Purpose: The homeodomain transcription factor CDX2 is a relatively specific immunohistochemical marker for gastrointestinal carcinoma. However, no study has comprehensively examined the relationship between CDX2 expression in colon cancer and clinical, pathologic, prognostic, and molecular features, including microsatellite instability and CpG island methylator phenotype (CIMP).

Experimental Design: Utilizing 621 colorectal cancers with clinical outcome and molecular data, CDX2 loss was detected in 183 (29%) tumors by immunohistochemistry.

Results: In multivariate logistic regression analysis, CDX2 loss was associated with female gender [odds ratio (OR), 3.32; P < 0.0001], CIMP-high (OR, 4.42; P = 0.0003), high tumor grade (OR, 2.69; P = 0.0085), stage IV disease (OR, 2.03; P = 0.019), and inversely with LINE-1 hypomethylation (for a 30% decline; OR, 0.33; P = 0.0031), p53 expression (OR, 0.55; P = 0.011), and β-catenin activation (OR, 0.60; P = 0.037), but not with body mass index, tumor location, microsatellite instability, BRAF, KRAS, PIK3CA, p21, or cyclooxygenase-2. CDX2 loss was not independently associated with patient survival. However, the prognostic effect of CDX2 loss seemed to differ according to family history of colorectal cancer (Pinteraction = 0.0094). CDX2 loss was associated with high overall mortality (multivariate hazard ratio, 2.40; 95% CI, 1.28-4.51) among patients with a family history of colorectal cancer; no such association was present (multivariate hazard ratio, 0.97; 95% CI, 0.66-1.41) among patients without a family history of colorectal cancer.

Conclusions: CDX2 loss in colorectal cancer is independently associated with female gender, CIMP-high, high-level LINE-1 methylation, high tumor grade, and advanced stage. CDX2 loss may be associated with poor prognosis among patients with a family history of colorectal cancer.

Translational Relevance

CDX2 plays a critical role in development and differentiation of intestinal epithelium and is clinically useful as a relatively specific marker for colorectal cancer. Understanding of CDX2 expression patterns in colorectal cancer is therefore important in clinical and pathology practice. We used a database of 621 colorectal cancers in two independent, prospective cohort studies with clinical information, adequate follow-up, and key molecular data. To our knowledge, this is the first study to comprehensively examine the relationship between CDX2 expression and clinicopathologic, prognostic, and molecular variables, including microsatellite instability, CpG island methylator phenotype, KRAS, BRAF, PIK3CA, p53, and β-catenin. Our results will alert pathologists and clinicians that sensitivity of CDX2 testing to identify a colorectal origin for metastatic tumors depends on patient and tumor characteristics, and that the prognostic significance of CDX2 loss is related to a family history of colorectal cancer.

The Caudal-related homeodomain transcription factor CDX2 regulates development and differentiation of intestinal epithelium (1). Immunohistochemical detection of CDX2 expression is clinically useful as a relatively specific marker for epithelial neoplasms of the gastrointestinal tract, particularly colon and rectum (25). Different lines of evidence suggest that CDX2 may suppress colorectal tumorigenesis (69) and CDX2 expression is often lost in colorectal cancers with high tumor grade, advanced tumor stage, or microsatellite instability (MSI-high; refs. 10, 11). Because colorectal cancer represents a heterogeneous group of tumors with diverse genetic/epigenetic signatures (12), the frequency of CDX2 loss likely depends on clinical and molecular features of colorectal cancer. Understanding features associated with CDX2 loss is crucially important in clinical and pathology practice. No previous study has comprehensively examined clinical, pathologic, and molecular features that influence the frequency of CDX2 expression (i.e., sensitivity of CDX2 testing) in colorectal cancer.

Promoter CpG island methylation is an important mechanism for silencing tumor suppressor genes in the carcinogenic process (13). A subset of colorectal cancers exhibits widespread promoter CpG island methylation, called the CpG island methylator phenotype (CIMP; ref. 14). CIMP-high colorectal cancers are associated with female gender, proximal tumor location, high tumor grade, MSI-high, BRAF mutation, and inactive WNT/β-catenin (1517). MSI and CIMP status reflect global genomic and epigenomic aberrations in tumor cells and influence clinical, pathologic, and molecular characteristics of colorectal cancer (12). Thus, a molecular classification based on MSI and CIMP status is increasingly important (12, 18). However, no prior study has examined CDX2 expression in relation to CIMP or deciphered independent relationship of CDX2 loss with clinical, pathologic, and molecular variables in colorectal cancer.

Utilizing a database of 621 colorectal cancers, we therefore examined CDX2 expression in relation to patient survival and molecular features such as MSI, CIMP, and LINE-1 methylation. We have found that CDX2 loss is independently associated with CIMP and high-level LINE-1 methylation. In addition, we have found a possible modifying effect of family history on the relation between CDX2 loss and patient survival.

Study group. We used the databases of two independent, prospective cohort studies: the Nurses' Health Study (n = 121,701 women followed since 1976; ref. 19), and the Health Professional Follow-up Study (n = 51,529 men followed since 1986; ref. 19). Every 2 y, participants have been sent follow-up questionnaires to update information on potential risk factors and to identify newly diagnosed cancers in themselves and their first-degree relatives (father, mother, and sibling). We defined a family history as the presence of colorectal cancer in any first-degree relative. We calculated body mass index (BMI; kg/m2) using self-reported height and weight. Study physicians, while blinded to exposure data, reviewed all records related to colorectal cancer, and recorded tumor-node-metastasis, tumor stage, and tumor location. We collected paraffin-embedded tissue blocks from hospitals where patients underwent tumor resections (19). We excluded cases preoperatively treated with radiation and/or chemotherapy. Tissue sections from all colorectal cancer cases were reviewed and confirmed by one of the investigators (S.O.). Based on availability of adequate tissue specimens and results, a total of 621 colorectal cancers (diagnosed up to 2003) were included. For survival analysis, we excluded the patients with any cancer at baseline and the patients with no follow-up data, resulting in analysis of 598 patients. Among our cohort studies, there was no significant difference in demographic features between cases with tissue available and those without available tissue (19). This current analysis represents a new analysis of CDX2 on the existing colorectal cancer database that has been previously characterized for CIMP, MSI, LINE-1 methylation, and clinical outcome (1922), which is analogous to novel studies using the well-described cell lines or animal models. We have not examined CDX2 expression or the relationship between CDX2 and clinical outcome or other molecular events in any of our previous studies. Written informed consent was obtained from all study subjects. Tissue collection and analyses were approved by the Harvard School of Public Health and Brigham and Women's Hospital Institutional Review Boards.

Measurement of mortality. Patients were observed until death or June 30, 2006, whichever came first. Ascertainment of deaths included reporting by the family or postal authorities. In addition, the names of persistent nonresponders were searched in the National Death Index. More than 98% of deaths in the cohorts were identified by these methods. The cause of death was assigned by physicians blinded to information on life-style exposures and molecular features in colorectal cancer.

Histopathologic evaluations. H&E-stained tissue sections were examined by one of the investigators (S.O.) unaware of other data. The tumor grade was categorized as low (≥50% gland formation) versus high (<50% gland formation). The presence and extent of extracellular mucin were categorized as 0% (no mucin), 1% to 49%, or ≥50% of the tumor volume. The presence and extent of signet ring cells were categorized as absent (0%) or present (>0%).

Sequencing of KRAS, BRAF, and PIK3CA and MSI analysis. Genomic DNA was extracted from tumor and PCR and Pyrosequencing targeted for KRAS (codons 12 and 13; ref. 23), BRAF (codon 600; ref. 24), and PIK3CA (exons 9 and 20; ref. 25) were done as previously described. MSI analysis was done, using 10 microsatellite markers (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67, and D18S487; ref. 26). MSI-high was defined as the presence of instability in ≥30% of the markers. MSI-low was defined as instability in 10% to 29% of the markers, and “microsatellite stable” tumors were defined as tumors without an unstable marker (26).

Real-time PCR to measure CpG island methylation. Sodium bisulfite treatment on genomic DNA and subsequent real-time PCR (MethyLight) were validated and done as previously described (27). We quantified DNA methylation in eight CIMP-specific promoters [CACNA1G, CDKN2A (p16), CRABP1, IGF2, MLH1, NEUROG1, RUNX3, and SOCS1; refs. 15, 20, 28]. CIMP-high was defined as the presence of ≥6 of 8 methylated promoters, CIMP-low as the presence of 1 of 8 to 5 of 8 methylated promoters, and CIMP-0 as the absence (0 of 8) of methylated promoters, according to the previously established criteria (20).

Pyrosequencing to measure LINE-1 methylation. To accurately quantify relatively high methylation levels in LINE-1 repetitive elements, we used Pyrosequencing as previously described (22, 29).

Immunohistochemistry for CDX2, p53, p21, β-catenin, and cyclooxygenase-2. Tissue microarrays (TMA) were constructed as previously described (19). Methods of immunohistochemical procedure, interpretation, and evaluation were previously described for p53 and P21 [CDKN1A (30, 31) and β-catenin (17); and cyclooxygenase-2 (COX-2; refs. 19, 26)]. For CDX2 staining, antigen retrieval was done in 10 mmol/L sodium citrate buffer (pH 6.0) in a decloaking chamber (Biocare Medical) for 2 min. Endogenous peroxidases were blocked using 0.03% hydrogen peroxide containing sodium azide for 30 min. A primary antibody [mouse monoclonal to CDX2 (CDX2-88), 1:50 dilution; Biogenex Laboratories] was applied, and the slides were maintained overnight at 4°C. The remaining procedure was done using Dako EnVision+ System (DAKO Corporation). The stained slides were counterstained with hematoxylin and blueing reagent. Nuclear CDX2 expression was recorded as no, weak, moderate, or strong expression (Fig. 1). CDX2 loss (CDX2-negative expression) was defined as no expression, based on examination of two tissue cores per tumor in TMAs. Appropriate positive and negative controls were included in each run of immunohistochemistry. A whole tissue section of colonic carcinoma known to be CDX2 positive was used as a positive control, and a whole tissue section of colonic carcinoma known to be CDX2 negative was used as a negative control. Each immunohistochemical maker was interpreted by one of the investigators (CDX2 by Y.B.; β-catenin by K.N.; p53, p21, and COX-2 by S.O.) unaware of other data. A random selection of 118 cases was examined for CDX2 by a second observer (K.S.) unaware of other data, and the κ coefficient between the two observers was 0.84 (P < 0.0001), indicating substantial agreement. For each of the other immunohistochemical markers, a second observer (S.O. for β-catenin; K.S. for p21; K.N. for p53 and COX-2) examined a random sample of 108 to 402 tumors, unaware of other data. The κ coefficients between the two observers were 0.65 for β-catenin (P < 0.0001; n = 402), 0.62 for p21 (P < 0.0001; n = 179), 0.75 for p53 (P < 0.0001; n = 118), and 0.62 for COX-2 (P < 0.0001; n = 108), indicating substantial agreement.

Fig. 1.

CDX2 expression in normal colonic mucosa and colon cancer. A, nuclear CDX2 expression in normal colonic epithelial cells (arrow). B, nuclear CDX2 expression in colon cancer cells (white arrow). C, loss of CDX2 expression in colon cancer cells (arrowhead).

Fig. 1.

CDX2 expression in normal colonic mucosa and colon cancer. A, nuclear CDX2 expression in normal colonic epithelial cells (arrow). B, nuclear CDX2 expression in colon cancer cells (white arrow). C, loss of CDX2 expression in colon cancer cells (arrowhead).

Close modal

Statistical analysis. All statistical analyses used SAS program (version 9.1, SAS Institute). All P values were two sided, and statistical significance was set at a P value of ≤0.05. Nonetheless, P values were conservatively interpreted, considering multiple hypothesis testing. For categorical data, the χ2 test was done and odds ratio (OR) with 95% confidence interval (CI) was computed. The κ coefficient was calculated to assess an agreement between the two interpreters in immunohistochemical analyses. We also examined the possibility of a nonlinear relation between CDX2 loss and LINE1 methylation, nonparametrically with restricted cubic spines (32). This flexible method allowed us to examine the relation to CDX2 loss without any categorization of LINE-1 methylation level, or without the assumption of a linear relationship of LINE-1 methylation level with CDX2 loss.

To assess independent relations of CDX2 loss with other variables, a multivariate logistic regression analysis was done. OR was adjusted for sex, age (continuous), BMI (<30 versus ≥30 kg/m2), family history of colorectal cancer (present versus absent), tumor location (rectum versus colon), tumor stage (I-III versus IV), tumor grade (low versus high), mucinous component (0 versus ≥1%), signet ring cell component (0 versus >0%), CIMP status (high versus low versus 0), MSI status (high versus low/microsatellite stable), LINE-1 methylation (continuous), BRAF, KRAS, PIK3CA, p53, p21, β-catenin, and COX-2. For CIMP (2.9% missing) and tumor grade (3.2% missing), we assigned indicator (“missing”) variables to those cases with missing data, and included those cases in multivariate analysis. For cases missing LINE-1 data (4.0% missing), we assigned the median LINE-1 methylation level. For cases with missing information in other variables [BMI (5.8% missing), tumor location (1.6%), stage (6.1%), MSI (0.8%), BRAF (2.7%), KRAS (0.5%), PIK3CA (11%), p53 (1.1%), p21 (2.6%), and COX-2 (0.5%)], we included those cases in a majority category of the missing variable, to minimize the number of indicator variables in multivariate analysis. We confirmed that excluding cases with missing information in any of the covariates did not substantially alter results (data not shown). An interaction was assessed by including the cross product of two variables of interest (without data-missing cases) in a multivariate logistic regression model, and the Wald test was done.

For survival analysis, Kaplan-Meier method was used to assess survival time distribution according to CDX2 status, and log-rank test was used to test significance of a deviation from the null hypothesis. For analyses of colorectal cancer–specific mortality, death as a result of colorectal cancer was the primary end point and deaths as a result of other causes were censored. To assess independent effect of CDX2 loss on mortality, we constructed a multivariate, stage-matched (stratified) conditional Cox proportional hazard model to compute a hazard ratio (HR) according to CDX2 status, adjusted for sex, age, BMI, family history of colorectal cancer, tumor location, tumor grade, mucinous component, signet ring cell component, CIMP, MSI, BRAF, KRAS, PIK3CA, LINE-1 methylation, p53, p21, β-catenin, and COX-2. Tumor stage (I, IIA, IIB, IIIA, IIIB, IIIC, IV, unknown) was used as a stratifying (matching) variable in Cox models. The proportionality of hazard assumption was satisfied by evaluating time-dependent variables, which were the cross-product of the CDX2 variable and survival time (P = 0.76 for colon cancer–specific mortality; P = 0.91 for overall mortality). For cases with missing data, data in any of the covariates were dealt as in the multivariate logistic regression analysis described above. An interaction was assessed by including the cross-product of the CDX2 variable and another variable of interest (without data-missing cases) in a multivariate Cox model, and the Wald test was done.

CDX2 loss in relation to clinical, pathologic, and molecular features in colorectal cancer. Among 621 colorectal cancers, we observed CDX2 expression in 438 tumors (71%) and CDX2 loss in 183 tumors (29%) by immunohistochemistry (Fig. 1). Table 1 shows the frequency of CDX2 loss in relation to various clinical, pathologic, and molecular features. CDX2 loss was associated with female gender (OR, 3.33; 95% CI, 2.22-4.98; P < 0.0001), proximal location (compared with rectum; OR, 4.08; 95% CI, 2.32-7.17; P < 0.0001), stage IV (compared with stage I; OR, 3.39; 95% CI, 1.87-6.16; P < 0.0001), high tumor grade (OR, 5.41; 95% CI, 3.03-9.67; P < 0.0001), mucinous component (≥50% versus 0%; OR, 2.29; 95% CI, 1.37-3.85; P = 0.0014), signet ring cell component (OR, 2.54; 95% CI, 1.34-4.80; P = 0.0032), CIMP-high (compared with CIMP-0; OR, 11.1; 95% CI, 6.47-19.1; P < 0.0001), MSI-high (OR, 4.39; 95% CI, 2.79-6.91; P < 0.0001), and BRAF mutation (OR, 5.78; 95% CI, 3.57-9.33; P < 0.0001), and inversely associated with KRAS mutation (OR, 0.57; 95% CI, 0.39-0.83; P = 0.0034), LINE-1 hypomethylation, p53 expression (OR, 0.44; 95% CI, 0.30-0.64; P < 0.0001), loss of p21 (OR, 0.38; 95% CI, 0.25-0.58; P < 0.0001), and β-catenin activation (OR, 0.46; 95% CI, 0.30-0.69; P = 0.0002).

Table 1.

Frequency of CDX2 loss in colorectal cancer according to clinical, pathologic and molecular features

Clinical or molecular featurenCDX2 lossUnivariate OR (95% CI)P
All cases 621 183 (29%)   
Sex     
    Men 242 38 (16%) Reference 
    Women 379 145 (38%) 3.33 (2.22-4.98) <0.0001 
Age (y)     
    ≤59 142 47 (33%)  
    60-69 264 81 (31%) 0.89 (0.58-1.38)  
    ≥70 215 55 (26%) 0.69 (0.44-1.11)  
BMI (kg/m2    
    <25 268 78 (29%)  
    25-30 217 64 (29%) 1.02 (0.69-1.51)  
    ≥30 100 29 (29%) 0.99 (0.60-1.65)  
Family history of colorectal cancer     
    (−) 477 147 (31%)  
    (+) 144 36 (25%) 0.75 (0.49-1.14)  
Tumor location     
    Rectum 122 17 (14%) Reference 
    Distal colon (splenic flexure to sigmoid) 200 49 (25%) 2.00 (1.09-3.67) 0.023 
    Proximal (cecum to transverse colon) 289 115 (40%) 4.08 (2.32-7.17) <0.0001 
Stage     
    I 130 28 (22%) Reference 
    II 193 56 (29%) 1.49 (0.88-2.51)  
    III 175 50 (29%) 1.46 (0.86-2.48)  
    IV 85 41 (48%) 3.39 (1.87-6.16) <0.0001 
Tumor grade     
    Low 545 136 (25%) Reference 
    High 56 36 (64%) 5.41 (3.03-9.67) <0.0001 
Mucinous component     
    0% 307 72 (23%) Reference 
    1-49% 143 46 (32%) 1.55 (1.00-2.40)  
    ≥50% 80 33 (41%) 2.29 (1.37-3.85) 0.0014 
Signet ring cell component     
    0% 447 114 (26%) Reference 
    >0% 43 20 (47%) 2.54 (1.34-4.80) 0.0032 
CIMP status (no. of methylated CIMP markers)     
    CIMP-0 (0) 268 42 (16%) Reference 
    CIMP-low (1-5) 240 67 (28%) 2.08 (1.35-3.22) 0.0008 
    CIMP-high (6-8) 95 64 (67%) 11.1 (6.47-19.1) <0.0001 
MSI status     
    MSS 459 105 (23%) Reference 
    MSI-low 58 20 (34%) 1.77 (0.99-3.18)  
    MSI-high 99 56 (57%) 4.39 (2.79-6.91) <0.0001 
BRAF mutation     
    (−) 516 120 (23%) Reference 
    (+) 88 56 (64%) 5.78 (3.57-9.33) <0.0001 
KRAS mutation     
    (−) 394 132 (34%) Reference 
    (+) 224 50 (22%) 0.57 (0.39-0.83) 0.0034 
PIK3CA mutation     
    (-) 475 132 (28%)  
    (+) 78 19 (24%) 0.84 (0.48-1.46)  
LINE-1 methylation     
    ≥70% 106 45 (42%) Reference 
    50-69% 412 104 (25%) 0.46 (0.29-0.71) 0.0005 
    <50% 78 17 (22%) 0.38 (0.20-0.73) 0.0034 
p53 expression     
    (−) 351 128 (36%) Reference 
    (+) 263 53 (20%) 0.44 (0.30-0.64) <0.0001 
p21     
    Expressed 121 56 (46%) Reference 
    Lost 484 120 (25%) 0.38 (0.25-0.58) <0.0001 
β-Catenin*     
    Inactive (score 0-2) 348 113 (32%) Reference 
    Active (score 3-5) 211 38 (18%) 0.46 (0.30-0.69) 0.0002 
COX-2 expression     
    (−) 99 38 (38%) Reference 
    (+) 519 144 (28%) 0.62 (0.39-0.97) 0.033 
Clinical or molecular featurenCDX2 lossUnivariate OR (95% CI)P
All cases 621 183 (29%)   
Sex     
    Men 242 38 (16%) Reference 
    Women 379 145 (38%) 3.33 (2.22-4.98) <0.0001 
Age (y)     
    ≤59 142 47 (33%)  
    60-69 264 81 (31%) 0.89 (0.58-1.38)  
    ≥70 215 55 (26%) 0.69 (0.44-1.11)  
BMI (kg/m2    
    <25 268 78 (29%)  
    25-30 217 64 (29%) 1.02 (0.69-1.51)  
    ≥30 100 29 (29%) 0.99 (0.60-1.65)  
Family history of colorectal cancer     
    (−) 477 147 (31%)  
    (+) 144 36 (25%) 0.75 (0.49-1.14)  
Tumor location     
    Rectum 122 17 (14%) Reference 
    Distal colon (splenic flexure to sigmoid) 200 49 (25%) 2.00 (1.09-3.67) 0.023 
    Proximal (cecum to transverse colon) 289 115 (40%) 4.08 (2.32-7.17) <0.0001 
Stage     
    I 130 28 (22%) Reference 
    II 193 56 (29%) 1.49 (0.88-2.51)  
    III 175 50 (29%) 1.46 (0.86-2.48)  
    IV 85 41 (48%) 3.39 (1.87-6.16) <0.0001 
Tumor grade     
    Low 545 136 (25%) Reference 
    High 56 36 (64%) 5.41 (3.03-9.67) <0.0001 
Mucinous component     
    0% 307 72 (23%) Reference 
    1-49% 143 46 (32%) 1.55 (1.00-2.40)  
    ≥50% 80 33 (41%) 2.29 (1.37-3.85) 0.0014 
Signet ring cell component     
    0% 447 114 (26%) Reference 
    >0% 43 20 (47%) 2.54 (1.34-4.80) 0.0032 
CIMP status (no. of methylated CIMP markers)     
    CIMP-0 (0) 268 42 (16%) Reference 
    CIMP-low (1-5) 240 67 (28%) 2.08 (1.35-3.22) 0.0008 
    CIMP-high (6-8) 95 64 (67%) 11.1 (6.47-19.1) <0.0001 
MSI status     
    MSS 459 105 (23%) Reference 
    MSI-low 58 20 (34%) 1.77 (0.99-3.18)  
    MSI-high 99 56 (57%) 4.39 (2.79-6.91) <0.0001 
BRAF mutation     
    (−) 516 120 (23%) Reference 
    (+) 88 56 (64%) 5.78 (3.57-9.33) <0.0001 
KRAS mutation     
    (−) 394 132 (34%) Reference 
    (+) 224 50 (22%) 0.57 (0.39-0.83) 0.0034 
PIK3CA mutation     
    (-) 475 132 (28%)  
    (+) 78 19 (24%) 0.84 (0.48-1.46)  
LINE-1 methylation     
    ≥70% 106 45 (42%) Reference 
    50-69% 412 104 (25%) 0.46 (0.29-0.71) 0.0005 
    <50% 78 17 (22%) 0.38 (0.20-0.73) 0.0034 
p53 expression     
    (−) 351 128 (36%) Reference 
    (+) 263 53 (20%) 0.44 (0.30-0.64) <0.0001 
p21     
    Expressed 121 56 (46%) Reference 
    Lost 484 120 (25%) 0.38 (0.25-0.58) <0.0001 
β-Catenin*     
    Inactive (score 0-2) 348 113 (32%) Reference 
    Active (score 3-5) 211 38 (18%) 0.46 (0.30-0.69) 0.0002 
COX-2 expression     
    (−) 99 38 (38%) Reference 
    (+) 519 144 (28%) 0.62 (0.39-0.97) 0.033 

Abbreviations: MSS, microsatellite stable; LINE-1, long interspersed nucleotide element-1.

*

β-Catenin activation score was calculated as previously described (17).

Relationship between CDX2 loss and LINE-1 hypomethylation. We examined the possibility of a nonlinear relationship between LINE-1 hypomethylation and CDX2 loss by using a nonparametric method with restricted cubic splines (32). This flexible method allowed us to evaluate the relationship without predetermined LINE-1 categorization or assumption of a linear relation. The OR for CDX2 loss decreased precipitously as LINE-1 methylation level decreased, indicating an inverse association between LINE-1 hypomethylation and CDX2 loss (Fig. 2A).

Fig. 2.

A, smoothing spline plot for the CDX2 loss/LINE-1 relation, showing OR for CDX2 loss (Y axis) according to LINE-1 methylation level (X axis) with high LINE-1 methylation level as a reference. Hatched lines, 95% CI. B, Kaplan-Meier curves for colon cancer-specific survival (top) and overall survival (bottom) according to CDX2 status in colorectal cancer.

Fig. 2.

A, smoothing spline plot for the CDX2 loss/LINE-1 relation, showing OR for CDX2 loss (Y axis) according to LINE-1 methylation level (X axis) with high LINE-1 methylation level as a reference. Hatched lines, 95% CI. B, Kaplan-Meier curves for colon cancer-specific survival (top) and overall survival (bottom) according to CDX2 status in colorectal cancer.

Close modal

Multivariate analysis to assess independent relations with CDX2 loss. To examine which variables were independently associated with CDX2 loss, we did multivariate logistic regression analysis (Table 2). CDX2 loss was significantly associated with female gender (OR, 3.32; 95% CI, 2.00-5.52; P < 0.0001), CIMP-high (versus CIMP-0; OR, 4.42; 95% CI, 1.98-9.86; P = 0.0003), and inversely with LINE-1 hypomethylation (for a 30% decline; OR, 0.33; 95% CI, 0.16-0.69; P = 0.0031). Considering multiple hypothesis testing, any of the associations with P values between 0.05 and 0.005 (including those with CIMP-low, high tumor grade, p53, stage IV, and β-catenin) might represent chance events. In multivariate analysis, CDX2 loss was not significantly associated with age, BMI, family history of colorectal cancer, tumor location, mucin, signet ring cells, MSI, KRAS, BRAF, PIK3CA, p21, or COX-2.

Table 2.

Multivariate analysis of the relations with CDX2 loss in colorectal cancer

Variable independently associated with CDX2 lossMultivariate OR (95% CI)P
Female gender 3.32 (2.00-5.52) <0.0001 
CIMP-high (vs CIMP-0) 4.42 (1.98-9.86) 0.0003 
LINE-1 hypomethylation (30% decrease as a unit) 0.33 (0.16-0.69) 0.0031 
CIMP-low (vs CIMP-0) 1.97 (1.21-3.18) 0.0059 
High tumor grade (vs low grade) 2.69 (1.29-5.61) 0.0085 
P53 expression 0.55 (0.34-0.87) 0.011 
Stage IV (vs stage I-III) 2.03 (1.12-3.66) 0.019 
β-Catenin activation 0.60 (0.37-0.97) 0.037 
Variable independently associated with CDX2 lossMultivariate OR (95% CI)P
Female gender 3.32 (2.00-5.52) <0.0001 
CIMP-high (vs CIMP-0) 4.42 (1.98-9.86) 0.0003 
LINE-1 hypomethylation (30% decrease as a unit) 0.33 (0.16-0.69) 0.0031 
CIMP-low (vs CIMP-0) 1.97 (1.21-3.18) 0.0059 
High tumor grade (vs low grade) 2.69 (1.29-5.61) 0.0085 
P53 expression 0.55 (0.34-0.87) 0.011 
Stage IV (vs stage I-III) 2.03 (1.12-3.66) 0.019 
β-Catenin activation 0.60 (0.37-0.97) 0.037 

NOTE. The multivariate logistic regression model included age, BMI, family history of colorectal cancer, tumor location, mucinous component, signet ring cells, MSI, KRAS, BRAF, PIK3CA, p21, COX-2, and the variables listed in the table.

We examined whether the relations of CDX2 loss with any one of the seven variables listed in Table 2 (sex, CIMP, LINE-1, tumor grade, p53, stage, and β-catenin) is influenced by any of the other six. There was no evidence for a significant interaction between any pair among these seven variables (all P values for interaction > 0.11).

Loss of CDX2 expression and patient survival. During follow-up of 598 patients who were eligible for survival analysis, there were 255 deaths, including 156 deaths attributed to colorectal cancer. We assessed the influence of CDX2 loss on patient mortality. In Kaplan-Meier analysis, the 5-year colorectal cancer–specific survival was 68% in patients with CDX2-lost tumors and 80% in patients with CDX2-expressing tumors (log-rank P = 0.0010), and the 5-year overall survival was 63% in patients with CDX2-lost tumors and 76% in patients with CDX2-expressing tumors (log-rank P = 0.0051; Fig. 2B). In univariate Cox regression analysis, compared with patients with CDX2-expressing tumors, those with CDX2-lost tumors experienced significantly higher colorectal cancer–specific (HR, 1.71; 95% CI, 1.24-2.36; P = 0.0012) and overall (HR, 1.44; 95% CI, 1.12-1.87; P = 0.0053) mortality (Table 3). In the multivariate Cox model adjusting for potential predictors of patient outcome, CDX2 loss was not significantly associated with cancer-specific mortality (multivariate HR, 1.02; 95% CI, 0.65-1.61) or overall mortality (multivariate HR, 1.16; 95% CI, 0.82-1.64; Table 3). Elimination of the prognostic influence of CDX2 loss in multivariate analysis was essentially due to adjusting for tumor stage: when we simply adjusted for tumor stage, adjusted HRs for colorectal cancer–specific mortality and overall mortality were 1.22 (95% CI, 0.87-1.72) and 1.16 (95% CI, 0.88-1.52), respectively. No other major confounder was identified.

Table 3.

CDX2 status in colorectal cancer and patient mortality

nColorectal cancer–specific mortality
Overall mortality
Deaths/person-yearsUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)Deaths/person-yearsUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)
CDX2 expression 426 96/3469 1 (Reference) 1 (Reference) 1 (Reference) 166/3469 1 (Reference) 1 (Reference) 1 (Reference) 
CDX2 loss 172 60/1293 1.71 (1.24-2.36) 1.22 (0.87-1.72) 1.02 (0.65-1.61) 89/1293 1.44 (1.12-1.87) 1.16 (0.88-1.52) 1.16 (0.82-1.64) 
   P = 0.0012 P = 0.25 P = 0.92  P = 0.0053 P = 0.29 P = 0.39 
nColorectal cancer–specific mortality
Overall mortality
Deaths/person-yearsUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)Deaths/person-yearsUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)
CDX2 expression 426 96/3469 1 (Reference) 1 (Reference) 1 (Reference) 166/3469 1 (Reference) 1 (Reference) 1 (Reference) 
CDX2 loss 172 60/1293 1.71 (1.24-2.36) 1.22 (0.87-1.72) 1.02 (0.65-1.61) 89/1293 1.44 (1.12-1.87) 1.16 (0.88-1.52) 1.16 (0.82-1.64) 
   P = 0.0012 P = 0.25 P = 0.92  P = 0.0053 P = 0.29 P = 0.39 

NOTE. The multivariate, stage-matched (stratified) Cox model included sex, age at diagnosis, year of diagnosis, BMI, family history of colorectal cancer, tumor location, grade, CIMP, MSI, BRAF, KRAS, PIK3CA, LINE-1 methylation, p53, p21, β-catenin, and COX-2.f.

Modifying effect of family history of colorectal cancer on the relation between CDX2 loss and mortality. We examined whether the prognostic influence of CDX2 loss was modified by any of the variables we examined. We found a significant modifying effect of family history of colorectal cancer on the relation between CDX2 loss and mortality [Pinteraction = 0.011 (for cancer specific mortality) and Pinteraction = 0.0094 (for overall mortality)], although this could be a chance event considering multiple hypothesis testing. Among patients with family history of colorectal cancer, CDX2 loss was associated with a significant increase in colorectal cancer–specific mortality in both stage-matched Cox regression analysis (HR, 3.06; 95% CI, 1.47-6.34) and multivariate analysis (HR, 2.53; 95% CI, 1.13-5.65; Table 4). In contrast, among patients without family history of colorectal cancer, CDX2 did not seem to be related with prognosis. Similar results were obtained when overall mortality was used as the end point (Table 4).

Table 4.

CDX2 status in colorectal cancer and patient mortality in strata of family history of colorectal cancer

Colorectal cancer–specific mortality
Overall mortality
No. of deaths/casesUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)No. of deaths/casesUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)
Family history (−)         
    CDX2 expression 75/321 1 (Reference) 1 (Reference) 1 (Reference) 131/321 1 (Reference) 1 (Reference) 1 (Reference) 
    CDX2 loss 48/137 1.63 (1.13-2.35) 1.03 (0.70-1.51) 0.81 (0.50-1.32) 71/137 1.33 (1.00-1.78) 0.98 (0.72-1.33) 0.97 (0.66-1.41) 
Family history (+)         
    CDX2 expression 22/106 1 (Reference) 1 (Reference) 1 (Reference) 36/106 1 (Reference) 1 (Reference) 1 (Reference) 
    CDX2 loss 12/35 1.99 (0.98-4.02) 3.06 (1.47-6.34) 2.53 (1.13-5.65) 18/35 1.84 (1.04-3.24) 2.34 (1.31-4.18) 2.40 (1.28-4.51) 
Pinteraction (CDX2 loss and family history)  0.63 0.010 0.011  0.32 0.0096 0.0094 
Colorectal cancer–specific mortality
Overall mortality
No. of deaths/casesUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)No. of deaths/casesUnivariate HR (95% CI)Stage-matched HR (95% CI)Multivariate HR (95% CI)
Family history (−)         
    CDX2 expression 75/321 1 (Reference) 1 (Reference) 1 (Reference) 131/321 1 (Reference) 1 (Reference) 1 (Reference) 
    CDX2 loss 48/137 1.63 (1.13-2.35) 1.03 (0.70-1.51) 0.81 (0.50-1.32) 71/137 1.33 (1.00-1.78) 0.98 (0.72-1.33) 0.97 (0.66-1.41) 
Family history (+)         
    CDX2 expression 22/106 1 (Reference) 1 (Reference) 1 (Reference) 36/106 1 (Reference) 1 (Reference) 1 (Reference) 
    CDX2 loss 12/35 1.99 (0.98-4.02) 3.06 (1.47-6.34) 2.53 (1.13-5.65) 18/35 1.84 (1.04-3.24) 2.34 (1.31-4.18) 2.40 (1.28-4.51) 
Pinteraction (CDX2 loss and family history)  0.63 0.010 0.011  0.32 0.0096 0.0094 

NOTE. The multivariate, stage-matched (stratified) Cox model included CDX2 variable stratified by family history, sex, age, year of diagnosis, BMI, tumor location, grade, CIMP, MSI, BRAF, KRAS, PIK3CA, LINE-1 methylation, p53, p21, β-catenin, and COX-2.

For any of the other variables including sex, age, BMI, tumor location, stage, grade, mucinous component, signet ring cells, CIMP, MSI, BRAF, KRAS, PIK3CA, LINE-1 methylation, p53, p21, β-catenin, and COX-2, we did not observe a significant interaction with CDX2 loss in survival analysis (all Pinteraction >0.05).

We conducted this study to examine loss of CDX2 expression in colorectal cancer, in relation to clinical, pathologic, and molecular features including the CIMP and patient mortality. CDX2 is used as an immunohistochemical marker to distinguish adenocarcinomas of a colorectal origin from those arising in other organs (25). We have shown that CDX2 loss is independently associated with female gender, CIMP-high, and high tumor grade, and inversely with LINE-1 hypomethylation and p53 expression. Moreover, we show that CDX2 loss is associated with advanced tumor stage (stage IV) and inferior survival. The relationship of CDX2 loss with high mortality is especially remarkable among patients with family history of colorectal cancer. Thus, this study provides useful information on CDX2 expression patterns in colorectal cancer and may refine the role of CDX2 immunohistochemistry in clinical and pathology practice.

Studying molecular alterations is important in cancer research (3337). Accumulating evidence suggests that CDX2 is a tumor suppressor gene (79). Heterozygous CDX2 knockout mice develop multiple hamartomatous polyps in the proximal colon (7). Furthermore, although APC± mice develop adenomatous polyposis of the small intestine, APC± CDX2± mice develop polyposis of the colon (8). The procarcinogen azoxymethane induced invasive colon adenocarcinoma in CDX2± mice but not wild-type littermates (9). Considering the potential role of CDX2 as a tumor suppressor, molecular correlates with CDX2 loss may be important to better understand the interrelationship of multiple genetic and epigenetic alterations during carcinogenic process.

CDX2 expression, a marker of intestinal differentiation, may be expressed in any gastrointestinal cancer but also in gynecologic cancer (38); based on abundant evidence (25), it is commonly used as a relatively specific marker for adenocarcinoma of the lower digestive tract. Previous results on the frequency and significance of CDX2 loss in colorectal cancer have not been uniform. Some studies investigating the utility of CDX2 as a marker for intestinal adenocarcinomas report that CDX2 is detected in 98% to 100% of cases, although staining pattern and intensity vary and focal expression/loss is quite common (35); other studies report loss of CDX2 expression in 14% to 37% of cases (2, 11), which agrees with our data. The discrepancy may be due to a difference in antibody sensitivity and specificity or methods of assessing CDX2 expression (e.g., TMA based versus whole tissue based). We must of course consider a limitation of TMA-based assessment; tumors with partial or heterogeneous positivity in whole sections might have been scored as “negative” in TMA cores. Despite this limitation, we were able to detect highly significant and independent relations of CDX2 loss with female gender, CIMP, and LINE-1 methylation, due to the large sample size and our meticulous laboratory and statistical evaluations.

Loss of CDX2 expression in colorectal cancer has previously been associated with high tumor grade, advanced tumor stage, and MSI-high (10, 11). Low-level CDX2 mRNA has been related with MSI-high in colorectal cancer (38). However, CIMP status has not been examined in these previous studies (10, 11, 39). Our multivariate analysis reveals that CDX2 loss is independently associated with CIMP-high, high tumor grade, and advanced tumor stage, whereas MSI-high is associated with CDX2 loss in only univariate analysis, but not in multivariate analysis. As there is a strong association between MSI-high and CIMP-high (12, 15), the association of MSI-high with CDX2 loss in previous studies (10, 11) and our univariate analysis likely reflects enrichment of CIMP-high in MSI-high colorectal cancers. In addition, we show independent relations of CDX2 loss with female gender and high-level LINE-1 methylation. In using CDX2 as a marker for colorectal cancer, clinicians should recognize that female gender, high tumor grade, advanced stage, high-level LINE-1 methylation, and CIMP-high increase the likelihood of CDX2 loss. In light of the lack of evidence for significant interactions between these variables, the presence of two or more of these features may further increase the likelihood of CDX2 loss. Thus, our findings have useful clinical implications and delineate caveats for interpreting absence of CDX2 expression in a metastatic tumor of an unknown origin, especially in a small biopsy specimen where a whole tumor profile is not available (somewhat analogous to the current TMA-based study); further studies are necessary to confirm these features of tumors with CDX2 loss.

The mechanisms for loss of CDX2 expression during colorectal carcinogenesis are not well characterized. CDX2 mutations occur infrequently in colorectal cancers with defective DNA mismatch repair (i.e., MSI-high; refs. 40, 41). In a population-based case-control study, CDX2 polymorphisms do not play a role in reduced CDX2 expression (38). Although not likely a major cause, loss of heterozygosity at the CDX2 locus (13q12-13) may account for loss of CDX2 expression in a small subset of colorectal cancer (42). A study using colon cancer cell lines has shown evidence for a dominant transcriptional repressor of CDX2 and indicated the possibility that an epigenetic alteration, such as promoter CpG island methylation, might be responsible for CDX2 silencing (43). Our finding of the relationship between CDX2 loss and CIMP-high supports the possibility of epigenetic CDX2 silencing in a subset of colorectal tumors.

An important question is whether loss of CDX2 expression can predict clinical outcome. Considering the role of CDX2 in promoting cellular differentiation (1) and inhibiting proliferation (8), CDX2 loss could conceivably contribute to aggressive tumor behavior. In malignancies other than colorectal cancer, CDX2 loss has been associated with poor prognosis (44, 45). In colorectal cancer, CDX2 loss has been associated with poor survival in only univariate analysis, but not in multivariate analysis (11), which is in agreement with our data. These results likely reflect the association of CDX2 loss with advanced tumor stage. Nonetheless, loss of CDX2 expression in colorectal cancer increases the likelihood of metastatic disease and its loss may therefore serve as a marker to warrant more intensive surveillance in colorectal cancer patients.

The effect of tumoral CDX2 loss on prognosis seems to differ according to family history of colorectal cancer. Specifically, CDX2 loss is independently associated with poor prognosis among patients with family history of colorectal cancer, but not among patients lacking such a family history. A family history of colorectal cancer approximately doubles the risk of developing the disease (46). Excluding studies on familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer, several large-scale studies address the influence of family history on colorectal cancer recurrence and survival (4751). Some studies (4951) have shown that family history of colorectal cancer is associated with a significant reduction in cancer recurrence or death, whereas other studies (47, 48) have reported no such significant effect of family history. To date, no large-scale study (other than our current study) has examined a potential modifying effect of family history of colorectal cancer on any molecular predictor of patient outcome. These intriguing findings on interactions between CDX2 loss, family history of colorectal cancer, and tumor behavior need to be confirmed in independent cohorts in the future.

In conclusion, our large-scale study has shown that loss of CDX2 expression in colorectal cancer is independently associated with female gender, CIMP-high, high-level LINE-1 methylation, high tumor grade, and advanced stage (stage IV). Our results should alert pathologists and clinicians that the usefulness of CDX2 testing to identify metastatic colorectal tumors may be affected by clinical and molecular variables. In addition, CDX2 loss in colorectal cancer is also independently associated with poor prognosis among patients with a family history of colorectal cancer. Loss of CDX2 expression may also serve as a marker to predict outcome for patients with a family history of colorectal cancer. Further studies are necessary to confirm our findings.

No potential conflicts of interest were disclosed.

Grant support: U.S. NIH grants P01 CA87969 (S. Hankinson), P01 CA55075 (W. Willett), P50 CA127003 (C.S. Fuchs), K07 CA122826 (S. Ogino), and K07 CA97992 (J.A. Meyerhardt); and in part by grants from the Bennett Family Fund and from the Entertainment Industry Foundation National Colorectal Cancer Research Alliance. K. Nosho was supported by a fellowship grant from the Japan Society for Promotion of Science. The content is solely the responsibility of the authors and does not necessarily represent the official views of National Cancer Institute or NIH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

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: Y. Baba, K. Nosho, and K. Shima contributed equally to this work.

We thank the Nurses' Health Study and Health Professionals Follow-up Study cohort participants who have generously agreed to provide us with biological specimens and information through responses to questionnaires, and Frank Speizer, Walter Willett, Susan Hankinson, Graham Colditz, Meir Stampfer, and many other staff members who implemented and have maintained the cohort studies.

1
Beck F. The role of Cdx genes in the mammalian gut.
Gut
2004
;
53
:
1394
–6.
2
Kaimaktchiev V, Terracciano L, Tornillo L, et al. The homeobox intestinal differentiation factor CDX2 is selectively expressed in gastrointestinal adenocarcinomas.
Mod Pathol
2004
;
17
:
1392
–9.
3
Moskaluk CA, Zhang H, Powell SM, Cerilli LA, Hampton GM, Frierson HF, Jr. Cdx2 protein expression in normal and malignant human tissues: an immunohistochemical survey using tissue microarrays.
Mod Pathol
2003
;
16
:
913
–9.
4
Barbareschi M, Murer B, Colby TV, et al. CDX-2 homeobox gene expression is a reliable marker of colorectal adenocarcinoma metastases to the lungs.
Am J Surg Pathol
2003
;
27
:
141
–9.
5
Werling RW, Yaziji H, Bacchi CE, Gown AM. CDX2, a highly sensitive and specific marker of adenocarcinomas of intestinal origin: an immunohistochemical survey of 476 primary and metastatic carcinomas.
Am J Surg Pathol
2003
;
27
:
303
–10.
6
Abate-Shen C. Deregulated homeobox gene expression in cancer: cause or consequence?
Nat Rev Cancer
2002
;
2
:
777
–85.
7
Chawengsaksophak K, James R, Hammond VE, Kontgen F, Beck F. Homeosis and intestinal tumours in Cdx2 mutant mice.
Nature
1997
;
386
:
84
–7.
8
Aoki K, Tamai Y, Horiike S, Oshima M, Taketo MM. Colonic polyposis caused by mTOR-mediated chromosomal instability in Apc+/Δ716 Cdx2+/− compound mutant mice.
Nat Genet
2003
;
35
:
323
–30.
9
Bonhomme C, Duluc I, Martin E, et al. The Cdx2 homeobox gene has a tumour suppressor function in the distal colon in addition to a homeotic role during gut development.
Gut
2003
;
52
:
1465
–71.
10
Hinoi T, Tani M, Lucas PC, et al. Loss of CDX2 expression and microsatellite instability are prominent features of large cell minimally differentiated carcinomas of the colon.
Am J Pathol
2001
;
159
:
2239
–48.
11
Lugli A, Tzankov A, Zlobec I, Terracciano LM. Differential diagnostic and functional role of the multi-marker phenotype CDX2/CK20/CK7 in colorectal cancer stratified by mismatch repair status.
Mod Pathol
2008
;
21
:
1403
–12.
12
Ogino S, Goel A. Molecular classification and correlates in colorectal cancer.
J Mol Diagn
2008
;
10
:
13
–27.
13
Teodoridis JM, Hardie C, Brown R. CpG island methylator phenotype (CIMP) in cancer: causes and implications.
Cancer Lett
2008
;
268
:
177
–86.
14
Toyota M, Ahuja N, Ohe-Toyota M, Herman JG, Baylin SB, Issa JP. CpG island methylator phenotype in colorectal cancer.
Proc Natl Acad Sci U S A
1999
;
96
:
8681
–6.
15
Weisenberger DJ, Siegmund KD, Campan M, et al. CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer.
Nat Genet
2006
;
38
:
787
–93.
16
Samowitz WS, Albertsen H, Herrick J, et al. Evaluation of a large, population-based sample supports a CpG island methylator phenotype in colon cancer.
Gastroenterology
2005
;
129
:
837
–45.
17
Kawasaki T, Nosho K, Ohnishi M, et al. Correlation of β-catenin localization with cyclooxygenase-2 expression and CpG island methylator phenotype (CIMP) in colorectal cancer.
Neoplasia
2007
;
9
:
569
–77.
18
Jass JR. Classification of colorectal cancer based on correlation of clinical, morphological and molecular features.
Histopathology
2007
;
50
:
113
–30.
19
Chan AT, Ogino S, Fuchs CS. Aspirin and the risk of colorectal cancer in relation to the expression of COX-2.
N Engl J Med
2007
;
356
:
2131
–42.
20
Ogino S, Kawasaki T, Kirkner GJ, Kraft P, Loda M, Fuchs CS. Evaluation of markers for CpG island methylator phenotype (CIMP) in colorectal cancer by a large population-based sample.
J Mol Diagn
2007
;
9
:
305
–14.
21
Ogino S, Nosho K, Kirkner GJ, et al. CpG island methylator phenotype, microsatellite instability, BRAF mutation and clinical outcome in colon cancer.
Gut
2009
;
58
:
90
–6.
22
Ogino S, Nosho K, Kirkner GJ, et al. A cohort study of tumoral LINE-1 hypomethylation and prognosis in colon cancer.
J Natl Cancer Inst
2008
;
100
:
1734
–8.
23
Ogino S, Kawasaki T, Brahmandam M, et al. Sensitive sequencing method for KRAS mutation detection by Pyrosequencing.
J Mol Diagn
2005
;
7
:
413
–21.
24
Ogino S, Kawasaki T, Kirkner GJ, Loda M, Fuchs CS. CpG island methylator phenotype-low (CIMP-low) in colorectal cancer: possible associations with male sex and KRAS mutations.
J Mol Diagn
2006
;
8
:
582
–8.
25
Nosho K, Kawasaki T, Ohnishi M, et al. PIK3CA mutation in colorectal cancer: relationship with genetic and epigenetic alterations.
Neoplasia
2008
;
10
:
534
–41.
26
Ogino S, Brahmandam M, Cantor M, et al. Distinct molecular features of colorectal carcinoma with signet ring cell component and colorectal carcinoma with mucinous component.
Mod Pathol
2006
;
19
:
59
–68.
27
Ogino S, Kawasaki T, Brahmandam M, et al. Precision and performance characteristics of bisulfite conversion and real-time PCR (MethyLight) for quantitative DNA methylation analysis.
J Mol Diagn
2006
;
8
:
209
–17.
28
Ogino S, Cantor M, Kawasaki T, et al. CpG island methylator phenotype (CIMP) of colorectal cancer is best characterised by quantitative DNA methylation analysis and prospective cohort studies.
Gut
2006
;
55
:
1000
–6.
29
Ogino S, Kawasaki T, Nosho K, et al. LINE-1 hypomethylation is inversely associated with microsatellite instability and CpG island methylator phenotype in colorectal cancer.
Int J Cancer
2008
;
122
:
2767
–73.
30
Ogino S, Meyerhardt JA, Cantor M, et al. Molecular alterations in tumors and response to combination chemotherapy with gefitinib for advanced colorectal cancer.
Clin Cancer Res
2005
;
11
:
6650
–6.
31
Ogino S, Kawasaki T, Kirkner GJ, et al. Down-regulation of p21 (CDKN1A/CIP1) is inversely associated with microsatellite instability and CpG island methylator phenotype (CIMP) in colorectal cancer.
J Pathol
2006
;
210
:
147
–54.
32
Durrleman S, Simon R. Flexible regression models with cubic splines.
Stat Med
1989
;
8
:
551
–61.
33
Cheng YW, Pincas H, Bacolod MD, et al. CpG island methylator phenotype associates with low-degree chromosomal abnormalities in colorectal cancer.
Clin Cancer Res
2008
;
14
:
6005
–13.
34
Jorissen RN, Lipton L, Gibbs P, et al. DNA copy-number alterations underlie gene expression differences between microsatellite stable and unstable colorectal cancers.
Clin Cancer Res
2008
;
14
:
8061
–9.
35
Etienne-Grimaldi MC, Formento JL, Francoual M, et al. K-Ras mutations and treatment outcome in colorectal cancer patients receiving exclusive fluoropyrimidine therapy.
Clin Cancer Res
2008
;
14
:
4830
–5.
36
Suehiro Y, Wong CW, Chirieac LR, et al. Epigenetic-genetic interactions in the APC/WNT, RAS/RAF, and P53 pathways in colorectal carcinoma.
Clin Cancer Res
2008
;
14
:
2560
–9.
37
Ortega P, Moran A, de Juan C, et al. Differential Wnt pathway gene expression and E-cadherin truncation in sporadic colorectal cancers with and without microsatellite instability.
Clin Cancer Res
2008
;
14
:
995
–1001.
38
Sullivan LM, Smolkin ME, Frierson HF, Jr., Galgano MT. Comprehensive evaluation of CDX2 in invasive cervical adenocarcinomas: immunopositivity in the absence of overt colorectal morphology.
Am J Surg Pathol
2008
;
32
:
1608
–12.
39
Rozek LS, Lipkin SM, Fearon ER, et al. CDX2 polymorphisms, RNA expression, and risk of colorectal cancer.
Cancer Res
2005
;
65
:
5488
–92.
40
Woodford-Richens KL, Halford S, Rowan A, et al. CDX2 mutations do not account for juvenile polyposis or Peutz-Jeghers syndrome and occur infrequently in sporadic colorectal cancers.
Br J Cancer
2001
;
84
:
1314
–6.
41
Wicking C, Simms LA, Evans T, et al. CDX2, a human homologue of Drosophila caudal, is mutated in both alleles in a replication error positive colorectal cancer.
Oncogene
1998
;
17
:
657
–9.
42
Sivagnanasundaram S, Islam I, Talbot I, Drummond F, Walters JR, Edwards YH. The homeobox gene CDX2 in colorectal carcinoma: a genetic analysis.
Br J Cancer
2001
;
84
:
218
–25.
43
Hinoi T, Loda M, Fearon ER. Silencing of CDX2 expression in colon cancer via a dominant repression pathway.
J Biol Chem
2003
;
278
:
44608
–16.
44
Hansel DE, Maitra A, Lin JW, et al. Expression of the caudal-type homeodomain transcription factors CDX 1/2 and outcome in carcinomas of the ampulla of Vater.
J Clin Oncol
2005
;
23
:
1811
–8.
45
Fan Z, Li J, Dong B, Huang X. Expression of Cdx2 and hepatocyte antigen in gastric carcinoma: correlation with histologic type and implications for prognosis.
Clin Cancer Res
2005
;
11
:
6162
–70.
46
Johns LE, Houlston RS. A systematic review and meta-analysis of familial colorectal cancer risk.
Am J Gastroenterol
2001
;
96
:
2992
–3003.
47
Kune GA, Kune S, Watson LF. The effect of family history of cancer, religion, parity and migrant status on survival in colorectal cancer. The Melbourne Colorectal Cancer Study.
Eur J Cancer
1992
;
28A
:
1484
–7.
48
Slattery ML, Kerber RA. The impact of family history of colon cancer on survival after diagnosis with colon cancer.
Int J Epidemiol
1995
;
24
:
888
–96.
49
Chan JA, Meyerhardt JA, Niedzwiecki D, et al. Association of family history with cancer recurrence and survival among patients with stage III colon cancer.
JAMA
2008
;
299
:
2515
–23.
50
Bass AJ, Meyerhardt JA, Chan JA, Giovannucci EL, Fuchs CS. Family history and survival after colorectal cancer diagnosis.
Cancer
2008
;
112
:
1222
–9.
51
Zell JA, Honda J, Ziogas A, Anton-Culver H. Survival after colorectal cancer diagnosis is associated with colorectal cancer family history.
Cancer Epidemiol Biomarkers Prev
2008
;
17
:
3134
–40.