecific hybridization. Each filter was hybridized at 65°C for 16 h with ~6 x 106 counts of labeled probe in a buffer containing 7% SDS, 250 mM Na3PO4 (pH 7.2), 1 mM EDTA, 0.5% Casein-Hammerstein, and 0.1 mg/ml of salmon sperm DNA. After the filters were washed with 4 and 1% SDS wash buffer [20 mM Na3PO4 (pH 7.2), 1 mM EDTA, and 4 or 1% SDS], they were exposed to Fuji Phosphoimager screens and scanned using a Fuji scanner BAS 2500. Spots were quantitated using an automated array analysis program, Grid Guru v1.0, developed at Millennium Pharmaceuticals, Inc.

Marker Scoring Algorithm and Statistical Analysis.
To correct for differences in hybridization efficiency, the digitized data from each microarray filter was normalized by the median intensity of all spots on that filter. Both array-based and gene-based hierarchical clustering were performed and visualized using Stanford’s Gene Cluster and Tree View software. Differentially expressed genes were ranked by calculating the Marker Score (defined below) for each gene.

To compute Marker Score, the samples were divided into control and tester groups. The starting point for the Marker Score is average fold change (ratio) of the tester samples above the control samples. We wanted to capture in this score both the degree of change (the expression ratio) and the number of tester samples showing differential expression. We did not want the score to be dominated by a small fraction of tester samples with very high values. To reduce this outlier effect, we chose to treat genes with expression ratios > 10 as not meaningfully different from those with ratios of 10. This desired performance from our Marker Score was accomplished by transforming the tester:control expression ratio using an asymptotic compression function before taking the average fold-change across tester samples. Our Marker Score has a value of 1 when the testers do not appear to be expressed more highly than the controls, and a value > 1 otherwise. A Marker Score cannot exceed a value of 10 for any gene.

The Marker Score Sg for gene g is therefore computed as the average of compressed tester:control ratios:


where C(r) is the compression function C(r) = A(1 - e-r/A) for r >= 1, and C(r) = 1 for r < 1; A is an upper asymptote on the fold-change value (we used 10); xgs is the expression value of gene g on sample s; xgQ is the Qth percentile of the control samples’ expression value; typically Q = 50; k is a constant reflecting the additive noise in the data, i.e., the fixed component of the variance in repeated measurements. We derived a value of 0.25 for this parameter from calibration experiments using our microarray technology (data not shown); and Ntester, the number of tester samples.

The marker scoring method was validated using a population permutation test (16) . In such a test, marker scores are calculated for all genes in the dataset, and the genes ranked by score. Next, the sample labels are randomly permuted (e.g., some tumor samples are moved into the normal group and vice versa) 100 times, and the marker scores recalculated and ranked for each permutation. A comparison is deemed significant (P = 0.05) if the scores of top markers using the actual sample labels exceeds the top scores generated by 95% of the permuted runs.

ISH of Tissue Microarrays.
Formalin-fixed, paraffin-embedded cervical tissue microarrays containing tissue cores from normal, LSILs, HSILs, SCCs, and ACAs were provided by Clinomics (Pittsfield, MA). Prehybridization treatment was performed with an automatic Tissue-Tek DRS 2000 Slide Stainer (Sakura, Torrance, CA) using a previously described protocol (17) . The cervical tissues were deparaffinized, rehydrated, and postfixed with 4% paraformaldehyde in PBS for 15 min. After washing with PBS, the tissue microarrays were digested with 2 µg/ml proteinase K at 37°C for 15 min and again incubated with 4% paraformaldehyde/PBS for 10 min. Tissue sections were subsequently incubated with 0.2 N HCL for 10 min, 0.25% acetic anhydride/0.1 M triethanolamine for 10 min, and dehydrated with graded ethanol. Antisense probes were labeled with 35S-UTP in an in vitro transcription reaction (Riboprobe Combination System; Promega, Madison, WI) using 500 ng of linearized plasmid DNA derived from IMAGE clones. Hybridizations were performed at 55°C for 18 h using probes labeled at 5 x 107 cpm/ml in 10 mM Tris-HCl (pH 7.6) buffer containing 50% formamide, 10% dextran sulfate, 1x Denhardt’s solution, 0.6 M NaCl, 10 mM DTT, 0.25% SDS, and 200 µg/ml tRNA. After hybridization, slides were washed with 5x SSC at 50°C for 10 min, 50% formamide/2x SSC at 50°C for 30 min, 10 mM Tris-HCl (pH 7.6)/500 mM NaCl/1 mM EDTA (TNE) at 37°C for 10 min, incubated in 10 µg/ml Rnase A in TNE at 37°C for 30 min, washed in TNE at 37°C for 10 min, incubated once in 2x SSC at 50°C for 20 min, twice in 0.2x SSC at 50°C for 20 min, and dehydrated with graded ethanol. Localization of mRNA transcripts was determined by dipping slides in Kodak NTB2 photoemulsion (Eastman Kodak, Rochester, NY) and exposing for 14–21 days at 4°C. The slides were counterstained using Myers hematoxylin and alcoholic eosin Y. ISH signal strength was scored manually under dark field microscopy using the following scale: 0, no signal above background; +, weak indeterminate signal; ++, weak to moderate determinate signal; and +++, strong to very strong signal.


   RESULTS
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Transcriptional Profiling of Cervical Tissues by cDNA Microarrays.
In an attempt to identify molecular markers that correlate with clinical stages of cervical abnormalities, we profiled 12 normal cervical tissues (9 from ectocervix and 3 from endocervix), 5 LSILs, 5 HSILs, 9 SCCs, and 3 ACAs on cDNA microarrays that contain 30,732 clones (designated as 30K microarray). To assess the power of the data sets to discriminate between diseased and normal tissue, a hierarchical clustering of the 34 sample data sets was performed on the basis of overall similarity in gene expression patterns (Fig. 1)Citation . The dendrogram shows that 10 of 12 normal cervical tissues and all LSIL samples cluster in one group (designated as control group) and 11 of 12 tumor samples and 3 of 5 HSIL samples cluster together in the other group (designated as diseased group). This segregation indicates that global gene expression profiles of normal ectocervical epithelium, normal endocervical epithelium, and LSIL are very similar, whereas the expression profiles of 3 of 5 HSIL samples more closely resemble cervical cancers. These findings indicate robust data sets that can distinguish control tissues from diseased tissues, even though the samples were taken from patients of different ages and from different clinical sites.



View larger version (26K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 1. Cluster diagram of cervical tissue samples. Dendrogram was created from hierarchical clustering of the transcriptional profiles of 34 normal, LSIL, HSIL, and cancerous cervical tissue samples. Each sample was labeled by its tissue type and an Id number. Necto, normal ectocervix; Nendo, normal endocervix; Tscc, squamous cell carcinoma; Taca, adenocarcinoma. — — — divides the 34 samples into two major groups: control group and diseased group. {circ} indicates incorrectly clustered samples.

 
Marker Selection.
To identify gene markers that would differentiate the control tissue group from the diseased group, marker scores were calculated for each clone on the 30K cDNA microarray from three marker selection paradigms: 9 SCCs versus control group (9 ectocervix, 3 endocervix, and 5 LSILs), 5 HSILs versus control group, and 3 ACAs versus control group. Whenever one evaluates the expression of many array elements simultaneously (in our case >30,000), there is a risk that the best performing genes could have arisen by chance alone. To assess this possibility and to evaluate the performance of the marker scoring algorithm, a permutation test was run for each of the disease versus control paradigms mentioned above. In this test, the clinical samples are randomly permuted 100 times, and the marker scores are recalculated and ranked for each permutation. High-quality markers should have scores that lie outside the 95th percentile of the permuted data. Fig. 2Citation shows the results of the permutation test for the SCC versus control comparison. Except for the marker ranked no. 1, all marker scores from actual data exceeded the 95th percentile of the permuted data, indicating that the scoring algorithm selected a marker set of high quality. Similar results were obtained for the other disease versus control comparisons.



View larger version (17K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 2. Permutation test of marker scoring algorithm for SCC versus control (normals + LSILs) comparison. In this test, marker scores for SCC were calculated for all 30,732 clones in the cDNA microarray data set, and the genes were ranked by score. The sample labels were then randomly permuted 100 times, and the marker scores were recalculated and ranked for each permutation. For each rank, the 5th, 50th, and 95th percentile of the marker scores from the 100 permutations, along with the corresponding actual score for the correctly labeled samples, were plotted. The marker scores of SCC from actual sample labels consistently exceeded the 95th percentile of the permuted runs. The top 100 marker ranks are shown.

 
Because the aim of the present study is to discover new markers associated with the transformation of cervical cells, up-regulated genes related to an immune response (i.e., immunoglobulins, MHCs) were excluded during marker selection. Clones with marker scores ranked in the top 50 from SCC or ACA paradigms, and clones ranked between 50 and 100 that were overexpressed in both SCC and ACA samples were selected for additional evaluation. Scores from the HSIL paradigm were not used independently to select markers because increased expression in tumors was considered essential for good marker performance. After consolidation, a total of 62 markers were selected. Their scores in SCC, ACA, and HSIL paradigms are shown in Table 1Citation . It was found that most of the up-regulated genes from SCC samples were also elevated in ACA. Although many markers selected from the SCC and/or ACA paradigms have scores > 3.0, only a few of the HSIL markers had scores > 2.0, indicating increasing expression as lesions progress from dysplasia to invasive carcinomas. Fig. 3Citation shows three genes from Table 1Citation that represent typical but distinct types of expression patterns among normal, LSIL, HSIL, SCC, and ACA tissues. MCM 6 was overexpressed in HSILs, SCCs, and ACAs. Some genes such as claudin 1 were overexpressed only in SCCs, whereas others (e.g., 37310_EST) were increased mainly in ACAs.


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

  		Table 1 Markers selected by transcriptional profiling

 


View larger version (32K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 3. Transcriptional profiles (TP) of MCM6, Claudin 1, and EST 37310 in normal, dysplastic, and cancerous cervical tissues by cDNA microarray hybridization. Each data point represents the average of duplicate microarray hybridizations. The TP intensity was normalized by the median intensity of all spots on the array. Endo, normal endocervical tissue; Ecto, normal ectocervical tissue.

 
In an attempt to reveal functional relationships between the 62 up-regulated genes from Table 1Citation , hierarchical clustering was performed based on the expression profiles across all clinical samples (Fig. 4)Citation . Interestingly, these overexpressed genes were clustered into two main groups. One group consists mainly of genes that encode either ECM proteins (collagen, laminin, fibronectin) or proteins responsible for cell-ECM interaction or ECM degradation and remodeling (e.g., osteonectin, matrix metalloproteinase, urokinase). The other cluster contains many genes involved in cell replication and proliferation. Examples include DNA replication licensing factors (MCM 4 and MCM 6), aurora-related kinase 2, topoisomerase 2A, and the oncogene B-Myb.



View larger version (44K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 4. Cluster diagram of the top 62 gene markers. The dendrogram was created by hierarchical clustering of 62 expression profiles across the 34 cervical tissue samples. Two major clusters were identified.

 
Marker Confirmation by ISH.
The 62 markers from Table 1Citation were also evaluated by ISH in a second independent set of clinical tissue samples. This step is important both to confirm transcriptional profiling results and to determine the cell types responsible for increased mRNA expression. ISH experiments were performed using tissue microarrays. Depending on the level of the paraffin block sectioned, 26–87 normal cervical tissue cores (from ectocervix and endocervix), 2–10 LSILs, 5–33 HSILs, and 10–21 cancer cores (including SCC, ACA, and poorly differentiated carcinomas) were examined. As a first assessment, we classified the 62 markers into three categories based on their ISH results: 35 genes/ESTs were expressed in a higher percentage of cervical cancer cores than normals, 24 showed no detectable signal, and 3 were not differentially expressed. In general, the ISH signal was detected either in the cervical epithelial cells or in the stroma (Table 2)Citation . As expected, genes that were up-regulated in the stroma typically encode proteins for ECM composition or remodeling, whereas many genes that are overexpressed in epithelial cells are responsible for cell growth and cell-ECM interactions. Interestingly, laminin {gamma} 2 and COL1A1, which were determined by ISH to be differentially expressed in the cervical epithelial cells, clustered with other ECM or ECM degradation genes that were up-regulated in the stroma (COL1A2, osteonectin, MMP11, MGP, OSF-2p1, and so forth; Fig. 4Citation ). Because the gene clustering in Fig. 4Citation was based on the similarity of gene transcriptional profiles across all tested cervical tissues, the coclustering of epithelial expressing LAMC2 and COL1A1 with other stroma-expressing genes might suggest coordinated gene regulation between cervical epithelium and its microenvironment during cancer progression.


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

  		Table 2 Genes overexpressed in cervical cancer by ISH

 
Photomicrographs of two representative genes, claudin 1 and MGP, are shown as examples of the ISH experiments (Fig. 5)Citation . There was little or no detectable signal from either claudin 1 (Fig. 5A)Citation or MGP (Fig. 5B)Citation probes in normal endo-/ectocervical tissues. Gene expression was elevated in HSIL and increased further in cervical tumors. Unlike claudin 1, where expression was limited to the epithelium, the differential expression of MGP was located in the stroma of cervical lesions and tumors. It is interesting to note that claudin 1 expression was not significantly elevated in the 5 HSIL and 3 ACA samples that were profiled on cDNA microarrays (Fig. 3)Citation . The increased sensitivity of ISH in this case could be attributable to the focal nature of the signal. Such focal signals are readily apparent by ISH but can be missed in RNA preparations of whole tissue homogenates.



View larger version (107K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 5. In situ hybridization detection of Claudin 1 (A) and MGP (B) mRNAs from a tissue microarray. For each gene, one representative tissue core for each tissue type was selected for presentation. Bright field panels show the H&E-stained normal ectocervix, normal endocervix, LSIL, HSIL, SCC, and ACA. The dark field panel of (A) shows low level expression of Claudin 1 in normal basal epithelial cells. The gene expression was elevated mildly in LSIL and increased significantly in HSIL, SCC, and ACA. The dark field panel for MGP (B) shows that gene expression was elevated in HSIL, SCC, and ACA but located in the stroma adjacent to the diseased epithelia.

 
Because cervical screening evaluates morphological changes of cells exfoliated from cervical epithelium, cells from stroma are unlikely to be present in a Pap test sample. We therefore focused our marker selection effort on those candidate markers that were differentially expressed in the epithelial cells of cervical dysplasias and invasive tumors. To understand the frequency with which each marker was elevated in different types of cervical lesions and tumors, a frequency calculation was performed using all tissue cores on the microarray. The calculation was based on a manual, semiquantitative scoring method. The signal was scored on a scale from 0 to +++: 0, no signal; +, weak indeterminate signal; ++, determinate, weak to moderate signal; and +++, strong to very strong signal. Among the 27 markers expressed in the epithelium, 17 were selected as top performing markers (Table 3)Citation . These markers displayed the greatest differential expression between tumors and normals, with preference given to markers having minimal signal in normal tissues. The other 10 markers either showed significant expression in normal endocervical tissues or had very weak signal intensity in disease tissues. To be considered positive, a tissue core had to have a signal score of ++ or higher. In cases where the microarray contained more than one tissue core from a single patient, a positive call required at least 50% of tissue cores to be ++ or higher. To better visualize the results, the selected gene markers are presented in the order of increasing frequency of positive cores for normal cervical tissues. It was found that the frequency of marker elevation is highly correlated with the stage of clinical abnormality and varies in a broad range from gene to gene at particular clinical stages. IFI27, FOSL2, CRIP 1, and NP25 had relatively high (>20%) positive cores from normal cervical tissues, whereas markers such as ITGB6 and CLDN1 were relatively lower in normals and started to increase in LSILs and HSILs. The appearance of positive cores for BST2 and LAMC2 took place even later in the tumor progression stage at the transition from high-grade premalignant lesions to invasive disease. These findings demonstrate the existence of markers that identify sequential molecular changes during cervical cancer development. An ideal screening marker should distinguish the most advanced dysplasias and cancers from benign and low grade lesions. Although the number of sample cores examined and the ISH quantitation method used in this study are not designed to determine the best genes for diagnostic purposes, markers such as MCM6, B-MYB, GAPB3, and CLDN1 appear to have some promise.


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

  		Table 3 ISHa Expression of top markers using a cervical tissue microarray

 

   DISCUSSION
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Cytological screening for cervical cancer and its precursors using a Pap test has successfully reduced the morbidity and mortality of cervical cancer. However, because the Pap test relies on morphological features of cervical epithelial cells, it is adversely affected by a high rate of false positive and false negative results. More objective diagnostic parameters are needed to identify truly dysplastic and malignant cells, especially in ASCUS and LSIL smears. Increased confidence in the diagnosis will improve patient management and reduce the cost of repeat testing and colposcopy. Using cDNA microarray technology, we identified 62 genes and ESTs that were up-regulated in low-grade and high-grade cervical lesions, as well as in invasive tumors. Some of these genes, urokinase, B-myb, thymidine kinase, and carbonic anhydrase, have been shown previously to be up-regulated in cervical SCCs or HPV-16-immortalized endocervical cell lines (18, 19, 20, 21) . Other genes, such as claudin 1, mucin 1, mesothelin, and topoisomerase II {alpha}, have been associated with cancers other than cervical, including breast, ovarian, lung, and other cancers (22, 23, 24, 25) . The presence of these genes among the markers identified in the present study lends support for the cDNA microarray and marker selection approach that was used.

Among the 35 ISH-confirmed genes and ESTs, we found a subset up-regulated in the epithelium of cervical lesions and tumors, whereas others were differentially expressed in the stroma. As expected, many of the genes overexpressed in the epithelium have a role in DNA replication and cell cycle regulation, whereas those elevated in stroma encode proteins for major ECM components or for ECM modeling. It was of interest that two genes overexpressed in the epithelium, LAMC2 and COL1A1, shared similar transcriptional profiling patterns with genes up-regulated in the stroma. This might indicate a coordinated response between cervical epithelium and its microenvironment as malignancy progresses. Such interactions between stromal, endothelial, and epithelial tumor cells have been demonstrated in cyclo-oxygenase 2-/- mice, where stromal gene expression was required for lung tumor growth (26) . The role of stromal/epithelial interactions in tumor development is of increasing interest (27) , and the up-regulated stromal genes identified in the present study may prove valuable in this regard.

New technologies aimed at increasing the accuracy of Pap tests are focused on two aspects: reducing sampling errors and reducing interpretation errors. Although liquid-based cytology (ThinPrep; Cytyc Corporation, Boxborough, MA; AutoCyte; TriPath Imaging, Burlington, NC) has significantly minimized the sampling errors, biomarker-based assays are the main hope for increasing interpretation accuracy. To date, numerous biomarkers have been discovered and developed, including high-risk HPV markers and cellular markers (p16/INK4A, MCM5, cytokeratins, and so forth; Refs. 28, 29, 30 ). These markers have been shown to dramatically improve the ability to detect precursor malignant cells. Unfortunately, the price for increased sensitivity is the high-positive rate of low-grade lesions, making it difficult to differentiate precancerous lesions that will regress from those that have acquired malignant characteristics and represent a high risk of progressing to invasive cancer. In our study, we showed that markers differed in their ability to detect cancers as well as low- and high-grade dysplasias. Some genes clearly appeared to be up-regulated at earlier stages of dysplasia than others. Because the vast majority of premalignant lesions regress spontaneously, the perfect marker would detect no normal cells and very few low-grade abnormalities (LSIL and ASCUS) but would detect a higher proportion of high-grade lesions and all cancers. As seen in Table 3Citation , no single marker meets these ideal criteria. Promising markers do exist, however. For example, MCM6 has a combined 77% sensitivity for detecting tumors and high-grade lesions, with a specificity of >98% for correctly excluding normals and low grade lesions. We fully expect, however, that ideal performance will be reached only by combining markers into sets.

It should also be emphasized that ISH is not likely to become the platform of choice for cervical smear analysis. ISH was used in the present study to validate candidate markers that, in the future, should be evaluated with a more technically robust procedure such as IHC. This latter work will require first the generation of appropriate immunogens and monoclonal antibodies and later the analysis of cervical smears from appropriate patient populations.

The current study found that virtually all genes overexpressed in cervical dysplasias or squamous carcinomas were also up-regulated in ACAs. Although ACAs and adenosquamous carcinomas account for only 10–20% of cervical cancers, they are responsible for up to 80% of all cervical cancer malpractice suits because of the low sensitivity of cytology and colposcopy for this type of disease (5) . The marker candidates from this study may improve the early detection of ACAs and endocervical lesions of clinical concern.

Because the goal of this study was to discover promising markers for cervical screening, we only focused on those genes that were overexpressed in the epithelial cells of premalignant lesions and cancers. However, those markers that showed differential expression in the stroma of diseased cervical tissues also have potential clinical use. Stromal enzyme inhibitors such as MMP inhibitors and antiadhesive molecules such as anti-integrin peptides or antibodies have been used as alternative strategies to treat a variety of cancers in human clinical trials and murine tumor models (31, 32, 33) . The genes identified in the current study should not only provide new opportunities for diagnostic markers but may improve our understanding of cervical cancer biology and help identify therapeutic drug targets.


   FOOTNOTES
 
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1 To whom requests for reprints should be addressed, at Millennium Pharmaceuticals, Inc., 45 Sidney Street, Cambridge, MA 02139. Phone: (617) 761-6971; Fax: (617) 444-1670; E-mail: schlegel{at}mpi.com Back

2 The abbreviations used are: ASCUS, atypical squamous cells of undetermined significance; AGUS, atypical glandular cells of undetermined significance; LSIL, low-grade squamous intraepithelial lesion; HSIL, high-grade squamous intraepithelial lesion; HPV, human papillomavirus; ISH, in situ hybridization; SCC, squamous cell carcinoma; ACA, adenocarcinoma; ECM, extracellular matrix; MGP, matrix Gla protein; EST, expressed sequence tag. Back

Received 7/ 1/02. Accepted 2/18/03.


   REFERENCES
Top
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Greenberg M. D., Sedlacek T. V., Campion M. J. Cervical neoplasia: are adjunctive tests to cervical cytology worthwhile?. Clin. Obstet. Gynecol., 38: 600-609, 1995.[Medline]
  2. Eddy D. M. Screening for cervical cancer. Ann. Intern. Med., 113: 214-226, 1990.
  3. National Cancer Institute Workshop. The 1988 Bethesda System for reporting cervical/vaginal cytologic diagnosis. J. Am. Med. Assoc., 262: 931-934, 1989.[Abstract/Free Full Text]
  4. Fahey M. T., Irwig L., Macaskill P. Meta-analysis of Pap test accuracy. Am. J. Epidemiol., 141: 680-689, 1995.[Abstract/Free Full Text]
  5. Cox J. T. Evaluation of abnormal cervical cytology. Clin. Lab. Med., 2: 303-343, 2000.
  6. Ostor A. G. Natural history of cervical intraepithelial neoplasia: a critical review. Int. J. Gynecol. Pathol., 12: 186-192, 1993.[Medline]
  7. Anderson M. C. Glandular lesions of the cervix. Baillieres Clin. Obstet. Gynaecol., 9: 105-119, 1995.[Medline]
  8. Wright T. C., Cox J. T., Massad L. S., Twiggs L. B., Wilkinson E. J. 2001 consensus guidelines for the management of women with cervical cytological abnormalities. J. Am. Med. Assoc., 287: 2120-2129, 2002.[Abstract/Free Full Text]
  9. Walboomers J. M., Jacobs M. V., Manos M. M., Bosch F. X., Kummer J. A., Shah K. V., Snijders P. J., Peto J., Meijer C. J., Munoz N. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. J. Pathol., 189: 12-19, 1999.[Medline]
  10. Wick M. J. Diagnosis of human papillomavirus gynecologic infections. Clin. Lab. Med., 2: 271-287, 2000.
  11. Sasieni P. D. Human papillomavirus screening and cervical cancer prevention. J. Am. Med. Wom. Assoc., 55: 216-219, 2000.
  12. Sasieni P. D. A systematic review of the role of human papilloma virus (HPV) testing within a cervical screening programme: summary and conclusions. Br. J. Cancer, 83: 561-565, 2000.[Medline]
  13. Human papillomavirus testing for triage of women with cytologic evidence of low-grade squamous intraepithelial lesions: baseline data from a randomized trial. The Atypical Squamous Cells of Undetermined Significance/Low-Grade Squamous Intraepithelial Lesions Triage Study (ALTS) Group. J. Natl. Cancer Inst. (Bethesda), 92: 397-402, 2000.[Abstract/Free Full Text]
  14. Solomon D., Schiffman M., Tarone R., the ALTS Study Group. Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial. J. Natl. Cancer Inst. (Bethesda), 93: 293-299, 2001.[Abstract/Free Full Text]
  15. Ried T., Heselmeyer-Haddad K., Blegen H., Schrock E., Auer G. Genomic changes defining the genesis, progression, and malignancy potential in solid human tumors: a phenotype/genotype correlation. Genes Chromosomes Cancer, 25: 195-204, 1999.[Medline]
  16. Efron B., Tibshirani R. J. . An introduction to the bootstrap. Chap. 15: permutation tests, CRC Press New York 1998.
  17. Duncan L. M., Deeds J., Cronin F. E., Donovan M., Sober A. J., Kauffman M., McCarthy J. J. Melastatin expression and prognosis in cutaneous malignant melanoma. J. Clin. Oncol., 19: 568-576, 2001.[Abstract/Free Full Text]
  18. Yang X., Nakao Y., Pater M. M., Tang S. C., Pater A. Expression of cellular genes in HPV16-immortalized and cigarette smoke condensate-transformed human endocervical cells. J. Cell Biochem., 66: 309-321, 1997.[Medline]
  19. Fujiwaki R., Hata K., Moriyama M., Iwanari O., Katabuchi H., Okamura H., Miyazaki K. Clinical value of thymidine kinase in patients with cervical carcinoma. Oncology, 61: 47-54, 2001.
  20. Loncaster J. A., Harris A. L., Davidson S. E., Logue J. P., Hunter R. D., Wycoff C. C., Pastorek J., Ratcliffe P. J., Stratford I. J., West C. M. Carbonic anhydrase (CA IX) expression, a potential new intrinsic marker of hypoxia: correlations with tumor oxygen measurements and prognosis in locally advanced carcinoma of the cervix. Cancer Res., 61: 6394-6399, 2001.[Abstract/Free Full Text]
  21. Riethdorf L., Riethdorf S., Petersen S., Bauer M., Herbst H., Janicke F., Loning T. Urokinase gene expression indicates early invasive growth in squamous cell lesions of the uterine cervix. J. Pathol., 189: 245-250, 1999.[Medline]
  22. Kramer F., White K., Kubbies M., Swisshelm K., Weber B. H. Genomic organization of claudin-1 and its assessment in hereditary and sporadic breast cancer. Hum. Genet., 107: 249-256, 2000.[Medline]
  23. Apostolopoulos V., Pietersz G. A., McKenzie I. F. MUC1 and breast cancer. Curr. Opin. Mol. Ther., 1: 98-103, 1999.[Medline]
  24. Keith W. N., Tan K. B., Brown R. Amplification of the topoisomerase II {alpha} gene in a non-small cell lung cancer cell line and characterisation of polymorphisms at the human topoisomerase II {alpha} and ß loci in normal tissue. Genes Chromosomes Cancer, 4: 169-175, 1992.[Medline]
  25. Hough C. D., Sherman-Baust C. A., Pizer E. S., Montz F. J., Im D. D., Rosenshein N. B., Cho K. R., Riggins G. J., Morin P. J. Large-scale serial analysis of gene expression reveals genes differentially expressed in ovarian cancer. Cancer Res., 60: 6281-6287, 2000.[Abstract/Free Full Text]
  26. Williams C. S., Tsujii M., Reese J., Dey S. K., DuBois R. N. Host cyclooxygenase-2 modulates carcinoma growth. J. Clin. Investig., 105: 1589-1594, 2000.[Medline]
  27. Bissel M. J., Radisky D. Putting tumors in context. Nat. Rev. Cancer, 1: 46-54, 2001.[Medline]
  28. Klaes R., Friedrich T., Spitkovsky D., Ridder R., Rudy W., Petry U., Dallenbach-Hellweg G., Schmidt D., von Knebel Doeberitz M. Overexpression of p16(INK4A) as a specific marker for dysplastic and neoplastic epithelial cells of the cervix uteri. Int. J. Cancer, 92: 276-284, 2001.[Medline]
  29. Williams G. H., Romanowski P., Morris L., Madine M., Mills A. D., Stoeber K., Marr J., Laskey R. A., Coleman N. Improved cervical smear assessment using antibodies against proteins that regulate DNA replication. Proc. Natl. Acad. Sci. USA, 95: 14932-14937, 1998.[Abstract/Free Full Text]
  30. Maddox P., Sasieni P., Szarewski A., Anderson M., Hanby A. Differential expression of keratins 10, 17, and 19 in normal cervical epithelium, cervical intraepithelial neoplasia, and cervical carcinoma. J. Clin. Pathol. (Lond.), 52: 41-46, 1999.[Abstract]
  31. Rowinsky E. K., Humphrey R., Hammond L. A., Aylesworth C., Smetzer L., Hidalgo M., Morrow M., Smith L., Garner A., Sorensen J. M., Von Hoff D. D., Eckhardt S. G. Phase I and pharmacologic study of the specific matrix metalloproteinase inhibitor BAY 12-9566 on a protracted oral daily dosing schedule in patients with solid malignancies. J. Clin. Oncol., 18: 178-186, 2000.[Abstract/Free Full Text]
  32. Brooks P. C., Stromblad S., Klemke R., Visscher D., Sarkar F. H., Cheresh D. A. Anti-integrin {alpha} v ß 3 blocks human breast cancer growth and angiogenesis in human skin. J. Clin. Investig., 96: 1815-1822, 1995.
  33. Lode H. N., Moehler T., Xiang R., Jonczyk A., Gillies S. D., Cheresh D. A., Reisfeld R. A. Synergy between an antiangiogenic integrin {alpha}, antagonist and an antibody-cytokine fusion protein eradicates spontaneous tumor metastases. Proc. Natl. Acad. Sci. USA, 96: 1591-1596, 1999.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 J. Clin. Pathol.Home page
P Conesa-Zamora, A Domenech-Peris, S Ortiz-Reina, F J Orantes-Casado, J Acosta-Ortega, J Garcia-Solano, and M Perez-Guillermo
Immunohistochemical evaluation of ProEx C in human papillomavirus-induced lesions of the cervix
J. Clin. Pathol., February 1, 2009; 62(2): 159 - 162.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
N. Shivapurkar, V. Stastny, N. Okumura, L. Girard, Y. Xie, C. Prinsen, F. B. Thunnissen, I. I. Wistuba, B. Czerniak, E. Frenkel, et al.
Cytoglobin, the Newest Member of the Globin Family, Functions as a Tumor Suppressor Gene
Cancer Res., September 15, 2008; 68(18): 7448 - 7456.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
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]

Home page
 Clin. Cancer Res.Home page
X.-S. Wang, Z. Zhang, H.-C. Wang, J.-L. Cai, Q.-W. Xu, M.-Q. Li, Y.-C. Chen, X.-P. Qian, T.-J. Lu, L.-Z. Yu, et al.
Rapid Identification of UCA1 as a Very Sensitive and Specific Unique Marker for Human Bladder Carcinoma.
Clin. Cancer Res., August 15, 2006; 12(16): 4851 - 4858.
[Abstract] [Full Text] [PDF]

Home page
 Ann OncolHome page
P. K. Wright
Response to "Primary cervical cancer screening by self-sampling of human papillomavirus DNA in internal medicine outpatient clinics", by Dannecker et al. (Ann Oncol 2004; 15: 863-869) and "Histopathological validation of the sentinel node concept in cervical cancer", by Barranger et al. (Ann Oncol 2004; 15: 870-874): DNA microarrays will be instrumental in the future diagnosis of cervical dysplasia and neoplasia
Ann. Onc., April 1, 2005; 16(4): 675 - 676.
[Full Text] [PDF]

Home page
 Cancer Res.Home page
B. Akgul, R. Garcia-Escudero, L. Ghali, H. J. Pfister, P. G. Fuchs, H. Navsaria, and A. Storey
The E7 Protein of Cutaneous Human Papillomavirus Type 8 Causes Invasion of Human Keratinocytes into the Dermis in Organotypic Cultures of Skin
Cancer Res., March 15, 2005; 65(6): 2216 - 2223.
[Abstract] [Full Text] [PDF]

Home page
 Clin. Chem.Home page
G. D. Sgarlato, C. L. Eastman, and H. H. Sussman
Panel of Genes Transcriptionally Up-regulated in Squamous Cell Carcinoma of the Cervix Identified by Representational Difference Analysis, Confirmed by Macroarray, and Validated by Real-Time Quantitative Reverse Transcription-PCR
Clin. Chem., January 1, 2005; 51(1): 27 - 34.
[Abstract] [Full Text] [PDF]

Home page
 J. Mol. Diagn.Home page
S.-i. Maruya, H.-W. Kim, R. S. Weber, J. J. Lee, M. Kies, M. A. Luna, J. G. Batsakis, and A. K. El-Naggar
Gene Expression Screening of Salivary Gland Neoplasms: Molecular Markers of Potential Histogenetic and Clinical Significance
J. Mol. Diagn., August 1, 2004; 6(3): 180 - 190.
[Abstract] [Full Text]

Home page
 Cancer Res.Home page
F. Thierry, M. A. Benotmane, C. Demeret, M. Mori, S. Teissier, and C. Desaintes
A Genomic Approach Reveals a Novel Mitotic Pathway in Papillomavirus Carcinogenesis
Cancer Res., February 1, 2004; 64(3): 895 - 903.
[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
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 Chen, Y.
Right arrow Articles by Schlegel, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chen, Y.
Right arrow Articles by Schlegel, R.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online