Cancer Research AACR Membership Protein Translation and Cancer
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
Cancer Research 67, 7113, August 1, 2007. doi: 10.1158/0008-5472.CAN-07-0260
© 2007 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
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 Gius, D.
Right arrow Articles by Rader, J. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Gius, D.
Right arrow Articles by Rader, J. S.

Molecular Biology, Pathobiology, and Genetics

Profiling Microdissected Epithelium and Stroma to Model Genomic Signatures for Cervical Carcinogenesis Accommodating for Covariates

David Gius1, Margo C. Funk2, Eric Y. Chuang1, Sheng Feng3, Phyllis C. Huettner4, Loan Nguyen2, C. Matthew Bradbury1, Mark Mishra1, Shuping Gao1, Barbara M. Buttin2, David E. Cohn2, Matthew A. Powell2, Neil S. Horowitz2, Bradford P. Whitcomb2 and Janet S. Rader2

1 Radiation Oncology Branch, Center for Cancer Research, National Cancer Institute, NIH, Bethesda, Maryland and 2 Division of Gynecologic Oncology, Department of Obstetrics and Gynecology; 3 Division of Biostatistics; and 4 Lauren V. Ackerman Laboratory of Surgical Pathology, Washington University School of Medicine, St. Louis, Missouri

Requests for reprints: David Gius, Radiation Oncology Branch, National Cancer Institute, NIH, 9000 Rockville Pike, Bethesda, MD 20892. Phone: 301-496-5457; Fax: 301-480-5439; E-mail: giusd{at}mail.nih.gov and Janet S. Rader, Departments of Obstetrics and Gynecology and Genetics, Washington University School of Medicine, 4911 Barnes-Jewish Hospital Plaza Box 8064, St. Louis, MO 63110. Phone: 314-362-3181; Fax: 314-362-2893; E-mail: raderj{at}wustl.edu.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 Conclusions
 References
 
This study is the first comprehensive, integrated approach to examine grade-specific changes in gene expression along the entire neoplastic spectrum of cervical intraepithelial neoplasia (CIN) in the process of cervical carcinogenesis. This was accomplished by identifying gene expression signatures of disease progression using cDNA microarrays to analyze RNA from laser-captured microdissected epithelium and underlying stroma from normal cervix, graded CINs, cancer, and patient-matched normal cervical tissues. A separate set of samples were subsequently validated using a linear mixed model that is ideal to control for interpatient gene expression profile variation, such as age and race. These validated genes were ultimately used to propose a genomically based model of the early events in cervical neoplastic transformation. In this model, the CIN 1 transition coincides with a proproliferative/immunosuppression gene signature in the epithelium that probably represents the epithelial response to human papillomavirus infection. The CIN 2 transition coincides with a proangiogenic signature, suggesting a cooperative signaling interaction between stroma and tumor cells. Finally, the CIN 3 and squamous cell carcinoma antigen transition coincide with a proinvasive gene signature that may be a response to epithelial tumor cell overcrowding. This work strongly suggests that premalignant cells experience a series of microenvironmental stresses at the epithelium/stroma cell interface that must be overcome to progress into a transformed phenotype and identifies the order of these events in vivo and their association with specific CIN transitions. [Cancer Res 2007;67(15):7113–23]


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 Conclusions
 References
 
Cervical malignancies present a unique opportunity to profile neoplastic transformation. The histologic transition from normal epithelial cells to preinvasive cervical intraepithelial neoplasia (CIN 1–3) and finally to squamous cell carcinoma antigen (SCCA) have been well characterized. Histologically, CIN 1 consists of immature basal-type cells involving the lower third of the epithelium. In CIN 2, these immature basal-type cells involve more than the lower third, whereas CIN 3 involves the full thickness of the epithelium. In addition, higher CIN grades exhibit nuclear crowding, pleomorphism, loss of cell polarity, and increased mitotic activity (1). These transitions seem to be well conserved and, as such, provide an intriguing system to use genomics to identify the early events in cervical cell transformation. The underlying stromal composition may vary slightly beneath the basement membrane; however, it is overwhelmingly characterized by fibroblasts. Complex molecular cross-talk between the epithelium and stroma has been implicated as a key component of neoplastic transformation, tumor invasion, and metastasis. Therefore, to construct an effective model for CIN progression in vivo, it is imperative to separately examine the different microenvironments involved in carcinogenesis.

Multidisciplinary studies have unequivocally implicated human papillomavirus (HPV) as the etiologic agent of neoplastic cervical lesions, with HPV DNA being found in 99.7% of invasive cervical carcinomas (2). Even so, most HPV infections are transient and spontaneously cleared by the host immune system. Thus, few women infected with HPV develop CIN 3 and fewer still progress to invasive carcinoma (3), suggesting that HPV does not act alone in the development of cervical cancer. Persistence of HPV infection and progression to cancer vary by age and race: Women over 30 years show a much higher persistence of HPV than younger women (4, 5), and progression to CIN 3 may vary by race (6).

To address interpatient variability, tissue heterogeneity, and experimental reproducibility, we used laser-capture microdissection (LCM) on 130 prospectively collected samples to isolate epithelial cells and stromal fibroblasts within 3 mm of the basement membrane from normal cervix, graded CINs, cancer, and patient compartment–matched "normal" cervical tissues. Microarray hybridization was done using only excellent quality RNA with triplicate replication per sample and a single universal reference across all samples. We identified specific markers in the epithelium and stromal compartments showing a progressive change in gene expression (increased or decreased) from viral cytopathic to invasive cancer and markers with a sharp change at CIN 3 and invasive cancer. Validation on a separate set of samples was improved by using a linear mixed model accommodating for age and race and showed several intriguing findings. Both preneoplastic epithelial cells and the surrounding fibroblasts induce the expression of proangiogenic factors, suggesting a potential cooperative communication between these differing cell types at the transition through CIN 2. In addition, distinct functional gene expression profiles were identified as cells progressed into the different CIN and SCCA subtypes. The in vivo model shows three distinct genomic signatures at the epithelium-stromal interface: (a) early viral/CIN 1 proproliferative/immunosuppression signature in the epithelium; (b) intermediate CIN 2 proangiogenic epithelium and stromal signature; and (c) late CIN 3/SCCA proinvasive epithelial and stromal signature.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 Conclusions
 References
 
Tissue collection, processing, slide preparation, and LCM. This study was approved by Washington University Human Studies Committee, and informed consent was obtained from all patients. Specimens excised from the cervix were placed in optimal cutting temperature embedding medium for frozen tissue specimens (Sakura Finetek USA). Approximately fifty 8-µm sections were cut per block. A reference slide was prepared at every 10th section and examined by a gynecologic pathologist (P.C.H.) who marked the exact location in the epithelium and stroma to be dissected and were validated in the LCM cap. The epithelium was more than 85% epithelial cells and the stroma was more than 85% fibroblasts. The mean depth of dissection was 2 mm (1–3 mm). For further details, see Supplementary Data.

RNA isolation and amplification. Immediately after LCM, each cap was placed in a 0.5 mL Eppendorf safe-lock tube containing 30 µL of extraction buffer from the PicoPure RNA Isolation Kit (Arcturus; see Supplementary Data). RNA quality were assessed using the Nanodrop Spectrophotometer (Nanodrop Technologies) and RNA 6000 Pico LabChip Kit in conjunction with the Agilent 2100 Bioanalyzer (Agilent Technologies), respectively. Only high-quality RNA with an 18S:28S rRNA ratio >1.0 was applied to the microarrays. An input of 10 to 20 ng of total cellular RNA containing poly(I)·poly(C) (Amersham Biosciences) as a nucleic acid carrier was amplified in two rounds of T7-based in vitro transcription, using the RiboAmp HS Amplification Kit (Arcturus) according to the manufacturer's instructions. A universal reference RNA (Stratagene) was in all microarray hybridizations.

Microdissected specimens. A total of 133 RNA samples were amplified from microdissected epithelium/tumor (86 samples) and stroma (47 samples). Represented in this total were 85 samples for microarray hybridization, including 26 epithelial and 9 stromal abnormal-normal patient matched pairs (52 and 18 samples, respectively) and 10 epithelial and 5 stromal samples from normal transformation zone. A validation set of 21 epithelial and 24 stromal samples ranging from normal tissue to cancer were obtained, microdissected, and processed in the same way as the discovery panel (shown in detail in Supplementary Table S1).

Microarray fabrication, hybridization, and analysis. The microarrays used for this study (National Cancer Institute ROSP 8 k human array) were prepared from the Research Genetics Named Genes set and contained 7,680 human cDNA clones enriched for known genes. cDNAs were spotted onto poly-L-lysine–coated slides using an OmniGrid arrayer (GeneMachines), as previously described (7). Probe labeling, microarray hybridization, and scanning were done as previously described (8) with few modifications. For all experiments, the cDNA probes from microdissected abnormal and normal specimens were compared with an amplified universal reference RNA (Stratagene). For each array, 3 µg of amplified RNA (aRNA) from a single microdissected sample and 6 µg of universal reference aRNA were labeled with Cy3-dUTP and Cy5-dUTP, respectively. Microarrays were scanned at 10 µm resolution, using a GenePix 4000A scanner (Axon Instruments, Inc.). The hybridized Cy3- and Cy5-labeled cDNA samples were scanned at 532 and 635 nm, respectively, and the resulting TIFF images were analyzed with GenePix Pro 3.0 software (Axon Instruments).

Data processing. Array elements that were flagged or had a median signal/background intensity <1.4 were excluded from further analysis. Global normalization of median signal and background intensities was done before background was subtracted from local signal intensities. The mean Cy3/Cy5 ratio of 3 hybridization replicates representing abnormal/reference was divided by the mean Cy3/Cy5 ratio of normal/reference to determine the indirect patient-matched abnormal/normal expression ratio that was used in all downstream data analyses.

Semiquantitative and real-time PCR. Reverse transcription was done on each aRNA sample used to confirm the microarray results. Using the RETROscript First Strand Synthesis kit (Ambion) and random decamers, we reverse-transcribed 2.5 µg of each aRNA sample according to the manufacturer's instructions. The resulting cDNA was semiquantitatively amplified in 25 to 30 cycles, using primer sets specifically designed for the 3'-biased products of the T7-based RNA amplification.

Online supplemental material. Supplemental Data includes the characterization of the patients validation set (Supplementary Table S1); a detailed description of analyses I and II; and additional details for the Materials and Methods for tissue processing, RNA isolation and amplification, and HPV typing.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 Conclusions
 References
 
LCM and preparation of samples of epithelium and underlying stroma from sequential neoplastic grades. To investigate the changes in genomic expression that occur as normal cervical tissue progresses to CIN and then to cervical carcinoma in specific microenvironments, we used LCM (Supplementary Figs. S1A–C) to isolate the epithelium and stromal compartments from the same area of cervix from each neoplastic grade. Dissected samples from the epithelium contained more than 85% epithelial cells and the stroma over 85% fibroblasts. The maximum height of the epithelium was 3 mm from the basement membrane with most samples being 1 to 2 mm. The stromal maximal dissection was less than or equal in height to the epithelium, usually 2 to 3 mm. The basement membrane was included in the epithelium. One hundred and thirty samples were meticulously marked for dissection by an experienced gynecologic pathologist (P.C.H.). This is critically important to this work because the proper histologic categories are necessary to produce the model of gene expression changes. Patient-matched neoplastic and normal epithelial samples were microdissected from five disease-free histologically normal specimens, three viral cytopathic effect (VCE), five CIN 1, four CIN 2, seven CIN 3, and six squamous cell carcinoma human cervical specimens. Patient-matched neoplastic and normal stromal samples were microdissected from two disease-free histologically normal specimens, two CIN 1, three CIN 2, two CIN 3, and two squamous cell carcinoma specimens. Patient characteristics include age, race, and HPV status (Table 2; Supplementary Table S1). A validation set of 21 epithelial and 24 stromal samples ranging from normal tissue to cancer were obtained, microdissected, and processed in the same way as the discovery panel (Table 2; Supplementary Table S1).


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

  		Table 2. Patient characteristics and HPV typing for microarray specimens

 
RNA was extracted from these samples and analyzed on cDNA microarrays. To ensure accurate hybridization, we used only RNA samples with a 28S:18S ratio >1.0 (Supplementary Fig. S1D). Even with the most rigorous technique, such high-quality RNA was obtained less often from stroma than from epithelium; thus, slightly fewer stromal samples were analyzed on the microarrays. To obtain enough RNA from LCM, we did two rounds of T7 linear amplification on all samples (Supplementary Fig. S1E). aRNA was then hybridized to cDNA microarrays against an amplified universal reference RNA in triplicate. After a pilot study in our laboratory5 showed that indirect (abnormal or normal to reference) and direct (abnormal to normal) hybridization gave similar results, we decided to use indirect hybridization, which provides more flexibility for data analysis. To avoid dye bias, we labeled all the experimental samples with the same dye (Cy3) and always labeled the reference RNA with Cy5. To achieve the statistical power and correct for interpatient and experimental variation, 246 cDNA microarrays and more than 50 quality control microarrays were used to evaluate LCM samples. The raw data for this analysis can be obtained online.

Analysis of epithelium and stromal samples from each neoplastic grade. We identified continuous trends in gene expression along the progression from normal cervical tissue to cancer and compared these trends in patient-matched normal and abnormal samples. To best identify the significant genes corresponding to disease progression in human samples, two different analysis methods were used. For analysis I, we used Functional Genomics v7.2 (Spotfire DecisionSite) to identify (a) increasing or decreasing abnormal/normal gene expression ratios, and (b) abrupt changes in gene expression between two CIN grades.

For analysis II, we generated a mixed linear model to fit the experimental design structure and accommodate multiple covariates. We identified variables and patterns that could be factored into the model to reduce interpatient variation and eliminate some possible false positives and therefore improve the pathway model. The false discovery rate was calculated using the multitest procedure in SAS by Benjamini and Hochberg (SAS Institute; ref. 9). Several interesting patterns were identified, modeling increasing and decreasing gene expression over disease progression and extent of gene expression within an individual neoplastic grade. In pattern 1 (nonparallel), gene expression in the adjacent normal specimens did not change as disease progressed (Fig. 1 ; Supplementary Table S2; Lists 2, 4, 6, 8, 10, 12, 14, and 16), whereas in pattern 2 (parallel), expression of target genes changed in both normal (immediately adjacent to abnormal) and neoplastic tissue with disease progression (Fig. 1; Supplementary Table S2; Lists 1, 3, 5, 7, 9, 11, 13, and 15). Pattern 1 identifies genes in neoplastic mechanisms whereby gene expression changes within histologic progression, whereas pattern 2 identifies genes in which the normal adjacent epithelium or stroma parallels the expression pattern of the neoplastic area and denote more globally changes occurring in the diseased cervix. These analytic methods also identified genes that changed their expression sharply at the threshold between CIN 3 and cancer and might therefore be markers of invasion (Fig. 1; Supplementary Table S2; Lists 5–8 and 13–16).


Figure 1
View larger version (50K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Differential expression along the progression from normal cervical tissue to cancer in abnormal versus patient-matched normal microdissected cervical specimens from microarray. A, differentially expressed genes in the epithelium. B, differentially expressed genes in stroma. Graphed values represent mean age- and race-adjusted intensities for each histopathologic grade. Black line, mean expression for histologically normal samples. Red and blue lines, increasing or decreasing expression, respectively. Colored boxes, different histologic grades. Y axis, the microarray fold change.

 
To compare the results across methods, we prepared a second population of microdissected epithelium and stroma cells for validation (Supplementary Table S1). We did semiquantitative reverse transcription-PCR (RT-PCR) on select genes from the top of each list obtaining <50% validation for analysis I and 68% (27 of 40) for analysis II. Validation required the identical trend in gene expression from normal to cancer as identified from the array data (Figs. 2 and 3 ; Table 1 ). Smaller P values represent the combination of (a) stronger upward or downward gene expression trends across the neoplastic progression and (b) smaller degrees of variance between values of the same neoplastic grade. This validation rate is high considering the genetic variability between individuals and environmental cofactors. Seven genes initially validated from analysis I failed to appear at the top of the 16 gene lists from analysis II. However, analysis II provided a list of genes that showed more consistent expression trends across cervical neoplasia in the two populations of cervical samples. The expression level from normal to cancer for the significant genes changed by a factor of 1.8 to 2.4 (Fig. 2).


Figure 2
View larger version (89K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Semiquantitative RT-PCR validation of selected genes from cervical epithelium and stroma. A, 21 microdissected epithelium. B, 24 microdissected stroma. Smaller P values represent the combination of stronger upward or downward gene expression trends across the neoplastic progression and smaller degrees of variance between values of the same neoplastic grade. Graphed values represent the percentage of maximum gene intensity/glyceraldehyde-3-phosphate dehydrogenase (Y axis; age- and race-adjusted intensities) for each histopathologic grade. The X axis is labeled with the corresponding histologic grade.

 

Figure 3
View larger version (61K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Schematic model for cervical carcinogenesis. Significant expression patterns using validated genes from analyses I and II are shown beneath the histologic grades; VCE, CIN 1 to 3, and SCCA. Yellow and red horizontal lines, basement membrane that separates the epithelium and stromal compartments. Genes coded in dark red, validated genes with increased expression; genes in dark blue, validated genes with decreased expression. Additional genes are also depicted in the model: gray, previously identified host and viral genes; light red or light blue, genes that increase or decrease, respectively, form the microarray data but are not yet validated. The CIN 1 transition coincides with a proproliferative/immunosuppression gene signature that appears to represent the epithelial response to HPV infection. The CIN 2 transition coincides with a proangiogenic stroma/epithelium signature, suggesting a cooperative signaling interaction between stromal and neoplastic cells responding to a shortage of nutrients necessary for proliferation. The CIN 3/SCCA transition coincides with a proinvasive gene signature that may be responding to tumor cell overcrowding.

 

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

  		Table 1. Differentially expressed genes from microarray analyses I and II

 
Epithelium and stromal gene expression patterns identify unique CIN progression pathways. The expression pattern of significant genes showed identical trends through the histologic grades of cervical carcinogenesis for both the discovery and validation sample set. For example, analysis II identified CENPF, CDKN2A, ITGAV, and KIF23 as showing nonparallel expression (Fig. 1). In contrast, MAPK7 and HINT1 are representative of genes showing parallel expression; the stroma beneath morphologically normal epithelium parallels the expression of the stroma beneath the adjacent neoplastic epithelium (Fig. 1B). IFNAR1 (Fig. 1A) was identified in analysis II as a sharp nonparallel increase in the epithelium. Visual inspection shows that IFNAR1 actually begins to increase in the neoplastic and adjacent normal epithelium at VRE but shows divergent and nonparallel expression at CIN 3. Overall, there was much less variability between the stroma underlying normal and neoplastic epithelium then within the neoplastic epithelium and adjacent normal and significant gene expression changes in the stromal fibroblasts did not begin until CIN 2.

These expression patterns can also be seen in the validation set. Microdissected histologically normal epithelial cell samples from patients with cancer (graphs 1 and 2, Fig. 2) show differential expression patterns paralleling neoplastic epithelium or stroma when compared with histologically normal samples from patients with no cervical pathology (graphs 3–8, Figs. 1 and 2) for IFNAR1, EMP1, IL1RN, ACCA1, and MMP3. MMP3 was initially identified by its sharp increase in expression at the threshold between CIN 3 and cancer in the stroma and epithelium (Fig. 1) and then validated (Fig. 2A and B). We show that microdissection of epithelium and stroma can identify genes showing opposite trends in expression. DSG3 shows decreasing expression in the epithelium and increasing expression in the stroma through disease progression (Figs. 1 and 2).

To determine whether our approach can identify potential clinical screening markers, we did semiquantitative RT-PCR for KIF23, using RNA from a third population of Pap smears (Supplementary Fig. S2). This gene has not previously been identified as a marker of carcinogenesis, but its expression more than doubled in the microarray and validation studies (Figs. 1 and 2). It was also overexpressed in "normal" cervical epithelium from cancer patients in the validation set, its nonparallel expression pattern being similar to that of CDKN2A, which is undergoing extensive clinical testing as a cervical cancer screening marker. The raw data and gene lists (Table 3) provided here will provide a framework for the establishment of an evolving model of cervical carcinogenesis at the epithelium-stromal interface and biomarker development.


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

  		Table 3. Up- and down-regulated genes alter in the CIN 1-3 transition signatures

 
Evaluation of HPV status in microarray samples. HPV status was characterized along the neoplastic progression by using general primer PCR, denaturing high-performance liquid chromatography, and sequencing for each specimen (Table 2 ; ref. 10). In epithelium, low-grade cervical lesions (VCE and CIN 1) were characterized by a mixture of low-risk (HPV 53, 84), intermediate-risk (HPV 39, 52, 56), and high-risk (HPV 16, 18, 31, 45) types. High-grade dysplasia (CIN 2 and 3) was characterized primarily by high-risk HPV (HPV 16, 18, 31, 59) and a few intermediate-risk HPV types (HPV 52, 58). Histopathologically confirmed squamous carcinoma cells contained only high-risk HPV types (HPV 16, 18, 31, and 45). There was a complete absence of HPV DNA from all disease-free cervical specimens. A few specimens contained more than one HPV type.


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
Conclusions
 References
 
This study is the first comprehensive, integrated approach to examining histologic, grade-specific changes in gene expression along the entire neoplastic spectrum of cervical carcinogenesis. This study is also the first to use LCM to isolate and describe the distinct cell populations of the epithelial and stromal microenvironments and to use patient-matched, histopathologically normal tissue as a control for each abnormal specimen. Using a linear mixed model for analysis of gene expression patterns, we propose the first genomically based model of early events in cervical neoplastic transformation (Fig. 3). This application of a linear mixed model is ideal for the complicated experimental design structure to control for interpatient gene expression profile variation. This ANOVA-based statistical analysis not only allows for accommodating for important covariants, such as age and race, but also estimates and separates the large between-patient variation from the total variation and thus increases the statistical power. Bioinformatic analysis of our results identified distinct expression pathways that coincided with the histopathologic transitions from normal to CIN 1, CIN 2, CIN 3, and cancer. Based on these results, we propose an initial evolving in vivo model for cervical carcinogenesis (Fig. 3) that we hope will serve as a template for other researchers to build more sophisticated models as additional and updated bioinformatics data become available (raw data for this analysis can be obtained online).

The results of these experiments also suggest that a potential communication exist between stromal fibroblasts and premalignant epithelial cells to induce proangiogenic signals that appear at the CIN 2 transition. This analysis suggests that the "angiogenic switch," as proposed by Hanahan and Folkman, may depend on cross-talk between stromal fibroblasts and CIN cells in the context of the differing tumor microenvironments (11). These authors suggest that the angiogenic switch may be critically important and a potentially rate-limiting step initiating early tumor progression as well as tumor establishment. These results extend this idea and suggest that the angiogenic switch in CIN 2 requires communication between both epithelial cells and the surrounding stromal fibroblasts and that each compartment individually induces the expression of angiogenic factors.

The proproliferative genomic signature for the CIN 1 transition. Early infection of the cervix with HPV should induce a series of molecular events in which the viral genome establishes itself as an episome within basal epithelial cells and uses host replication machinery to increase its copy number. The natural viral life cycle requires the expression of the viral oncogenes E6 and E7, which dysregulate the host cell cycle by interacting with p53 and pRb, respectively. The histopathologic presentation that corresponds to early viral infection is referred to as the VCE (koilocytotic atypia) and CIN 1. Our data suggest an initial series of genetic events in the basal epithelial cells that directly allows for viral replication that may also indirectly establish an environment permissive for neoplastic progression. Nine genes were validated by PCR (Table 3 ) and fall into two function physiologic groups: (a) genes that favor cellular proliferation (CDKN2A, CENPF, KIF23, ITGAV, and ACAA1); and (b) genes that alter the cellular immune system (IFNAR1, IL1RN, EMP1, and L1RN). These genes were used to propose a proproliferative/immunosuppression genomic signature that corresponds with the CIN 1 transition (Fig. 3).

Viral replication requires cell entry into the S-phase of cell cycle and expression of the E7 protein inactivates pRb resulting in the release of "unchecked" E2F family transcription factors that favor progression through the late G1 cell cycle checkpoint (11). An intercellular redundancy (or secondary) pathway that also inhibits pRb activity involves phosphorylation by the CDK4/cyclinD1 complex and this process is negatively regulated by CDKN2A (p16INK4a), whose primary action is to inhibit CDK4. Up-regulation of CDKN2A in early CIN and cervical cancer might represent the host's response to inactivation of pRb (12). This is consistent with the idea that infected cells may inhibit both primary and secondary pathways that normally prevent cell division and thus create a permissive state that is favorable for proliferation and ultimately transformation. Thus, we propose that the up-regulation of CDKN2A in early CIN and cervical cancer might be the host's response to inactivation of pRb and release of E2F.

It is also proposed that the HPV life cycle should activate machinery that induces processes for viral DNA transport, packaging, and subsequent viral shedding. In this regard, KIF23 and CENPF are increased the CIN 1 transition. KIF23 is a microtubule-dependent molecular motor that is essential for midzone/midbody formation at the mitotic spindle during cytokinesis (13). Interestingly, KIF23 is also increased in PAP smears women with SCCA (Supplementary Fig. S2). CENPF is a centromere passenger protein that acts as a scaffold to recruit and assemble kinetochores during the cell cycle. It ensures proper segregation of chromosomes by facilitating interactions between the centromere and mitotic spindle and overexpression of CENPF has been associated with human cancers (14). Thus, we suggest that the induction KIF23 and CENPF are necessary for the early viral life cycle and this may indirectly establishes a transformation-permissive microenvironment.

Our microarray analysis showed that the integrin ITGAV increases during early HPV infection and sharply increases at the CIN 3 interphase. Integrins are heterodimeric membrane proteins that regulate cell adhesion, cancer invasion, and altered patterns of integrin expression have been reported in neoplastic epithelium. For example, integrin {alpha}Vß3 has been implicated in the inhibition of anoikis (15); thus, up-regulation of ITGAV might contribute to the survival of HPV-infected cells that rise through the suprabasal epithelium. Indeed, HPV-infected cells have been found to enter the S-phase after they detach from the basal membrane (16), whereas uninfected basal cells exit the cell cycle after losing contact with the basement membrane (Fig. 3). Thus, altered integrin expression may favor HPV survival (Fig. 3).

The expression of acetyl-CoA acyltransferase 1 (ACAA1) was also increased in CIN 1 epithelial cells. ACAA1 is operative in the ß-oxidation system of the peroxisomes and has never been implicated in tumorigenesis. However, peroxisome proliferator-activated receptors (PPAR) act as transcription factors, binding with retinoid X receptor (RXR), and have been implicated in tumorigenesis (17). Moreover, PPAR{gamma} has been shown to be down-regulated in cervical tumors (18). Because ACAA1 is regulated by PPAR, we speculate that a common PPAR ligand, coactivator, or RXR regulatory mechanism may be involved in the progression of early cervical neoplasia (Fig. 3). Finally, these results suggest that these five genes (CDKN2A, CENPF, KIF23, ITGAV, and ACAA1) may play a role in the natural viral life cycle by supporting S-phase entry, proliferation, DNA segregation, and HPV survival and based on these results we propose a proproliferative early CIN 1 genomic signature.

The immunosuppression genomic signature for the CIN 1 transition. It has been previously suggested that the early HPV life cycle should create a microenvironment that allows infected cells to conceal their presence from the host's immune system. This analysis identified four genes (Table 3) that have been associated with the regulation of the immune system (IFNAR1, IL1RN, EMP1, and L1RN). IFNAR1 encodes a membrane protein that stimulates Janus-activated kinase/signal transducers and activators of transcription (JAK/Stat) signaling when complexed by IFNAR2 and bound by IFN-{alpha} (19). The IFN-{alpha}–JAK/Stat signaling pathway is critical for host defense against viral infection by stimulating transcription of antiviral, antiproliferative, and antitumor genes. This suggests a mechanism whereby increased IFNAR1 is up-regulated in HPV-infected cells as a host feedback response to inhibition of the IFN-{alpha}-JAK/Stat pathway, as happens in a multiple myeloma cell line (20). This result is consistent with the induction of IFN-inducible genes identified in infected cell lines and may be a host response to infection (21).

EMP1 expression is associated with squamous cell differentiation and is suppressed by retinoic acid receptor signaling (22), and transfection studies have shown an S-phase arrest along with decreased cell growth (23). IL1RN is a member of the interleukin 1 (IL-1) cytokine family and functions as a competitive receptor antagonist blocking binding of IL-1. EMP1 and IL1RN may play a role in cervical cells via interaction with the ligand gated ion channel, P2RX7 (Fig. 2). It has been shown that the binding of EMP1 and other EMP proteins to the COOH-terminal domain of P2RX7 activates channel formation and membrane blebbing (24). Furthermore, activation of P2RX7 results in extracellular secretion of IL1RN (25). The microarray analysis suggests that the concurrent down-regulation of EMP1 and IL1RN may be affecting this pathway, with a final downstream effect of decreased IL1RN secretion. Because IL1RN competes with IL-1 at the IL-1 receptor, decreased expression of this antagonist may result in the increased downstream effects of IL-1. This might include the inhibition of apoptosis in HPV-infected cells and the activation of cancer promoting nuclear factor-{kappa}B and c-Jun/activator protein-1, which are increased in cervical cancer (26, 27). Based on these results, we suggest that the altered expression of IFNAR1, IL1RN, EMP1, and L1RN may be necessary for the viral concealment from the host's immune system that may also indirectly establish permissive microenvironment for transformation (Fig. 3). These genes were used to propose an immunosuppression genomic signature for the CIN 1 transition.

The analysis of these results suggests that the epithelial genomic effects of early HPV viral infection, in the transition to koilocytotic atypia and CIN 1, share a very similar profile and these genes fall into two major categories. The first involves the uncoupling of cell cycle regulation that promotes proliferation (CDKN2A, KIF23, CENPF, ITGAV, and EMP1) and the second appears to allow virally infected cells to evade the host immune system (IFNAR1 and IL1RN). This seems logical because after viral infection of a host cell, it should both initiate intracellular processes necessary for viral replication and at the same time hide itself from the host immune surveillance systems. Thus, we refer to this transition as the "proproliferative/immunosuppression genomic signature."

The proangiogenic stroma/epithelium interaction genomic signature at the CIN 2 transition. This study suggests that during the CIN 2 transition, epithelial cells may be responding to the lack of local nutrients via a communication between stromal fibroblasts and epithelial cells to activate neovascularization pathways. This may result from the increasing metabolic demands from rapidly dividing epithelial cells that begin to overwhelm the existing vascular network. At the CIN 2 transition, six genes were validated by PCR (Table 3) that all fall into a physiologic group that may affect angiogenesis. The expression of five of these genes (Table 3) was altered in stromal cells (HINT1 and TAGLN2 were up-regulated, whereas MAP2K7, DAB2, and TBX19 were down-regulated), whereas only one gene was altered in epithelial cells (KAL1 was increased).

HINT1 is a member of the conserved histidine triad protein family involved in a variety of regulatory networks including chromatin remodeling, proliferation, and immune response (28). HINT1 is a repressor of MITF, an essential mast cell protein that regulates expression of many important mast cell products, including the mast cell proteases mMCP-5 and mMCP-6. In the murine model of cervical cancer, angiogenesis was activated concurrently with infiltration of the stroma by mast cells and subsequent expression of mMCP-4 and mMCP-6 (29) and this is consistent with our results from cervical biopsies. Thus, we propose that increased expression of HINT1 in CIN 2 stroma may be a marker of mast cell infiltration and inflammation that promotes vascularization and angiogenesis (Fig. 3).

We also detected up-regulation of TAGLN2, a homologue of transgelin that is a marker for hepatocellular carcinoma (29). The function of TAGLN2 is largely unknown but the gene becomes up-regulated as prostatic stromal fibroblasts transform into myofibroblasts in response to transforming growth factor ß (TGF-ß; ref. 30) and may be associated with early formation of a permissive environment for epithelial cell transformation. Thus, TAGLN2 could be one of several factors that influence angiogenesis and downstream cell invasion by contributing to a "reactive stroma" microenvironment that has been described histopathologically as desmoplasia (31).

We also report that the stress kinase MAP2K7 is down-regulated in CIN stroma. MAP2K7 is widely expressed and mediates the cell responses to proinflammatory cytokines and environmental stresses. This protein is a negative regulator of mast cell proliferation, as shown in mkk7(–/–) mast cells that are hyperproliferative (32). Therefore, down-regulation of MAP2K7 at the CIN 2 transition may be involved in release of negative regulation of cell growth, proliferation, and angiogenesis.

DAB2 is a widely expressed adaptor molecule involved in several receptor-medicated signaling pathways, including cellular organization and macrophage adhesion, and TGF-ß was decreased in this analysis. The function of DAB2 in the stroma is less understood, although it appears to be a negative regulator of integrin {alpha}IIb/ß3 and thereby an inhibitor of integrin-mediated fibrinogen adhesion in platelets and platelet activation (33). Thus, we suggest that DAB2 down-regulation in the stroma may be a proangiogenic mechanism leading to increased platelet activation and as such may establish a microenvironment permissive to vascularization, invasion, and potentially metastasis.

Decreased expression of TBX19, a transcriptional activator of POMC, which encodes polypeptide hormone precursors, such as the {alpha}-melanocyte–stimulating hormone ({alpha}-MSH), was also observed. {alpha}-MSH induces mast cell apoptosis (34) via signaling transduced by its receptor, MC1R. Reduced stromal expression of TBX19 might decrease the production of {alpha}-MSH, thereby promoting mast cell survival and production of inflammatory and angiogenic cytokines that should induce neovascularization.

One example of an epithelial gene that may be responding to nutrient deprivation is KAL1 (Table 3), which encodes anosmin-1, a secreted protein found in the basement membrane and interstitial matrices of multiple tissues during organogenesis (35). Anosmin-1 enhances the cellular response to FGF2 by acting as a modulatory coligand that binds the FGFRI-FGF2-HSPG complex (Fig. 3). Increased KAL1 expression in cervical epithelium should result in protein secretion that locally mediates angiogenesis by directly interacting with underlying stromal cells. This idea is consistent with several studies that have detected an increase in angiogenic growth factors and microvessel density in cervical cancer (21, 36). Thus, this result suggests a complex interplay between stromal fibroblasts and epithelial cells that may provide diffusible signals to initiate proangiogenic cascades responding to a lack of adequate nutrients.

The results presented above suggest that the predominant cellular and physiologic stress at the CIN 2 transition may be nutrient deprivation resulting from the increased metabolic demands of cell division. This seems like a logical assumption because the histologic appearance of CIN 2 shows an increase in the overall number of dividing cells that are beginning to fill the entire space above the basement membrane. These results also may suggest that a fibroblast/epithelial cell communication or interaction may be critical in how CIN 2 cells respond to a shortage of nutrients by inducing angiogenic factors that may be a more generalized intermediate step in carcinogenesis. Thus, we propose that the CIN 2 transition is driven by metabolic demands and thus, suggest a "proangiogenic stroma/epithelium genomic signature."

The proinvasive genomic signature for the CIN 3/SCCA transition. It seems logical to assume that, as cervical cells progress from CIN 2 (proangiogenic) to CIN 3 and SCCA, pressures from the local microenvironment should continue to the point where overcrowding might become the predominant cell stress. If this is the case, then specific genes that function to modulate and overcome this cellular stresses should be activated. In this analysis, seven genes were validated by PCR (Table 3) and five of these fall into a functional class of proteins involved in invasion (DSG3, MMP3, MMP14, BRWD1, and TNFRSF12A). Thus, based on these results, it seemed logical to hypothesize that the predominant cellular stress present at the CIN 3 transition may be due to cellular overcrowding (Fig. 3).

DSG3 is down-regulated in the epithelium and is up-regulated in the stroma at the CIN 3 transition and is a key component of desmosomes. There have been no reports of desmosomal junctions in stromal components; however, the existence of simplified desmoglein-containing desmosomes between adjacent periodontal ligament fibroblasts has been shown (37). Because this type of fibroblast is known to contract (for tooth eruption and repair of periodontal tissue), it was suggested that desmosomes protect gap junctions against cell contraction or strong occlusal forces (38). In light of this evidence, we hypothesize a role for such modified desmosomes in the fibroblasts of reactive stroma that may establish a microenvironment favorable for cell invasion.

In this analysis, two metalloproteinases, MMP3 and MMP14, were increased at the CIN 3 transition (Fig. 3). Metalloproteinases function by degrading specific components of the extracellular matrix and it would seem logical to assume that their increased expression might be responding to tumor cell overcrowding. MMP3 increases from CIN 1 in epithelium and then very sharply increases in CIN 3 and SCCA tumor cells and stroma. MMP3 is an extracellularly secreted matrix metalloproteinase that is overexpressed in a variety of tumor types (39) and has a wide range of specificity for both extracellular matrix and nonmatrix substrates, suggesting extensive effects on tumorigenesis (40). Because MMP3 contributes to angiogenesis by releasing sequestered angiogenic growth factors, it may be responding to the stress of nutrient deprivation as premalignant cells begin to overcrowd. Similar to MMP3, MMP14 is also a marker for invasion that has been shown to confer invasive ability to both normal and neoplastic cells and recently detected in high-grade cervical dysplasia as well as invasive cervical cancer (41). MMP14 activates pro-MMP2, which is overexpressed in premalignant and malignant lesions of the cervix and associates with increased invasive potential (42).

BRWD1, a proposed transcriptional regulator involved in chromatin remodeling via interaction with SMARCA4 (a component of the SWI/SNF complex), is overexpressed in stroma at the CIN 3 transition. SMARCA4 is required to recruit Sp1, AP2, and polymerase II to the MMP2 promoter and enhances transcription (43). These results suggest that induction of BRWD1 in tumor stroma may contribute to the invasive phenotype indirectly by increasing the expression of specific downstream matrix metalloproteinases such as MMP2 (Fig. 3). We also observed up-regulation of the TWEAK receptor, TNFRSF12A (FN14) that is a member of the TNF superfamily. TNFRSF12A stimulates proliferation, migration/invasion, and may also play a role in tumor angiogenesis (44, 45). Thus, we suggest that the induction of metalloproteinases either directly or indirectly at the CIN 3 transition is in response to cellular over crowding.

It is well established that abnormal CIN 3 cells histologically involve the full thickness of the epithelium and this observation along with the results discussed above suggest that the predominant microenvironmental stress may be cellular overcrowding. Thus, it seems logical that these CIN 3 cells should induce factors to overcome this stress by removing any local physical barriers to expansion. In this regard, several differentially regulated genes were identified that may induce an invasive phenotype by altering barriers to tumor cell expansion either directly (MMP3 and MMP14) or indirectly (DSG3, BRWD1, NOMO2, and TNFRSF12A; Fig. 3). Thus, we propose that the transition into CIN 3 be refer to as the "proinvasive genomic signature."


   Conclusions
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Conclusions
References
 
The genomic results in this study were developed based on changes in gene expression in a linear manner that were correlated with histologic changes to formulate an in vivo model for cervical carcinogenesis. This model identifies three apparent critical transitions. The first step appears to be consistent with the normal HPV life cycle after infection, including dysregulation of cell cycle regulation, which would like allow viral replication, and altering the host immune surveillance systems, so as the virally infected cells might evade detection. The CIN 2 transition suggests a stromal/epithelial cell interaction that induces several signals to induce neovacuolization pathways. This would suggest an intermediate step in progression in which the accumulation of CIN 2 cells, in an enclosed microenvironment, is outstripping the local nutrient sources and the logical response would be to activate angiogenesis.

Finally, the transition into CIN 3 coincides with the induction of factors that can disrupt physical barriers to tumor cell expansion and invasion into neighboring tissues. This may be a late step in transformation where abnormal epithelial cells are experiencing overcrowding and the obvious response would be to eliminate any physical barriers to cell expansion. Although this work validates many preexisting ideas for transformation in vitro, the real novelty is identification of the order of these events and their association with specific CIN transitions that allows the construction of the first in vivo model for cervical cancer.


   Acknowledgments
 
Grant support: Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research, the Radiation Oncology Branch, and grants DK51612 and CA82722; NIH grants CA95713 and CA94141; and Howard Hughes Medical Institute, where M.C. Funk is a medical student research fellow.

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.


   Footnotes
 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

D. Gius and M.C. Funk contributed equally to this work.

Current address for E.Y. Chuang: Biomedical Engineering Group, Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan.

5 Unpublished data. Back

Received 1/19/07. Revised 2/28/07. Accepted 4/19/07.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Conclusions
 References
 

  1. Wright TC, Kurman RJ, Ferenczy A. Precancerous lesions of the cervix. In: Kurman RJ, editor. Blaustein's pathology of the female genital tract. New York: Springer-Verlag; 1994. p. 229–77.
  2. Walboomers JM, Jacobs MV, Manos MM, et al. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide. J Pathol 1999;189:12–9.[CrossRef][Medline]
  3. Sawaya GF, Brown AD, Washington AE, Garber AM. Current approaches to cervical-cancer screening. N Engl J Med 2001;344:1603–7.[Free Full Text]
  4. Baseman JG, Koutsky LA. The epidemiology of human papillomavirus infections. J Clin Virol 2005;32:16–24.[CrossRef]
  5. Castle PE, Schiffman M, Herrero R, et al. A prospective study of age trends in cervical human papillomavirus acquisition and persistence in Guanacaste, Costa Rica. J Infect Dis 2005;191:1808–16.[CrossRef][Medline]
  6. Khan MJ, Partridge EE, Wang SS, Schiffman M. Socioeconomic status and the risk of cervical intraepithelial neoplasia grade 3 among oncogenic human papillomavirus DNA-positive women with equivocal or mildly abnormal cytology. Cancer 2005;104:61–70.[CrossRef][Medline]
  7. Eisen MB, Brown PO. DNA arrays for analysis of gene expression. Methods Enzymol 1999;303:179–205.[Medline]
  8. Chuang YY, Chen Y, Gadisetti, et al. Gene expression after treatment with hydrogen peroxide, menadione, or t-butyl hydroperoxide in breast cancer cells. Cancer Res 2002;62:6246–54.[Abstract/Free Full Text]
  9. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 1995;57:289–300.
  10. Li J, Gerhard DS, Zhang Z, et al. Denaturing high-performance liquid chromatography for detecting and typing genital human papillomavirus. J Clin Microbiol 2003;41:5563–71.[Abstract/Free Full Text]
  11. Hanahan D, Folkman J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 1996;86:353–64.[CrossRef][Medline]
  12. Dray M, Russell P, Dalrymple C, et al. p16(INK4a) as a complementary marker of high-grade intraepithelial lesions of the uterine cervix. I. Experience with squamous lesions in 189 consecutive cervical biopsies. Pathology 2005;37:112–24.[Medline]
  13. Zhu C, Bossy-Wetzel E, Jiang W. Recruitment of MKLP1 to the spindle midzone/midbody by INCENP is essential for midbody formation and completion of cytokinesis in human cells. Biochem J 2005;389:373–81.[CrossRef][Medline]
  14. de la Guardia C, Casiano CA, Trinidad-Pinedo J, Báez A. Cenp-F gene amplification and overexpression in head and neck squamous cell carcinomas. Head Neck 2001;23:104–12.[CrossRef][Medline]
  15. Zhang Y, Lu H, Dazin P, Kapila Y. Squamous cell carcinoma cell aggregates escape suspension-induced, p53-mediated anoikis: fibronectin and integrin {alpha}v mediate survival signals through focal adhesion kinase. J Biol Chem 2004;279:48342–9.[Abstract/Free Full Text]
  16. Doorbar J, Foo C, Coleman N, et al. Characterization of events during the late stages of HPV16 infection in vivo using high-affinity synthetic Fabs to E4. Virology 1997;238:40–52.[CrossRef][Medline]
  17. Berger J, Moller DE. The mechanisms of action of PPARs. Annu Rev Med 2002;53:409–35.[CrossRef][Medline]
  18. Jung T-I, Baek W-K, Suh S-I, et al. Down-regulation of peroxisome proliferator-activated receptor {gamma} in human cervical carcinoma. Gynecol Oncol 2005;97:365–73.[CrossRef][Medline]
  19. Parmar S, Platanias L. Interferons: mechanisms of action and clinical applications. Curr Opin Oncol 2003;15:431–9.[CrossRef][Medline]
  20. Walters D, Jelinek D. A role for Janus kinases in crosstalk between ErbB3 and the interferon-{alpha} signaling complex in myeloma cells. Oncogene 2004;23:1197–205.[CrossRef][Medline]
  21. Chang YE, Laimins LA. Microarray analysis identifies interferon-inducible genes and stat-1 as major transcriptional targets of human papillomavirus type 31. J Virol 2000;74:4172–82.
  22. Marvin KW, Fujimoto W, Jetten AM. Identification and characterization of a novel squamous cell-associated gene related to PMP22. J Biol Chem 1995;270:28910–6.[Abstract/Free Full Text]
  23. Wang HT, Kong JP, Ding F, et al. Analysis of gene expression profile induced by EMP-1 in esophageal cancer cells using cDNA microarray. World J Gastroenterol 2003;9:392–8.[Medline]
  24. Wilson HL, Wilson SA, Surprenant A, North RA. Epithelial membrane proteins induce membrane blebbing and interact with the P2X7 receptor C terminus. J Biol Chem 2002;277:34017–23.[Abstract/Free Full Text]
  25. Wilson HL, Francis SE, Dower SK, Crossman DC. Secretion of intracellular IL-1 receptor antagonist (type 1) is dependent on P2X7 receptor activation. J Immunol 2004;173:1202–8.[Abstract/Free Full Text]
  26. Nair A, Venkatraman M, Maliekal TT, Nair B, Karunagaran D. NF-{kappa}B is constitutively activated in high-grade squamous intraepithelial lesions and squamous cell carcinomas of the human uterine cervix. Oncogene 2003;22:50–8.[CrossRef][Medline]
  27. Prusty BK, Das BC. Constitutive activation of transcription factor AP-1 in cervical cancer and suppression of human papillomavirus (HPV) transcription and AP-1 activity in HeLa cells by curcumin. Int J Cancer 2005;113:951–60.[CrossRef][Medline]
  28. Weiske J, Huber O. The histidine triad protein Hint1 triggers apoptosis independent of its enzymatic activity. J Biol Chem 2006;281:27356–66.[Abstract/Free Full Text]
  29. Shi Y, Wang H, Yin Y, et al. Identification and analysis of tumour-associated antigens in hepatocellular carcinoma. Br J Cancer 2005;92:929–34.[CrossRef][Medline]
  30. Untergasser G, Gander R, Lilg C, Lepperdinger G, Plas E, Berger P. Profiling molecular targets of TGF-B1 in prostate fibroblast-to-myofibroblast transdifferentiation. Mech Ageing Dev 2005;126:59–69.[CrossRef][Medline]
  31. Dvorak H. Tumors: Wounds that do not heal. Similarities between tumor stroma generation and wound healing. N Engl J Med 1986;315:1650–9.[Medline]
  32. Sasaki T, Wada T, Kishimoto H, et al. The stress kinase mitogen-activated protein kinase kinase (MKK)7 Is a negative regulator of antigen receptor and growth factor receptor-induced proliferation in hematopoietic cells. J Exp Med 2001;194:757–68.[Abstract/Free Full Text]
  33. Huang C-L, Cheng J-C, Liao C-H, et al. Disabled-2 is a negative regulator of integrin {alpha}IIbß3-mediated fibrinogen adhesion and cell signaling. J Biol Chem 2004;279:42279–89.[Abstract/Free Full Text]
  34. Sarkar A, Sreenivasan Y, Manna SK. {alpha}-Melanocyte-stimulating hormone induces cell death in mast cells: involvement of NF-{kappa}B. FEBS Lett 2003;549:87–93.[CrossRef][Medline]
  35. Hardelin J-P, Julliard AK, Moniot B, et al. Anosmin-1 is a regionally restricted component of basement membranes and interstitial matrices during organogenesis: Implications for the developmental anomalies of X chromosome-linked Kallmann syndrome. Dev Dyn 1999;215:26–44.[CrossRef][Medline]
  36. Coussens LM, Raymond WW, Bergers G, et al. Inflammatory mast cells up-regulate angiogenesis during squamous epithelial carcinogenesis. Genes Dev 1999;13:1382–97.[Abstract/Free Full Text]
  37. Shore R, Berkovitz B, Moxham B. Intercellular contacts between fibroblasts in the periodontal connective tissues of the rat. J Anat 1981;133:67–76.[Medline]
  38. Yamaoka Y, Sawa Y, Ebata N, Yoshida S, Kawasaki T. Desmosomal proteins in cultured and intact human periodontal ligament fibroblasts. Tissue Cell 1999;31:605–9.[CrossRef][Medline]
  39. Lochter A, Galosy S, Muschler J, Freedman N, Werb Z, Bissell MJ. Matrix tetalloproteinase stromelysin-1 triggers a cascade of molecular alterations that leads to stable epithelial-to-mesenchymal conversion and a premalignant phenotype in mammary epithelial cells. J Cell Biol 1997;139:1861–72.[Abstract/Free Full Text]
  40. Egeblad M, Werb Z. New functions for the matrix metalloproteinases in cancer progression. Nat Rev Cancer 2002;2:161–74.[Medline]
  41. Zhai Y, Hotary KB, Nan B, et al. Expression of membrane type 1 matrix metalloproteinase Is associated with cervical carcinoma progression and invasion. Cancer Res 2005;65:6543–50.[Abstract/Free Full Text]
  42. Brummer O, Bohmer G, Hollwitz B, Flemming P, Petry K-U, Kuhnle H. MMP-1 and MMP-2 in the cervix uteri in different steps of malignant transformation—an immunohistochemical study. Gynecol Oncol 2002;84:222–7.[CrossRef][Medline]
  43. Ma X-J, Salunga R, Tuggle JT, et al. Gene expression profiles of human breast cancer progression. Proc Natl Acad Sci U S A 2003;100:5974–9.[Abstract/Free Full Text]
  44. Winkles JA, Tran NL, Berens ME. TWEAK and Fn14: New molecular targets for cancer therapy? Cancer Lett 2006;235:11–7.[CrossRef][Medline]
  45. Tran NL, McDonough WS, Savitch BA, et al. Increased fibroblast growth factor-inducible 14 expression levels promote glioma cell invasion via Rac1 and nuclear factor-{{kappa}}B and correlate with poor patient outcome. Cancer Res 2006;66:9535–42.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Cancer Res.Home page
Z. Zhang, I. Borecki, L. Nguyen, D. Ma, K. Smith, P. C. Huettner, D. G. Mutch, T. J. Herzog, R. K. Gibb, M. A. Powell, et al.
CD83 Gene Polymorphisms Increase Susceptibility to Human Invasive Cervical Cancer
Cancer Res., December 1, 2007; 67(23): 11202 - 11208.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
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 Gius, D.
Right arrow Articles by Rader, J. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Gius, D.
Right arrow Articles by Rader, J. S.

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