Loss of IGF2 Imprinting: A Potential Marker of Colorectal Cancer Risk

See allHide authors and affiliations

Science  14 Mar 2003:
Vol. 299, Issue 5613, pp. 1753-1755
DOI: 10.1126/science.1080902


Loss of imprinting (LOI), an epigenetic alteration affecting the insulin-like growth factor II gene (IGF2), is found in normal colonic mucosa of about 30% of colorectal cancer (CRC) patients, but it is found in only 10% of healthy individuals. In a pilot study to investigate the utility of LOI as a marker of CRC risk, we evaluated 172 patients at a colonoscopy clinic. The adjusted odds ratio for LOI in lymphocytes was 5.15 for patients with a positive family history [95% confidence interval (95% CI), 1.70 to 16.96; probability P = 0.002], 3.46 for patients with adenomas (95% CI, 1.14 to 11.37; P = 0.026), and 21.7 for patients with CRC (95% CI, 3.48 to 153.6; P = 0.0005). LOI can be assayed with a DNA-based blood test, and it may be a valuable predictive marker of an individual's risk for CRC.

Molecular genetics has played an important role in cancer risk assessment for rare hereditary cancer syndromes, such as CRC in familial adenomatous polyposis coli and hereditary nonpolyposis CRC (1). However, these syndromes cumulatively account for <1% of all cancer cases (2, 3). A critical challenge is to develop genetic tests to identify individuals in the general population with an elevated risk of developing cancer.

Genomic imprinting is a form of gene silencing that is epigenetic in origin; that is, it does not involve alterations in the DNA sequence but rather changes in DNA methylation and likely other modifications heritable during cell division (4). Loss of imprinting (LOI) of the insulin-like growth factor II gene (IGF2) was discovered in embryonal tumors of childhood, such as Wilms' tumor, but it is found commonly in many types of cancer, including ovarian, lung, liver, and colon (4). In Wilms' tumors, especially those with late onset, LOI results in the increased expression of IGF2 (5), an important autocrine growth factor for many cancers, including CRC (6–11).

Previously, we found LOI of IGF2 in the tumors and matched normal colonic mucosa of 30% of CRC patients, as compared with 10% of individuals without CRC (12). To determine whether LOI in normal tissue is associated with either a family history or personal history of colorectal neoplasia, we conducted a cross-sectional analysis of patients who were undergoing a colonoscopy for any medical indication and who had consented to contemporaneous mucosal biopsies for research purposes (13, 14). The medical indications included routine screening for individuals who were older than 50 years (45%), gastrointestinal symptoms (21%), family history (21%), or past history of cancer (12%). The study population was modestly enriched for a past history of colonic adenoma or cancer (10% in the general U.S. population of this age). A total of 421 patients agreed to participate between 1999 and 2001; 191 patients were informative for either an Apa I or CA repeat polymorphism within exon 9 of IGF2, allowing analysis of imprinting status, and in 172 patients, samples were adequate for study. We obtained both proximal and distal colonic mucosal specimens (up to four adjacent cold forceps biopsies from each region); peripheral blood lymphocytes (PBLs) for RNA and DNA analysis; and information from the patients on family history and personal history of neoplasia, environmental exposures, medications, and diet.

We examined the relation between LOI in PBLs and age, because it has been proposed that altered IGF2 methylation is age related, suggesting that epigenetic abnormalities are acquired over time (15). However, in our study, there was no statistically significant relation between age and LOI in bivariate analyses (P = 0.52, Wilcoxon rank-sum test) or in multivariate analyses, adjusting for race, family history, sex, or colorectal neoplasia.

Interestingly, the odds of LOI in individuals with a family history of CRC were 5.15 times that of individuals with no family history (P = 0.002) (Table 1). Similarly, the odds of LOI in participants with a personal history of colorectal neoplasia (adenomatous polyps or cancer) were 4.72 times that of participants with no history (P = 0.002) (Table 1), indicating a strong association between LOI and colorectal neoplasia. Even when patients with a positive family history were excluded from the analysis, the odds of LOI in PBLs among participants with colorectal neoplasia were 5.80 times that of participants without neoplasia [95% confidence intervals (CI), 1.88 to 16.29;P = 0.0007]. Consistent with the model of CRCs progressing from adenomas (16), the odds of LOI in patients with past or present adenomas but no CRC were 3.46 times that of patients with no past or present colorectal neoplasia (P = 0.026) (Table 1). The odds of LOI in patients with past or present CRC were 21.7 times that in patients without colorectal neoplasia (P = 0.0005) (Table 1). These data strongly suggest that LOI is associated with both initiation and progression of colorectal neoplasia.

Table 1

Association of LOI of IGF2 in PBLs with a family history of CRC and with a personal history of colorectal neoplasia. OR, odds ratio. Baseline category for OR indicated by dashes.

View this table:

The relation between LOI in PBLs and LOI in the colon was determined in 168 patients. Of these, all 25 patients with LOI in the blood also showed LOI in the colon (P < 0.0001) [kappa statistic = 86.5% (17)]. In 21 patients, LOI was limited to the colon, and there was no statistically significant association with family or personal history of colorectal neoplasia. In 31 patients, imprinting was examined in both the right and left colon (10 with LOI in the blood and 21 without). There was complete concordance of left and right colonic mucosa in all patients; that is, 10 patients with LOI in the blood showed LOI in both the left and right colon, 13 patients with normal imprinting in the blood showed LOI in both the left and right colon, and the other 8 patients showed normal imprinting in all three samples. Thus, LOI in the colon appeared to be a diffuse abnormality affecting both the proximal and distal colon. The absence of a cancer association in individuals with LOI in the colon but not in the blood suggests that in these individuals, the epigenetic abnormality was time- and/or tissue-limited.

A blood test of risk assessment is most practical if it is based on DNA rather than RNA. We recently identified in cancer patients a differentially methylated region (DMR) of three closely clustered CpG sites within IGF2, showing hypomethylation in CRC with LOI (18). To determine whether a similar hypomethylation defect occurs in the PBLs and colon of patients without colorectal neoplasia, we examined 24 samples: 12 from normal tissues (6 PBLs and 6 matched normal colonic mucosa) with normal imprinting and 12 from normal tissues (6 PBLs and 6 matched normal colonic mucosa) with LOI. In all 12 tissues with normal imprinting, IGF2 showed a normal pattern of half-methylation (fig. S1). In contrast, in 11 of 12 samples from normal tissue with LOI, IGF2 showed hypomethylation of the IGF2 DMR; in the other sample, IGF2 showed partial methylation of both alleles but was nevertheless abnormal (fig. S1). The significance of hypomethylation between normal tissues with and without LOI was P < 0.0001 (Fisher's exact test). Thus, aberrant IGF2 methylation is linked to LOI in normal colon and lymphocytes, just as it is in CRC.

We found a significant association of LOI with family history or personal history of colorectal neoplasia. When present in PBLs, LOI appeared to be a systemic abnormality, because it was always accompanied by LOI in both the proximal and distal colon. Thus, LOI appeared to be a constitutional defect in some patients and, in these patients, was associated with a positive family history and personal history of CRC. It was not clear whether LOI is present in the germ line or acquired postnatally. Because LOI has not been found commonly in normal pediatric tissues, such as the kidney (4,5), the simplest explanation of these findings is that the stringency of maintenance of IGF2 imprinting postnatally is itself genetically determined and is related to cancer risk. In addressing this issue, it will be important to determine the frequency of LOI in children, who were not included in our study. LOI appears to be common, as it was present in 14% of the patient group we studied (which by design was enriched for CRC) and in 10% of the general population (18, 19).

The prevalence of LOI in the general population (10%) is at least 10 times that of all known CRC-predisposing genetic mutations combined (2, 3). Consequently, conventional genetic mutation screening for cancer risk has been targeted at defined populations with a strong family history and not for screening and surveillance in the general population. In contrast, an LOI blood test might be of value for population screening. Two of the three criteria for such screening (20) have largely been met for CRC: its impact on public health and the benefit of intervention. Our study included 172 patients, and 25 of these patients showed LOI in PBLs. The third criterion for population screening, as with any genetic test, can be determined only through evaluation of a large prospective cohort of patients, establishing the positive and negative predictive values (20). It is also important to determine whether LOI is associated with other malignancies and whether this epigenetic abnormality is itself genetically determined.

Supporting Online Material

Materials and Methods

Fig. S1

References and Notes

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: afeinberg{at}


View Abstract

Stay Connected to Science

Navigate This Article