IGF2BP3 (IMP3) expression is a marker of unfavourable prognosis in colorectal cancer

Study population

We utilized a database of two U.S. nationwide prospective cohort studies, the Nurses’ Health Study (N=121,701 women followed since 1976), and the Health Professionals Follow-up Study (N=51,529 men followed since 1986).¹⁵ Colorectal cancer cases were ascertained by means of biennial questionnaire and the National Death Index, which was also used for ascertainment of death. The cause of death was determined through medical record review by study physicians. We collected formalin-fixed paraffin-embedded tissue blocks from hospitals where surgical resections had been performed.¹⁵ We collected diagnostic biopsy specimens from rectal cancer patients who had received preoperative treatment, to minimize treatment-induced bias or artefact. Based on tissue availability and IGF2BP3 expression data, we included 671 stage I–IV colorectal cancer cases diagnosed up to 2004 (Table 1). Patients were observed until death or January 1, 2011, whichever came first. Informed consent was obtained from all study subjects. This study was approved by the Human Subjects Committees of Harvard School of Public Health and Brigham and Women’s Hospital.

Histopathological evaluation

Haematoxylin and eosin-stained sections of all cases were examined by a pathologist (S.O.) unaware of other data. Tumour differentiation was categorized as well-moderate or poor (>50% vs. ≤50% gland formation). Tumour growth pattern at the tumour margin was categorized as expansile, intermediate, or infiltrative, as previously described.¹⁶ The presence and extent of mucinous and/or signet ring cell component was recorded.¹⁷ A subset of cases (N>100) was reviewed by a second pathologist (T.M.), and concordance was as follows (all p<0.0001): κ=0.72 for tumour differentiation; κ=0.73 for tumour growth pattern; Spearman r=0.87 for mucin (%); Spearman r=0.65 for signet ring cells (%).¹⁶

DNA extraction, Pyrosequencing of KRAS, BRAF and PIK3CA, and microsatellite instability (MSI) analyses

DNA was extracted from paraffin-embedded tissue. Mutation status for KRAS (codons 12 and 13),¹⁸ BRAF (codon 600),¹⁹ and PIK3CA (exons 9 and 20)²⁰ was determined by Pyrosequencing. PCR-based MSI analysis was performed using 10 microsatellite markers (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67 and D18S487).²¹ MSI-high was defined as instability in ≥30% of the markers, and MSI-low/microsatellite stability (MSS) as instability in <30% of the markers.²¹

Methylation analyses for CpG islands and LINE-1

Using real-time PCR (MethyLight) on bisulfite-converted DNA, we quantified DNA methylation in eight CIMP-specific promoters (CACNA1G, CDKN2A, CRABP1, IGF2, MLH1, NEUROG1, RUNX3 and SOCS1).²² CIMP-high was defined as the presence of ≥6/8 methylated promoters, and CIMP-low/0 as 0–5/8 methylated promoters, in accordance with previously established criteria.²² In order to accurately quantify methylation changes within LINE-1, we used Pyrosequencing as previously described.²³

Immunohistochemical analysis

Tissue microarrays (TMA) were constructed as previously described.^24,25 Each tumour was represented by a minimum of two 0.6 mm tissue cores, which has previously been shown to be sufficient for assessment of IGF2BP3 expression.²⁶ Immunohistochemistry was performed as previously described for TP53,²⁷ CTNNB1 (β-catenin),¹⁵ PTGS2 (cyclooxygenase-2)²⁴ and FASN.¹⁷ For IGF2BP3, antigen retrieval from deparaffinised tissue sections was achieved by microwaving in a pressure cooker for 10 minutes in Citra Antigen Retrieval Solution, pH 6 (BioGenex, San Ramon, CA). Tissue sections were incubated for 15 minutes with Peroxidase Blocking Reagent (Dako North America, Carpinteria, CA). A mouse monoclonal antibody against IGF2BP3 (IMP3; clone 69.1; Dako; 1:100 dilution) was applied, and the sections were incubated for 16 hours at 4°C. Visualization was achieved using EnVision^™+/HRP (Dako), diaminobenzidine (Dako), and hematoxylin counterstain. Human placenta was used as a positive control tissue. Omission of primary antibody served as a negative control. Cytoplasmic staining for IGF2BP3 was scored as absent, weak, moderate, or intense. For subsequent analyses, absent or weak staining was categorized as negative, while moderate or intense staining was categorized as positive.

Each immunohistochemical maker was interpreted by one of the investigators (IGF2BP3 and CTNNB1 by T.M.; TP53, PTGS2 and FASN by S.O.) unaware of other data. For agreement studies, over 100 randomly-selected cases were examined for each marker by a second observer (IGF2BP3 by P.L.; CTNNB1 by S.O.; TP53, PTGS2 and FASN by T.M.) unaware of other data. The concordance between observers (all p<0.0001) was κ=0.70 for IGF2BP3, κ=0.78 for TP53, κ=0.80 for CTNNB1, κ=0.69 for PTGS2, κ=0.80 for FASN, indicating substantial agreement.

Statistical analysis

All statistical analyses were performed using SAS (Version 9.2, SAS Institute, Cary, NC). All p values were two-sided. Our primary hypothesis was that IGF2BP3 expression was associated with worse clinical outcome. When multiple hypothesis testing was performed for exploratory analyses of clinicopathological and molecular associations, or interactions, the p value for significance was adjusted by Bonferroni correction to p=0.0024 (=0.05/21). For categorical data, the chi-square test was performed. Mean age was compared using a t-test.

Kaplan-Meier method and log-rank test were used for survival analyses. For analyses of colorectal cancer-specific mortality, deaths as a result of other causes were censored. To control for confounding, multivariate Cox proportional hazards regression models were used. The multivariate model initially included sex, age at diagnosis (continuous), year of diagnosis (continuous), BMI (<30 vs. ≥30 kg/m²), family history of colorectal cancer in any first-degree relative (present vs. absent), tumour location (proximal vs. distal), tumour differentiation (well-moderate vs. poor), tumour growth pattern (expansile-intermediate vs. infiltrative), mucinous component (continuous), signet ring cell component (continuous), MSI (high vs. low/MSS), CIMP (high vs. low/0), KRAS, BRAF, and PIK3CA mutation, LINE-1 methylation (continuous), and expression of TP53, nuclear CTNNB1, PTGS2 and FASN. Because we observed non-proportionality in the hazards for disease stage over time, and to avoid overfitting, stage (I, II, III, IV or unknown) was used as a stratifying variable using the “strata” option in the SAS “proc phreg” command. A backward elimination was performed with a threshold of p=0.20, to avoid overfitting. Cases with missing information in any of the covariates were assigned to the majority category of the given covariate to avoid overfitting. We confirmed that excluding cases with missing covariate information did not substantially alter results (data not shown). The proportionality of hazards assumption was satisfied for all variables in the multivariate models by evaluating time-dependent variables, which were the cross-product of a given variable and survival time. Interaction was assessed by the Wald test on the cross product of IGF2BP3 and another variable of interest (without data-missing cases) in a multivariate Cox model.

IGF2BP3 status in colorectal cancer and normal colorectal mucosa

Among the database of 671 stage I–IV colorectal cancers in the two prospective cohort studies, 234 (35%) showed IGF2BP3 positivity (Figure 1). In contract, IGF2BP3 was negative in normal colorectal epithelial cells in all the 403 adjacent normal mucosal specimens that were evaluable. In normal colorectal tissue, only lymphocytes in the germinal centres of mucosa-associated lymphoid tissue stained positively for IGF2BP3 (Figure 1). As demonstrated in Table 1, IGF2BP3 positivity was significantly associated with poor differentiation (p=0.0003). Because multiple hypothesis testing required a Bonferroni corrected significance level of p=0.0024, the relationships between IGF2BP3 positivity and stage III–IV disease (p=0.0081), BRAF mutation (p=0.031), and LINE-1 hypomethylation (p=0.020) were considered of borderline significance; these findings require validation in an independent dataset.

IGF2BP3 positivity and survival of colorectal cancer patients

Among the 671 patients, there were 328 deaths, including 194 colorectal cancer-specific deaths, during a median follow-up of 160 months (interquartile range, 123–207 months) for censored cases. In Kaplan-Meier analysis, IGF2BP3 positivity was significantly associated with shorter colorectal cancer-specific (log-rank p<0.0001) and overall survival (log-rank p=0.0004) (Figure 2).

We used Cox multivariate regression models to assess the prognostic influence of IGF2BP3 positivity independent of clinical, pathological, and other tumour molecular features (Table 2). Univariate hazard ratio (HR) for IGF2BP3 positivity was 1.76 [95% confidence interval (CI), 1.32–2.33] for colorectal cancer-specific survival, and 1.48 (95% CI, 1.19–1.85) for overall survival. The multivariate HR for IGF2BP3 positivity was 1.37 (95% CI, 1.02–1.84) for colorectal cancer-specific survival, and 1.32 (95% CI, 1.05–1.66) for overall survival. The attenuation of the effect of IGF2BP3 positivity in the multivariate analysis occurred largely as a result of adjusting for disease stage. When adjusted for disease stage alone, the HR for IGF2BP3 positivity was 1.39 (95% CI, 1.04–1.85) for colorectal cancer-specific survival, and 1.30 (95% CI, 1.04–1.63) for overall survival (Table 2). No other major confounder was present.

Given that stage appeared to be a mediator of the association between IGF2BP3 expression and survival, we performed Kaplan-Meier analyses stratified by disease stage (Supplementary figures 1 and 2). A significant difference in survival time distribution according to IGF2BP3 expression was observed only for colorectal cancer-specific survival in stage II disease (Supplementary figure 1, B; log-rank p=0.0005), and with borderline significance for stage II overall survival (Supplementary figure 2, B; log-rank p=0.071). Similarly, in stage-specific Cox regression, IGF2BP3 was significantly associated with colorectal cancer-specific survival only in stage II disease (multivariate HR=2.50; 95% CI, 1.14–5.47). We did not, however, observe a significant interaction between disease stage and IGF2BP3 positivity, indicating that the prognostic association of IGF2BP3 did not differ significantly by stage.

Exploratory analyses of interactions between IGF2BP3 and other covariates

We further examined whether any of the clinical, pathological, or molecular variables modified the influence of IGF2BP3 positivity on cancer-specific mortality. We found no significant interaction between IGF2BP3 and any of the covariates examined (all P_interaction>0.10). Notably, the effect of IGF2BP3 positivity did not significantly differ between the two cohort studies (P_interaction=0.33).