A clinical trial was recently conducted to evaluate the safety and efficacy of a selective inhibitor of cyclooxygenase-2 (celecoxib) in hereditary nonpolyposis colon cancer patients. In a randomized, placebo-controlled phase I/II multicenter trial, hereditary nonpolyposis colon cancer patients and gene carriers received either celecoxib at one of two doses or placebo. The goal was to evaluate the effects of these treatment arms on a number of endoscopic and tissue-based biomarker end points after 12 months of treatment. As part of this trial, we analyzed gene expression by cDNA array technology in normal descending (rectal) colonic mucosa of patients before and after treatment with celecoxib or placebo. We found that treatment of patients with celecoxib at recommended clinical doses (200 and 400 mg p.o. bid), in contrast to treatment with placebo, leads to changes in expression of >1,400 genes in the healthy colon, although in general, the magnitude of changes is <2-fold. Twenty-three of 25 pairs of colon biopsies taken before and after celecoxib treatment can be classified correctly by the pattern of gene expression in a leave-one-out cross-validation. Immune response, particularly T- and B-lymphocyte activation and early steps of inflammatory reaction, cell signaling and cell adhesion, response to stress, transforming growth factor-β signaling, and regulation of apoptosis, are the main biological processes targeted by celecoxib as shown by overrepresentation analysis of the distribution of celecoxib-affected genes across Gene Ontology categories. Analysis of possible cumulative effects of celecoxib-induced changes in gene expression indicates that in healthy colon, celecoxib may suppress the immune response and early steps of inflammation, inhibit formation of focal contacts, and stimulate transforming growth factor-β signaling. (Cancer Epidemiol Biomarkers Prev 2006;15(7):1382–91)

Nonsteroidal anti-inflammatory drugs (NSAID) inhibit colorectal carcinogenesis and decrease the risk of death from colorectal cancer as shown in several epidemiologic studies (reviewed in refs. 1-4). Many of the NSAIDs inhibit the activity of cyclooxygenases (COX), which are involved in the synthesis of prostaglandins from arachidonic acid (5). Two major isoforms of COX have been identified: COX-1 is constitutively expressed in most tissues and considered to be responsible for synthesis of prostaglandins that mediate normal physiologic functions, including maintenance of colonic mucosal integrity. COX-2 is usually not detectable under normal conditions in most tissues but can be induced by a wide variety of growth stimulatory and proinflammatory factors. COX-2 is thought to be mainly involved in inflammation, regulation of cell growth, apoptosis, and angiogenesis. However, it has also been shown that COX-2 is expressed constitutively in several tissues, including healthy gastric mucosa (6). It is implicated in the maintenance of mucosal integrity in the healthy stomach as well (review ref. 7).

Several lines of evidence suggest that COX-2 plays an important role in carcinogenesis. It has been shown that the concentration of prostaglandin E2, a major product of the prostaglandin synthesis pathway controlled by COX-1 and COX-2, is higher in various human and animal tumors compared with adjacent normal tissue (8-11), and that some prostaglandins (including prostaglandin E2) stimulate growth of colon tumor cells (12, 13). Importantly, prostaglandin levels in tumor tissues correlate with increased expression of COX-2 but not COX-1 (10, 14, 15). In addition to stimulation of cell proliferation, prostaglandin E2 enhances cell motility and angiogenesis and inhibits apoptosis and immune surveillance, phenomena that can be important for carcinogenesis (2, 16, 17). Prostaglandins act through prostanoid receptors, and homozygous deletion of prostaglandin E2 receptor EP4 reduces formation of aberrant crypt foci (ACF), preneoplastic colon lesions, in azoxymethane-treated mice (18), whereas knockout of the genes encoding prostaglandin E2 receptors EP1 and EP2 decreases the number and size of intestinal polyps in mice (19, 20) and inhibits polyp-related angiogenesis (21). Expression of COX-2 is increased in 40% of colon adenomas and in about 90% of adenocarcinomas (11, 14).

Several observational, case-control and, recently, double-blinded placebo-controlled studies have shown a protective effect of NSAIDs on sporadic colorectal cancer occurrence and mortality; compelling evidence was obtained in clinical trials for patients with familial adenomatous polyposis (FAP) syndrome (reviewed in refs. 2, 22-26). It was shown in several clinical trials that nonspecific NSAIDs aspirin and sulindac and the COX-2-specific inhibitor celecoxib reduce the number and size of colonic polyps in FAP patients (24, 27, 28). Particularly, it was found in a large, double blinded, placebo-controlled study that treatment of FAP patients for 6 months with 400 mg of celecoxib (twice daily) reduced the mean number of colorectal polyps by 28%, with no sign of adverse side effects (29).

The pathway of colorectal carcinogenesis in FAP patients is initiated by the loss of the wild-type allele of the APC gene and leads to abnormal activation of the Wingless/Wnt signaling pathway and to chromosomal instability (30-32). This pathway is considered to account also for ∼85% of sporadic colorectal cancer. Another major pathway of colorectal carcinogenesis is recognized by microsatellite instability and mostly pseudodiploid karyotypes of colorectal cancer and is thought to be responsible for the remaining 15% of sporadic colorectal cancer (33, 34) and for most of the colorectal cancer in patients with hereditary nonpolyposis colorectal cancer (HNPCC) or Lynch syndrome (35-37). HNPCC is an autosomal dominant syndrome characterized by a relatively small number of adenomas, development of colorectal cancer at a relatively young age, frequent occurrence of extracolonic tumors, and predominance of proximal colorectal cancer relative to distal colorectal cancer (38). Patients with HNPCC carry a germ line mutation in one allele of any several DNA mismatch repair genes (most frequently in hMSH2 or hMLH1, but in rare cases, in hPMS1, hPMS2, and hMSH6). It remains unknown whether NSAIDs may be useful as cancer preventive agents in HNPCC.

Celecoxib, a selective COX-2 inhibitor, was approved by the Food and Drug Administration for prevention of colon cancer in patients with FAP, and a clinical trial was recently conducted to evaluate the safety and efficacy of celecoxib in HNPCC patients. In a randomized, placebo-controlled phase I/II multicenter trial, HNPCC gene carriers received either celecoxib at one of two doses or placebo. The goal was to evaluate the effects of these treatment arms on a number of endoscopic and tissue-based biomarker end points after 12 months of treatment (39). As part of this trial, we studied gene expression in normal colonic mucosa of HNPCC patients and gene carriers to determine if there are differences in gene expression profiles at baseline and after treatment of these patients with celecoxib.

Methods of RNA extraction and amplification, cDNA labeling and microarray hybridization, data acquisition, and statistical analysis were described in detail in our publication on differences of gene expression in proximal and distal colon (40) and will be only outlined here.

Patients and Biopsy Samples

All HNPCC patients were known to carry germ line mutations in the hMLH1 or hMSH2 genes or have a family history of colorectal cancer with microsatellite instability, and were enrolled during a multi-institutional (M.D. Anderson Medical Center; National Medical Navy Center, National Cancer Institute; Creighton University; and University of Toronto) “Phase I-II multiple-dose safety and efficacy study of a selective inhibitor of cyclooxygenase-2 (celecoxib) in HNPCC patients and carriers.” Study design and dose selection were based on the results of previous clinical trial in FAP patients (29). Patients were randomized to receive celecoxib at 200 mg p.o. bid, 400 mg p.o. bid, or placebo p.o. bid for 12 months. Protocols used in the study were approved by Institutional Review Boards of the National Cancer Institute and respective clinical performance sites (M.D. Anderson Medical Center; National Naval Medical Center, National Cancer Institute; Creighton University; and University of Toronto). Before enrollment, informed consent was obtained from all of the patients. Standard pinch biopsies of normal colonic mucosa were taken during colonoscopy at the beginning of the study (baseline or pretreatment samples) and 12 months after celecoxib or placebo treatment (posttreatment samples), flash frozen in liquid nitrogen, and stored at −80°C. More than 80% of the cells in standard colon pinch biopsies were colonic crypt cells as determined by H&E staining. Although we have studied almost all available pretreatment biopsies from descending (rectum) and ascending colon (from 50 patients; ref. 40), we decided to analyze posttreatment biopsies only from descending (rectum) colon of randomly selected patients, and it seemed, after code disclosure, that we have data for both pretreatment and posttreatment biopsies of 39 patients. All but one patient have germ line mutations in hMLH1, hMSH2, or PMS1; 14 patients were treated with placebo and 25 with celecoxib.

RNA Extraction and Amplification

Total RNA was isolated from flash-frozen specimens (two standard pinch biopsies per one RNA sample), homogenized with a Disposable Generator and Micro-H Omni Homogenizer in lysis buffer RLT (Qiagen, Chatsworth, CA), and purified using Qiagen's RNeasy Mini kit columns (Qiagen) according to the manufacturer's instructions. mRNA was amplified according to a modified Eberwine's protocol (41). SuperScript II reverse transcriptase (Life Technologies, Gaithersburg, MD) was used for first-strand cDNA synthesis, starting with 5 μg of total RNA. First-strand cDNA synthesis was done with an oligo(dT24) primer containing 5′-T7 promoter-primer. The second strand was synthesized with a mixture of RNase H, DNA Polymerase, and DNA Ligase. Purified cDNA was transcribed in vitro with T7 MEGAscript kit (Ambion, Austin, TX) according to the manufacturer's instructions, and antisense RNA (aRNA) was purified using RNeasy mini spin columns (Qiagen). This and similar protocols of linear aRNA amplification are shown to retain the RNA ratio characteristics of unamplified RNA especially when starting with a relatively large amount (in our case, 5 μg) of total RNA (42-44). We have also used total RNA isolated from HCT116 cells for RNA amplification and found good correlation (Pearson correlation coefficient was >0.9) for gene expression data when arrays were hybridized with total RNA and aRNA samples, which were independently amplified using 5 μg of the total RNA.

cDNA Labeling and Microarray Hybridization

Fluorescent-labeled cDNA was synthesized by reverse transcription of colon aRNA and human testis aRNA (prepared from total testis RNA as described for colon aRNA; Clontech, Inc., Palo Alto, CA) with Superscript II reverse transcriptase (Invitrogen, San Diego, CA) and random oligonucleotide primers in the presence of Cy3-dUTP or Cy5-dUTP (Amersham Pharmacia Biotech, Piscataway, NJ), respectively. For each hybridization experiment, 5 μg of colon aRNA and 6 μg of testis aRNA (common reference) were used to prepare a mixture of labeled cDNAs. Purified labeled colon and testis cDNAs were dissolved in hybridization buffer containing 3.7 × saline-sodium phosphate-EDTA, 2.5 × Denhardt's solution (Sigma, St. Louis, MO), 0.5 μg/μL of poly(A)40-60 (Amersham Pharmacia Biotech) and Human C0t1 DNA (Invitrogen), 0.125 μg/μL yeast tRNA (Sigma), and 10% SDS in TE buffer.

Microarrays containing cDNA clones (from Research Genetics, Inc., Huntsville, AL) spotted on lysine-coated glass slides were obtained from the Advanced Technology Center (National Cancer Institute). Detailed information on printed cDNA can be found on the mAdb web site (http://nciarray.nci.nih.gov/). Microarrays contain 9,128 sequence-verified cDNAs, among which 7,102 represent named genes and 1,179 expressed sequence tag clusters, including 8,556 cDNAs with UniGene cluster ID (mapping to 7,777 unique UniGene clusters).

After hybridization at 65°C for 16 to 18 hours with 40 μL of hybridization mixture, slides were washed at room temperature in 2 × SSC-0.1% SDS (5 minutes), 2 × SSC (2 minutes), 1 × SSC (1 minute), 0.2 × SSC (30 seconds) and dipped in 0.05% SSC. Slides were dried by centrifugation at 150 × g for 5 minutes at room temperature.

All array hybridization and pretreatment image analysis were finished before trial codes were broken.

Data Acquisition and Analysis

Microarrays were scanned with an Axon 4000 laser scanner, and pretreatment image analysis was done with GenePix Pro 3.0 Software (Axon Instruments, Inc., Sunnyvale, CA). Images were also visually inspected, and questionable spots were flagged and excluded from the analysis. Data analysis was done with the BRB-ArrayTools (version 3.2) software package developed by the Biometric Research Branch of the Division of Cancer Treatment and Diagnosis of the National Cancer Institute and The EMMES Corp. (45).

Briefly, background intensities were subtracted, and data were filtered for minimal spot intensity (100 units) in each of the two channels and for missing values (not more than in 20% of arrays). Fluorescence intensity ratio data were log transformed and normalized by subtracting the median log ratio from all log ratios on the array.

The Class comparison and Class Prediction modules of BRB-ArrayTools were used to determine if the pattern of gene expression allowed discrimination of colonic mucosa samples at baseline and after treatment of patients for 12 months with celecoxib or placebo. First, a paired t test was done on average differences in gene expression log ratios of pairs of samples (pretreatment and posttreatment biopsies) from the same patients, to select genes that showed univariately statistically significant (at a stated P, routinely at P < 0.001) differences in expression between baseline and posttreatment colonic mucosa samples of patients treated with celecoxib or placebo. In a case of multiple classes (in the comparison of printing batches, see Results), an F test was used for identification of the differentially expressed genes, which is based on comparing mean log ratios between classes relative to the variation expected in the mean differences (45). The multivariate class label permutation test was used to compute the number of false positives among selected genes (46, 47). Several multivariate classification methods available in BRB ArrayTools 3.2 (compound covariate predictor, diagonal linear discriminant analysis, nearest neighbor and nearest centroid predictors, and support vector machine) were used for classification of pretreatment and posttreatment samples from celecoxib or placebo treated patients and gave comparable results, which will be illustrated with the Compound Covariate Predictor (CCP) and Support Vector Machine Predictor (SVMP) classifiers. CCP was calculated as a linear combination of log ratio differences weighed by univariate t values (48). A positive sign was assigned to t test values for genes that show higher log ratios in posttreatment biopsies and a negative sign for genes that show higher log ratios in pretreatment biopsies. The CCP value was calculated for each pair of samples and, as a classification threshold, the sign (positive or negative) of the CCP was used for classifying. SVMP was calculated as a linear function of the log ratio differences for genes selected in a paired t test with weights estimated by linear kernel support vector machine algorithm to minimize the number of misclassified samples. The misclassification rate was estimated by a leave-one-out cross-validation. Specifically, a pair of samples (from one patient) was omitted, and a CCP and a SVMP were developed from scratch using the remaining samples. The cross-validation procedure includes performing t tests for a selection of genes on arrays for which differences in expression are significant at a stated P for remaining samples, recalculating the CCP, SVMP, and classification threshold for the remaining pairs of samples, and applying the new CCP, SVMP, and threshold values to classify the omitted pair of samples. This was done independently for each omitted pair of samples. The ratio of pairs of samples correctly classified in cross-validation to the total number of sample pairs yields the misclassification rate. Permutation P for a classifier was calculated by performing 2,000 random permutations of class labels and repeating the cross-validation procedure for each permutation. The proportion of random permutations that gave the same or smaller misclassification rate as was obtained with the true class labels is presented as a (permutation) P for the classifier, and P < 0.0005 was reported when no random permutation of the class label was found out of 2,000 permutations with the same or smaller misclassification rate as for the true class labeling.

The Class Comparison and Class Prediction algorithms use t statistics and P cutoffs for selection of genes whose changes in expression are considered significant. We also used another algorithm to determine significant changes in gene expression in colon biopsies after treatment of patients with celecoxib, significance analysis of microarrays (SAM; ref. 49), as implemented in BRB-ArrayTools (45). SAM is based on a comparison of the observed and expected (estimated through random class label permutations) sorted modified t statistics (relative differences) but uses an adjustable threshold, Δ (distance from expected relative differences) and (user defined) false discovery rate as cutoffs for selection of significant genes (49).

To identify biological processes that are most likely targeted in colonic mucosa of patients treated with celecoxib, we structured 1,584 cDNA clones for which changes in expression were statistically significant by SAM (see below) according to Gene Ontology categories (50, 51). Because arrays that were used in this study were printed with a relatively small (n = 9,000) and biased (toward known genes implicated mostly in tumorigenesis) set of cDNAs, we used overrepresentation analysis, in which the number of genes belonging to a Gene Ontology category that were affected by celecoxib treatment (relative to the total number of celecoxib affected genes) is compared with the number of genes printed on the array that belong to the same Gene Ontology category (relative to the total number of genes printed on the array). Two programs were used for overrepresentation analysis: Expression Analysis Systematic Explorer (EASE; ref. 52) and GoMiner (53). Fisher exact probability test is used by GoMiner to estimate the probability of getting an enrichment of a category by chance, and EASE score (derived from the Fisher test penalized by removing one affected gene from the given category and then calculating Fisher exact probability for that category; thus, a category with more genes will be favored against a category with a smaller number of genes) is used in EASE. MatchMiner (54) was used to map UniGene Cluster ID into HUGO symbols (gene symbols approved by Human Genome Organization: http://www.gene.ucl.ac.uk/nomenclature), which are used by GoMiner (see ref. 55).

Real-time PCR

One-step Taqman real-time reverse transcription-PCR was done to study expression of several cDNAs using an ABI Prism 7700 Sequence Detection System. Primers and hybridization FAM-labeled probes were selected with PrimerExpress software (Applied Biosystems, Foster City, CA) by using complete cDNA sequences that have the same UniGene cluster ID as cDNAs printed on array. Six patients (12 RNAs) were randomly selected from 11 patients that had received celecoxib at 400 mg p.o. bid. Taqman Gold reverse transcription-PCR kit (Applied Biosystems) and the manufacturer's protocol (30 minutes at 48°C for reverse transcription reaction, 10 minutes at 95°C for activation of TaqGold Polymerase, and 40 cycles consisting of 15 seconds at 95°C and 1 minute at 60°C) were used, and 50 ng of total RNA were assayed in 25 μL of one-step reverse transcription-PCR reaction mixture with gene-specific primers and probe in triplicates for each sample and each gene. Serial dilutions of a mixture of colon total RNAs from several other colon biopsies were used to calculate PCR efficiency for a gene in the range of 10 to 100 ng of total RNA per reaction. Gene expression was normalized to the amount of ribosomal 18S RNA in a sample, for which reverse transcription-PCR reactions were done on the same 96-well plate in separate wells, together with serial dilutions of colon total RNA to calculate PCR efficiency for ribosomal 18S RNA. Data were analyzed with Q-Gene software (56).

Twenty-five pairs of pretreatment and posttreatment descending colon biopsies from patients exposed to celecoxib (from 14 patients receiving 200 mg bid and 11 patients receiving 400 mg bid of celecoxib) and 14 pairs of biopsies from patients treated with placebo were analyzed on a total of 88 arrays (with some samples analyzed in duplicate). Upon statistical analysis, data on duplicate arrays were averaged and treated as one entry in the analysis. Mean Pearson correlation coefficient (r) for gene expression log ratios between duplicate arrays was 0.84 (95% confidence interval, 0.78-0.91) for 10 different aRNA pairs (for two duplicate arrays of the same printing set hybridized with the same aRNA, r = 0.95).

Paired t test identified 173 genes that showed statistically significant (at a univariate P < 0.001) differences in expression in pretreatment and posttreatment biopsies from 25 patients (58 arrays) who received celecoxib (Table 1). Only 14 genes were differentially expressed (at P < 0.001) in pretreatment and posttreatment biopsies of 14 patients (30 arrays) who received placebo (Supplementary Table S1). The probability of getting 173 and 14 differentially expressed genes out of 9,128 by chance (multivariate class label permutation Ps) is 0.001 and 0.094, respectively. The magnitude of gene expression differences is rather small: maximal fold differences are 1.76 for 92 overexpressed genes and 2.10 for 81 genes underexpressed in posttreatment biopsies of celecoxib-treated patients. For five selected genes, array data showing overexpression after celecoxib treatment were confirmed by real-time reverse transcription-PCR (Supplementary Tables S2 and S3).

Twenty-three of 25 pairs of samples (92%) from celecoxib-treated patients were classified correctly in a leave-one-out cross-validation using the CCP and SVMP. Random permutation of class label showed that the probability of getting as small a cross-validated misclassification rate (8%) as obtained with the true class label by chance (CCP/SVMP permutation Ps) is 0.001 for both classifiers. In contrast, less than half of the samples from placebo-treated patients were correctly classified in cross-validation with CCP (43%, permutation P = 0.589) or SVMP (21%, P = 0.682).

Thus, the patterns of gene expression in pretreatment and posttreatment biopsies from celecoxib-treated patients are different, but no such differences, exceeding those expected by chance, are seen in the placebo-treated patients. Celecoxib-treated colonic mucosa biopsies can be correctly identified by the pattern of gene expression, again in contrast to biopsies from patients who received placebo.

One of the known confounding variables that can affect the results of array gene expression analysis is the use of different printing sets of arrays (i.e., arrays which were printed on different dates with the same cDNA clones). We identified (using F test) genes that showed statistically significant variation in log ratios between printing sets (6,130 and 6,831 genes at P < 0.001 and P < 0.005, respectively), excluded them, and repeated the analysis for the remaining 2,841 or 2,140 genes. Class comparison and class prediction analysis confirm the existence of a distinguishing gene expression signature in colonic mucosa of celecoxib-treated patients, which is robust and can withstand withdrawal of two thirds of the genes present on the array. Using SVMP classifier, up to 96% of pretreatment and posttreatment biopsies from celecoxib-treated patients can be classified correctly in a leave-one-out cross-validation (with one misclassified pair of biopsies out of 25 pairs) based on the expression of 2,841 genes.

Another confounding issue in the analysis is the possibility that the observed differences between the celecoxib- and placebo-treated groups may be a function of the difference in population size. To rule out this possibility, we randomly selected 250 sets of 14 patients from the 25 celecoxib-treated patients. Class comparison and class prediction were done based on the expression of 2,841 genes. Analysis of the distribution of the number of misclassified samples for the 250 sets of 14 patients showed that the probability of getting by chance a misclassification rate such as that for placebo-treated patients is <0.0002. This indicates that the difference in outcome of our analysis of placebo-treated patients compared with celecoxib-treated patients cannot be dismissed as a result of a smaller number of samples in the placebo group.

Among patients who received celecoxib, 14 were given the 200 mg dose, and 11 were given the 400 mg dose. Most (130 genes or 75%) of the 173 genes whose differences in expression are significant at P < 0.001 showed greater fold differences in expression between pretreatment and posttreatment biopsies for patients treated with 400 mg of celecoxib compared with patients who received 200 mg of celecoxib. The Wilcoxon matched pairs signed rank t test shows that the median fold differences in expression of 173 genes for the group of patients treated with 400 mg (1.46) and 200 mg (1.31) of celecoxib are statistically significant (P < 0.0001). These dose-dependent changes in expression are more characteristic for genes overexpressed in posttreatment biopsies (Fig. 1) than for underexpressed genes (Fig. 2): 91% (84 of 92) of overexpressed genes display a dose response effect compared with only 57% (46 of 81) of underexpressed genes.

SAM is less restrictive in defining the significance of gene expression changes, although as shown in an experiment (49) and in simulation of microarray hybridization experiments (57), it can still generate reliable lists of genes that show reproducible changes in expression between compared samples. When gene expression patterns in pretreatment and posttreatment colonic biopsies of patients treated with celecoxib were compared, 1,584 genes were found to be significant by SAM with a median false discovery rate of 9.9% and Δ = 0.63 (for log 2 of relative difference). Of 1,584 genes, 756 were overexpressed, and 828 genes were underexpressed in posttreatment biopsies (Supplementary Table S4). All 173 genes that showed differences in expression significant at univariate P < 0.001 were present in the list of 1,584 genes, and differences in expression of 1,454 of 1,584 genes were significant at P < 0.05. A dose effect is still present: 996 genes (63%) display greater fold differences in gene expression in patients treated with 400 mg of celecoxib compared with patients treated with 200 mg of celecoxib. Differences in the median fold difference are significant by Wilcoxon matched pairs signed rank t test for the whole set of 1,584 genes (1.30 and 1.24 for patients who received 400 and 200 mg of celecoxib, respectively; P < 0.0001), and in a separate comparison for 756 genes overexpressed in posttreatment biopsies (median fold differences = 1.31 and 1.19; P < 0.0001) but not for 828 genes underexpressed in posttreatment biopsies.

The 1,584 cDNA clones for which changes in posttreatment biopsies of celecoxib-treated patients are identified as significant by SAM are mapped to 1,448 unique UniGene Cluster ID, among which 696 (48%) are overexpressed, and 752 (52%) are underexpressed after celecoxib treatment. Analysis of the distribution of these genes across Gene Ontology categories indicates that immune response, including immune cell activation and inflammation (as part of “response to wounding” category; EASE score for the category “inflammatory response” by itself is 0.065), response to stress, cell adhesion, and regulation of apoptosis are the main biological processes affected by changes in gene expression in colonic mucosa of patients treated with celecoxib (Supplementary Table S5). Genes controlling cell communication and cell-to-cell signaling are also overrepresented among genes affected by celecoxib, and changes in their activity are likely responsible for shaping the list of most affected molecular functions, which include receptor binding and cytokine and signal transducer activity (Supplementary Table S6). There is also an enrichment in the number of genes whose expression has been changed by celecoxib that encode proteins residing in the plasma membrane and proteins that form the extracellular matrix (Supplementary Table S7).

The data show that treatment of patients with celecoxib at recommended clinical doses (200 and 400 mg p.o. bid), in contrast to treatment with placebo, leads to changes in expression of a substantial number of genes in healthy colon.

It is generally accepted that NSAIDs and, particularly, COX-2-specific inhibitors prevent colon tumorigenesis by inhibiting tumor cell proliferation and tumor-related angiogenesis, reducing carcinogen activation and inflammation, enhancing immune surveillance, and promoting apoptosis in tumor cells through COX-dependent and COX-independent mechanisms (1, 58-61). However, the relative importance of each of these effects still remains to be established and, in addition, because most of the data supporting these conclusions were obtained in experiments using tumor cells, it is not known which, if any, effects occur in normal colonic mucosa of healthy patients.

The normal colorectal mucosa has three main elements: a single layer of surface epithelial cells forming crypts (composed of absorptive cells, goblet cells, undifferentiated precursor cells, specialized enteroendocrine cells, and rare Paneth cells, which are supported by a thin basement membrane of collagen and other proteins), lamina propria (that extends between crypts and contains many cell types: fibroblasts, lymphocytes, plasma cells, eosinophils, macrophages, and mast cells), and muscularis mucosa. The colonic epithelium is an actively proliferating and dynamic system in which cells originating from stem cells on the bottom of crypts migrate toward the intercrypt table while undergoing differentiation before being shed into the lumen. Standard colon pinch biopsies contain mostly crypt cells (80%), some lamina propria cells, and some colon-associated lymphoid cells. Thus, the observed changes in gene expression after celecoxib treatment may originate from changes in different cells, reflecting changes in a particular cell characteristic, or in the balance of proliferation and differentiation processes in colonic mucosa.

In this study, we only considered a subset of the total set of human genome transcripts. It is possible that a study of a larger number of transcripts might reveal genes more significantly up-regulated and down-regulated. However, even with this limited set, some interesting relationships emerge:

Immune Response, Inflammation, and Angiogenesis. Among the 103 genes belonging to the Gene Ontology category “immune response” and affected by celecoxib, 22 are overexpressed, and 81 are underexpressed in posttreatment biopsies of celecoxib-treated patients (Supplementary Table S8), indicating that immune reactions may be somewhat suppressed in these patients. The disparity in the number of overexpressed and underexpressed genes is even more pronounced for genes controlling the cell defense response (Supplementary Table S9). Of 30 genes affected by celecoxib that belong to Gene Ontology category “inflammatory response,” 21 are underexpressed (Supplementary Table S10), again indicating that the inflammatory reaction, especially in the early steps, could be muted in the colonic mucosa of celecoxib-treated patients. Among genes affected by celecoxib treatment, five encode proteins controlling early steps of the inflammation reaction: CD40 (TNFSF5), LCK and zap70 protein kinases, IL2, and IL4, with all five being underexpressed (Supplementary Tables S4 and S10). Angiogenesis occurs during wound healing and in tumorigenesis, and COX-2-stimulated tumor angiogenesis can influence tumor development not only in late but also in the very early stages (62), whereas suppression of angiogenesis is considered an important component of the cancer-preventing effect of COX-2 inhibitors (reviewed in ref. 63). Nine genes belonging to the “angiogenesis” Gene Ontology category are affected by celecoxib, with five being overexpressed, and four being underexpressed (Supplementary Table S11). The hierarchical structure of the Gene Ontology categories does not reflect, naturally, the complexity of regulation of biological processes. For example, an important role in angiogenesis belongs to cell adhesion proteins, particularly to integrins (64, 65). Integrins α1, α2, and β3 are overexpressed in tumor-associated angiogenesis, known to be suppressed by NSAID, and all of them are underexpressed after celecoxib treatment (Supplementary Table S10).

Cell Adhesion. Cell-cell and cell-extracellular matrix adhesion play important roles in the maintenance of mechanical and functional integrity of colonic mucosa and in control of cell polarity, proliferation, and differentiation (66), and proteins implicated in cell adhesion are frequently targeted in colon tumorigenesis (67). Of 77 genes that encode proteins in the “cell adhesion” Gene Ontology category, 29 are overexpressed, and 48 are underexpressed after celecoxib treatment (Supplementary Table S13). Forty of these genes (16 overexpressed and 24 underexpressed after celecoxib treatment) encode proteins that target focal contacts, cellular structures linking intracellular actin cytoskeleton to extracellular matrix at points of cell contacts with substratum (Supplementary Table S1).

Analysis of possible effects of changes in expression of these 40 genes done with the PathwayAssist 2.5/MedScan program (Ariadne Genomics, Inc., Rockville, MD) and then manually reviewed indicates that most of the genes overexpressed after celecoxib treatment encode proteins that inhibit formation of focal contacts, whereas genes underexpressed by celecoxib encode mainly proteins that stimulate formation of focal contacts. Cell signaling, including signaling involved in reorganization of cell cytoskeleton and focal contacts, is mainly controlled by a number of different intracellular protein tyrosine kinases and usually does not require changes in gene expression and may supersede such changes (68). Nevertheless, if celecoxib-induced changes in gene expression are translated into corresponding changes in protein quantities and activities, the cumulative effect of celecoxib treatment is most likely the inhibition of cell-cell and cell-extracellular matrix adhesion in colonic mucosa. The observation that subapoptotic doses of sulindac sulfide cause loss of focal adhesion contacts and reorganization of the actin cytoskeleton in human colon carcinoma cells (69) supports this possibility.

Transforming Growth Factor-β Signaling Pathway. The genes for both type I and type II transforming growth factor-β (TGF-β) receptors and bone morphogenetic protein receptors are overexpressed and so is the gene for SMAD4, the common mediator of TGF-β receptor signaling, whereas the gene for SMAD7, an antagonist of TGF-β/bone morphogenetic protein receptor type I signaling, is underexpressed after celecoxib treatment (Supplementary Table S14). Increased activity of the TGF-β receptor signaling pathway may be responsible for the observed overexpression of the gene for SRY-interacting protein (SIP1), a transcription factor known to be induced by TGF-β signaling (70). It is interesting that rat epithelial cells genetically engineered to overexpress COX-2 have reduced TGF-β receptor type II level, which can be increased by treatment of the cells with sulindac sulfide (71).

Obviously, observed changes in gene expression are mediated by changes in the activity of transcription factors, with many genes encoding transcription factors also being affected by celecoxib treatment.

Cell Proliferation and Apoptosis. There are 139 genes in the Gene Ontology category “cell proliferation,” which are affected by celecoxib treatment (Supplementary Table S15), although the enrichment of affected genes of this category is not significant according to EASE analysis (EASE score 0.45). Among them, 80 are overexpressed, and 59 are underexpressed in posttreatment biopsies of celecoxib-treated patients, reflecting modulation of the balance between cell proliferation and differentiation rather then stimulation of cell proliferation. Consideration of genes implicated in cell death and apoptosis regulation leads to the same conclusion: 25 of the genes encoding apoptosis-inducing and antiapoptotic proteins are induced, and 35 are underexpressed after celecoxib treatment (Supplementary Table S16). It is interesting that genes encoding proteins that are involved in induction of apoptosis by extracellular signals (DAP3, DPF2, and MAP3K5) are induced by celecoxib, whereas genes encoding proteins implicated in induction of apoptosis via death domain receptors [BID, FADD, and TNFRSF10A (DR4)] are suppressed by celecoxib.

The results show that celecoxib, at clinically approved doses, and in contrast to placebo, induces changes in expression of a substantial number of genes. The targeted biological processes fit reasonably to the current models of cancer preventive action of NSAIDs. In fact, a comparison of gene expression changes induced by celecoxib in normal colonic mucosa with changes in gene expression observed in tumor cells in culture treated with NSAID reveals similarly affected genes. Among 10 known genes whose expression was changed after treatment of human rectal carcinoma cells with NSAIDs (72), two that were induced by the COX-2 inhibitor NS-398 are found in the list of 1,448 genes affected by celecoxib, and both are overexpressed in posttreatment biopsies of celecoxib-treated patients. Three genes induced by celecoxib treatment are present in both our list of 1,448 genes and the list of 31 known (with UniGene ID) genes induced by treatment of human colorectal cell lines with sulindac sulfide (73). Even when the observed gene expression changes were compared with changes induced by celecoxib treatment of a human mammary epithelial cell line (obviously with significant tissue-specific differences in gene expression compared with colonic mucosa cells; ref. 74), there were still 31 similarly reacting genes (21 induced and 10 suppressed). In experiments with cultured tumor cells, the conditions of treatment with NSAIDs are usually selected to achieve significant if not complete cell growth arrest and eventually apoptotic cell death. Gene expression changes observed under such conditions define cancer therapeutic effects of NSAIDs rather than possible cancer preventive action.

We have recently completed a study of patterns of gene expression in human colonic aberrant crypt foci (ACFs). ACFs are considered to be an early preneoplastic lesion in the colon (75), and the ability to decrease the frequency of ACF after carcinogen treatment is used routinely in experimental models of colon cancer to estimate the cancer preventive effect of different agents (76). It is particularly well established that NSAIDs (77-79) and celecoxib (80-83) decrease the frequency of carcinogen-induced ACFs in the colon in animal models. It has also been shown that the number of ACF in patients who received sulindac sulfide is decreased (84). It seems that we can distinguish ACFs from normal colonic mucosa by the pattern of gene expression. Among the 90 genes for which changes in expression in ACF compared with normal colonic mucosa were identified by SAM (false discovery rate = 9.9%, Δ = 0.547) as significant, 22 genes are present in the list of 1,584 genes whose expression was affected by celecoxib treatment. However, for 21 of 22 genes, the observed changes in expression after celecoxib treatment and in ACF compared with normal mucosa were in the opposite direction: genes that were overexpressed in ACF were underexpressed after celecoxib treatment and vice versa (Table 6 in the accompanying article). It is possible that the cancer preventive effects of celecoxib are directly reflected by the alterations in the expression of a subset of genes required in early stages of cellular transformation.

Grant support: Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.

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.

Note: Supplementary data for this article are available at Cancer Epidemiology Biomarkers and Prevention Online (http://cebp.aacrjournals.org/).

We thank Richard Simon (Biometric Research Branch, Center for Cancer Research, National Cancer Institute) for helpful advices.

1
Dannenberg AJ, Subbaramaiah K. Targeting cyclooxygenase-2 in human neoplasia: rationale and promise.
Cancer Cell
2003
;
4
:
431
–6.
2
Gupta RA, Dubois RN. Colorectal cancer prevention and treatment by inhibition of cyclooxygenase-2.
Nat Rev Cancer
2001
;
1
:
11
–21.
3
Ricchi P, Pignata S, Iaffaioli RV, Daniele B. Cyclo-oxygenase inhibition in colorectal adenomas and cancer.
J Clin Gastroenterol
2003
;
37
:
281
–7.
4
Steele VE, Hawk ET, Viner JL, Lubet RA. Mechanisms and applications of non-steroidal anti-inflammatory drugs in the chemoprevention of cancer.
Mutat Res
2003
;
523–524
:
137
–44.
5
Smith WL, DeWitt DL, Garavito RM. Cyclooxygenases: structural, cellular, and molecular biology.
Annu Rev Biochem
2000
;
69
:
145
–82.
6
Zimmermann KC, Sarbia M, Schror K, Weber A-A. Constitutive cyclooxygenase-2 expression in healthy human and rabbit gastric mucosa.
Mol Pharmacol
1998
;
54
:
536
–40.
7
Peskar BM, Maricic N, Gretzera B, Schuligoi R, Schmassmann A. Role of cyclooxygenase-2 in gastric mucosal defense.
Life Sci
2001
;
69
:
2993
–3003.
8
Bennett A, Del Tacca M. Proceedings: prostaglandins in human colonic carcinoma.
Gut
1975
;
16
:
409
.
9
Bennett A, Tacca MD, Stamford IF, Zebro T. Prostaglandins from tumours of human large bowel.
Br J Cancer
1977
;
35
:
881
–4.
10
Rigas B, Goldman IS, Levine L. Altered eicosanoid levels in human colon cancer.
J Lab Clin Med
1993
;
122
:
518
–23.
11
Sano H, Kawahito Y, Wilder R, et al. Expression of cyclooxygenase-1 and -2 in human colorectal cancer.
Cancer Res
1995
;
55
:
3785
–9.
12
Qiao L, Kozoni V, Tsioulias GJ, et al. Selected eicosanoids increase the proliferation rate of human colon carcinoma cell lines and mouse colonocytes in vivo.
Biochimica et Biophysica Acta (BBA) - Lipids and Lipid Metabolism
1995
;
1258
:
215
–23.
13
Sheng H, Shao J, Washington MK, DuBois RN. Prostaglandin E2 increases growth and motility of colorectal carcinoma cells.
J Biol Chem
2001
;
276
:
18075
–81.
14
Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, DuBois RN. Up-regulation of cyclooxygenase 2 gene expression in human colorectal adenomas and adenocarcinomas.
Gastroenterology
1994
;
107
:
1183
–8.
15
Fujita T, Matsui M, Takaku K, et al. Size- and invasion-dependent increase in cyclooxygenase 2 levels in human colorectal carcinomas.
Cancer Res
1998
;
58
:
4823
–6.
16
Fujino H, Regan JW. Prostaglandin F(2α) stimulation of cyclooxygenase-2 promoter activity by the Fp(B) prostanoid receptor.
Eur J Pharmacol
2003
;
465
:
39
–41.
17
Majima M, Amano H, Hayashi I. Prostanoid receptor signaling relevant to tumor growth and angiogenesis.
Trends Pharmacol Sci
2003
;
24
:
524
–9.
18
Mutoh M, Watanabe K, Kitamura T, et al. Involvement of prostaglandin E receptor subtype Ep4 in colon carcinogenesis.
Cancer Res
2002
;
62
:
28
–32.
19
Watanabe K, Kawamori T, Nakatsugi S, et al. Role of the prostaglandin E receptor subtype Ep1 in colon carcinogenesis.
Cancer Res
1999
;
59
:
5093
–6.
20
Sonoshita M, Takaku K, Sasaki N, et al. Acceleration of intestinal polyposis through prostaglandin receptor Ep2 in Apc(Delta 716) knockout mice.
Nat Med
2001
;
7
:
1048
–51.
21
Seno H, Oshima M, Ishikawa TO, et al. Cyclooxygenase 2- and prostaglandin E(2) receptor Ep(2)-dependent angiogenesis in Apc(Delta716) mouse intestinal polyps.
Cancer Res
2002
;
62
:
506
–11.
22
Kubba AK. Non steroidal anti-inflammatory drugs and colorectal cancer: is there a way forward?
Eur J Cancer
1999
;
35
:
892
–901.
23
Baron JA, Sandler RS. Nonsteroidal anti-inflammatory drugs and cancer prevention.
Annu Rev Med
2000
;
51
:
511
–23.
24
Sandler RS, Halabi S, Baron JA, et al. A randomized trial of aspirin to prevent colorectal adenomas in patients with previous colorectal cancer.
N Engl J Med
2003
;
348
:
883
–90.
25
Marnett LJ, DuBois RN. Cox-2: a target for colon cancer prevention.
Annu Rev Pharmacol Toxicol
2002
;
42
:
55
–80.
26
Ricchi P, Zarrilli R, Di Palma A, Acquaviva AM. Nonsteroidal anti-inflammatory drugs in colorectal cancer: from prevention to therapy.
Br J Cancer
2003
;
88
:
803
–7.
27
Labayle D, Boyer J, Drouhin F, Zarka Y, Fischer D. Sulindac in familial adenomatous polyposis.
Lancet
1994
;
343
:
417
–8.
28
Giardiello FM. Sulindac and polyp regression.
Cancer Metastasis Rev
1994
;
13
:
279
–83.
29
Steinbach G, Lynch PM, Phillips RKS, et al. The effect of celecoxib, a cyclooxygenase-2 inhibitor, in familial adenomatous polyposis.
N Engl J Med
2000
;
342
:
1946
–52.
30
Fearon ER, Vogelstein B. A genetic model for colorectal tumorigenesis.
Cell
1990
;
61
:
759
–67.
31
Kinzler KW, Vogelstein B. Lessons from hereditary colorectal cancer.
Cell
1996
;
87
:
159
–70.
32
Grady WM, Markowitz SD. Genetic and epigenetic alterations in colon cancer.
Annu Rev Genomics Hum Genet
2002
;
3
:
101
–28.
33
Kim H, Piao Z, Kim JW, et al. Expression of Hmsh2 and Hmlh1 in colorectal carcinomas with microsatellite instability.
Pathol Res Pract
1998
;
194
:
3
–9.
34
Liu B, Nicolaides NC, Markowitz S, et al. Mismatch repair gene defects in sporadic colorectal cancers with microsatellite instability.
Nat Genet
1995
;
9
:
48
–55.
35
Aaltonen LA, Peltomaki P, Mecklin JP, et al. Replication errors in benign and malignant tumors from hereditary nonpolyposis colorectal cancer patients.
Cancer Res
1994
;
54
:
1645
–8.
36
Aaltonen LA, Peltomaki P, Leach FS, et al. Clues to the pathogenesis of familial colorectal cancer.
Science
1993
;
260
:
812
–6.
37
Konishi M, Kikuchi-Yanoshita R, Tanaka K, et al. Molecular nature of colon tumors in hereditary nonpolyposis colon cancer, familial polyposis, and sporadic colon cancer.
Gastroenterology
1996
;
111
:
307
–17.
38
Lynch HT, Lynch J. Lynch syndrome: genetics, natural history, genetic counseling, and prevention.
J Clin Oncol
2000
;
18
:
19
–31S.
39
Annie Yu HJ, Lin KM, Ota DM, Lynch HT. Hereditary nonpolyposis colorectal cancer: preventive management.
Cancer Treat Rev
2003
;
29
:
461
–70.
40
Glebov OK, Rodriguez LM, Nakahara K, et al. Distinguishing right from left colon by the pattern of gene expression.
Cancer Epidemiol Biomarkers Prev
2003
;
12
:
755
–62.
41
Van Gelder RN, von Zastrow ME, Yool A, Dement WC, Barchas JD, Eberwine JH. Amplified RNA synthesized from limited quantities of heterogeneous cDNA.
Proc Natl Acad Sci U S A
1990
;
87
:
1663
–7.
42
Gomes LI, Silva RL, Stolf BS, et al. Comparative analysis of amplified and nonamplified RNA for hybridization in cDNA microarray.
Anal Biochem
2003
;
321
:
244
–51.
43
Nygaard V, Loland A, Holden M, et al. Effects of mRNA amplification on gene expression ratios in cDNA experiments estimated by analysis of variance.
BMC Genomics
2003
;
4
:
11
.
44
Wang E, Miller LD, Ohnmacht GA, Liu ET, Marincola FM. High-fidelity mRNA amplification for gene profiling.
Nat Biotechnol
2000
;
18
:
457
–9.
45
Simon R, Lam A. Brb-Arraytools user guide, version 3.2., Biometric Research Branch, National Cancer Institute http://linusncinihgov/brb; 2004.
46
Korn EL, McShane LM, Troendle JF, Rosenwald A, Simon R. Identifying pre-post chemotherapy differences in gene expression in breast tumours: a statistical method appropriate for this aim.
Br J Cancer
2002
;
86
:
1093
–6.
47
Reiner A, Yekutieli D, Benjamini Y. Identifying differentially expressed genes using false discovery rate controlling procedures.
Bioinformatics
2003
;
19
:
368
–75.
48
Radmacher MD, McShane LM, Simon R. A paradigm for class prediction using gene expression profiles.
J Comput Biol
2002
;
9
:
505
–11.
49
Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response.
Proc Natl Acad Sci U S A
2001
;
98
:
5116
–21.
50
Harris MA, Clark J, Ireland A, et al. The Gene Ontology (GO) database and informatics resource.
Nucleic Acids Res
2004
;
32
Database issue:
D258
–61.
51
Ashburner M, Ball CA, Blake JA, et al. Gene Ontology: tool for the unification of biology. The Gene Ontology Consortium.
Nat Genet
2000
;
25
:
25
–9.
52
Hosack DA, Dennis G, Jr., Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE.
Genome Biol
2003
;
4
:
R70
.
53
Zeeberg BR, Feng W, Wang G, et al. Gominer: a resource for biological interpretation of genomic and proteomic data.
Genome Biol
2003
;
4
:
R28
.
54
Bussey KJ, Kane D, Sunshine M, et al. Matchminer: a tool for batch navigation among gene and gene product identifiers.
Genome Biol
2003
;
4
:
R27
.
55
Zeeberg BR, Riss J, Kane DW, et al. Mistaken identifiers: gene name errors can be introduced inadvertently when using Excel in bioinformatics.
BMC Bioinformatics
2004
;
5
:
80
.
56
Muller PY, Janovjak H, Miserez AR, Dobbie Z. Processing of gene expression data generated by quantitative real-time RT-PCR.
Biotechniques
2002
;
32
:
1372
–4, 1376, 1378–9.
57
Singhal S, Kyvernitis CG, Johnson SW, Kaiser LR, Liebman MN, Albelda SM. Microarray data simulator for improved selection of differentially expressed genes.
Cancer Biol Ther
2003
;
2
:
383
–91.
58
Subbaramaiah K, Dannenberg AJ. Cyclooxygenase 2: a molecular target for cancer prevention and treatment.
Trends Pharmacol Sci
2003
;
24
:
96
–102.
59
Tegeder I, Pfeilschifter J, Geisslinger G. Cyclooxygenase-independent actions of cyclooxygenase inhibitors.
FASEB J
2001
;
15
:
2057
–72.
60
Gasparini G, Longo R, Sarmiento R, Morabito A. Inhibitors of cyclo-oxygenase 2: a new class of anticancer agents?
Lancet Oncol
2003
;
4
:
605
–15.
61
Iniguez MA, Rodriguez A, Volpert OV, Fresno M, Redondo JM. Cyclooxygenase-2: a therapeutic target in angiogenesis.
Trends Mol Med
2003
;
9
:
73
–8.
62
Chang SH, Liu CH, Conway R, et al. Role of prostaglandin E2-dependent angiogenic switch in cyclooxygenase 2-induced breast cancer progression.
Proc Natl Acad Sci U S A
2004
;
101
:
591
–6.
63
Wang D, DuBois RN. Cyclooxygenase 2-derived prostaglandin E2 regulates the angiogenic switch.
Proc Natl Acad Sci U S A
2004
;
101
:
415
–6.
64
Dormond O, Ruegg C. Regulation of endothelial cell integrin function and angiogenesis by COX-2, cAMP and protein kinase A.
Thromb Haemost
2003
;
90
:
577
–85.
65
Ruegg C, Dormond O, Mariotti A. Endothelial cell integrins and COX-2: mediators and therapeutic targets of tumor angiogenesis.
Biochim Biophys Acta
2004
;
1654
:
51
–67.
66
Efstathiou JA, Pignatelli M. Modulation of epithelial cell adhesion in gastrointestinal homeostasis.
Am J Pathol
1998
;
153
:
341
–7.
67
Buda A, Pignatelli M. Cytoskeletal network in colon cancer: from genes to clinical application.
Int J Biochem Cell Biol
2004
;
36
:
759
–65.
68
Schlaepfer DD, Mitra SK, Ilic D. Control of motile and invasive cell phenotypes by focal adhesion kinase.
Biochimica et Biophysica Acta (BBA) - Molecular Cell Research
2004
;
1692
:
77
–102.
69
Weyant MJ, Carothers AM, Bertagnolli ME, Bertagnolli MM. Colon cancer chemopreventive drugs modulate integrin-mediated signaling pathways.
Clin Cancer Res
2000
;
6
:
949
–56.
70
Lin SY, Elledge SJ. Multiple tumor suppressor pathways negatively regulate telomerase.
Cell
2003
;
113
:
881
–9.
71
Tsujii M, DuBois RN. Alterations in cellular adhesion and apoptosis in epithelial cells overexpressing prostaglandin endoperoxide synthase 2.
Cell
1995
;
83
:
493
–501.
72
Zhang Z, DuBois RN. Detection of differentially expressed genes in human colon carcinoma cells treated with a selective COX-2 inhibitor.
Oncogene
2001
;
20
:
4450
–6.
73
Bottone FG, Jr., Martinez JM, Alston-Mills B, Eling TE. Gene modulation by COX-1 and COX-2 specific inhibitors in human colorectal carcinoma cancer cells.
Carcinogenesis
2004
;
25
:
349
–57.
74
Levitt RJ, Buckley J, Blouin MJ, Schaub B, Triche TJ, Pollak M. Growth inhibition of breast epithelial cells by celecoxib is associated with upregulation of insulin-like growth factor binding protein-3 expression.
Biochem Biophys Res Commun
2004
;
316
:
421
–8.
75
Mori H, Yamada Y, Hirose Y, et al. Chemoprevention of large bowel carcinogenesis; the role of control of cell proliferation and significance of β-catenin-accumulated crypts as a new biomarker.
Eur J Cancer Prev
2002
;
11
Suppl 2:
S71
–5.
76
Pereira MA. Prevention of colon cancer and modulation of aberrant crypt foci, cell proliferation, and apoptosis by retinoids and NSAIDS.
Adv Exp Med Biol
1999
;
470
:
55
–63.
77
Charalambous D, Farmer C, O'Brien PE. Sulindac and indomethacin inhibit formation of aberrant crypt foci in the colons of dimethyl hydrazine treated rats.
J Gastroenterol Hepatol
1996
;
11
:
88
–92.
78
Kishimoto Y, Takata N, Jinnai T, et al. Sulindac and a cyclooxygenase-2 inhibitor, Etodolac, increase APC mRNA in the colon of rats treated with azoxymethane.
Gut
2000
;
47
:
812
–9.
79
Roy HK, Karolski WJ, Ratashak A. Distal bowel selectivity in the chemoprevention of experimental colon carcinogenesis by the non-steroidal anti-inflammatory drug nabumetone.
Int J Cancer
2001
;
92
:
609
–15.
80
Buecher B, Thouminot C, Menanteau J, et al. Fructooligosaccharide associated with celecoxib reduces the number of aberrant crypt foci in the colon of rats.
Reprod Nutr Dev
2003
;
43
:
347
–56.
81
Brown WA, Skinner SA, Malcontenti-Wilson C, et al. Non-steroidal anti-inflammatory drugs with different cyclooxygenase inhibitory profiles that prevent aberrant crypt foci formation but vary in acute gastrotoxicity in a rat model.
J Gastroenterol Hepatol
2000
;
15
:
1386
–92.
82
Yamada Y, Yoshimi N, Hirose Y, et al. Suppression of occurrence and advancement of β-catenin-accumulated crypts, possible premalignant lesions of colon cancer, by selective cyclooxygenase-2 inhibitor, celecoxib.
Jpn J Cancer Res
2001
;
92
:
617
–23.
83
Rao CV, Indranie C, Simi B, Manning PT, Connor JR, Reddy BS. Chemopreventive properties of a selective inducible nitric oxide synthase inhibitor in colon carcinogenesis, administered alone or in combination with celecoxib, a selective cyclooxygenase-2 inhibitor.
Cancer Res
2002
;
62
:
165
–70.
84
Takayama T, Katsuki S, Takahashi Y, et al. Aberrant crypt foci of the colon as precursors of adenoma and cancer.
N Engl J Med
1998
;
339
:
1277
–84.