Temporal stability and age-related prevalence of loss of imprinting of the insulin-like growth factor-2 gene

Abstract

Background:

Loss of genomic imprinting (LOI) of the insulin-like growth factor-2 gene (IGF2) is an epigenetic change involving abnormal activation of the normally silent maternally inherited allele. LOI of IGF2 gene is found in tumor tissue, normal adjoining mucosa and peripheral blood lymphocytes (PBL) of some patients with colorectal cancer (CRC), suggesting that this alteration precedes and is a risk factor for CRC. However, whether LOI of IGF2 is transitory or remains a permanent epigenetic alteration is unknown.

Results:

Four-hundred patients, mean age 60.7 years (range 15–95), 287 (80%) Caucasian were studied. This included 210 (51.4%) patients with no colorectal neoplasia, and 190 (48.6) with colorectal neoplasia. LOI of IGF2 was present in all age strata examined, and no statistically significant association across age strata (p trend > 0.05) was noted. Forty-nine patients had repeat analysis of blood imprinting status at a mean follow up time of 38.2 ± 12.9 months. All but three patients had the same imprinting status at follow up (94% agreement, kappa 0.79, p < 0.001). Genomic imprinting was stable for patients with and without colorectal neoplasia.

Methods:

Standard RT-PCR assays for imprinting analysis of IGF2 were performed on PBL from ApaI informative individuals recruited at baseline and repeated 1 to 3 years later. Prevalence of LOI of IGF2 was also evaluated according to age strata.

Conclusion:

LOI of the IGF2 gene in PBL appears to be a stable epigenetic phenomenon in most patients. Furthermore, LOI of IGF2 was not associated with age, suggesting an inherited or congenital epigenetic event. These findings support the concept that LOI of IGF2 may be a useful risk factor for CRC predisposition.

Introduction

Genomic imprinting is an epigenetic modification that results in the silencing of a specific allele dependent on its parental origin. Genomic imprinting does not involve alterations in the DNA sequence but, rather, methylation and likely other modifications heritable during cell division.¹ Loss of genomic imprinting (LOI) is seen when there is relaxation of the silenced allele and monoallelic expression changes to biallelic expression. The insulin-like growth factor-2 gene (IGF2) is the first gene reported to be normally imprinted and expressed exclusively from the paternal allele in both human and mouse. IGF2 is the first imprinted gene displaying LOI in human cancers. LOI of IGF2 results in activation of the normally silent maternal allele, with increase in IGF2 protein expression and potential increase cancer risk. Loss of imprinting of IGF2 gene was first discovered in embryonal tumors of childhood, such as Wilms tumor, but is found commonly in the tumor specimens of many types of cancer, including ovarian, lung, liver and colon.² LOI leads to increased expression of IGF2 an important autocrine growth factor for many malignancies including colorectal cancer (CRC).^3-7 The mitogenic effects of IGF2 protein are mediated by signaling through the IGF1 receptor, and blocking this receptor inhibits tumor cell growth.^8-11

Human studies have found an independent positive association between LOI of IGF2 and both personal and family history of colorectal neoplasia.^12,13 LOI of IGF2 occurs in colorectal cancer tissue, adjoining normal colorectum, and in peripheral blood lymphocytes (PBL).¹² Furthermore, animal studies using a LOI of IGF2 mouse model support an increased intestinal neoplasia risk and abnormal colonic mucosal differentiation.¹⁴ Also, LOI of IGF2 appears to be associated with a human colorectal cancer phenotype involving younger age at diagnosis, more advanced disease, right side colonic location, and poorly differentiated or mucinous carcinoma.^15,16 These findings support a role for LOI of IGF2 in colorectal carcinogenesis.

Presently, it is unknown whether LOI of IGF2 is present in the germline or acquired after birth. Association of LOI with younger age and imprinting status stability would support the hypothesis that this phenomenon is a germline rather than an acquired event. We therefore, examined the prevalence of LOI of IGF2 in children and adults. Moreover, previous investigations have examined imprinting status using cross-sectional study designs, which limits interpretation of the data as only association of the epigenetic status with the disease are calculated. Consequently, we evaluated whether LOI of IGF2 is a stable epigenetic phenomenon in patients who had baseline and follow up imprinting determination in PBL. We hypothesized that the imprinting status would remain stable in the PBL of individuals regardless of the baseline imprinting status or the neoplasia state.

Results

A total of 400 informative patients (221 women), mean age 60.7 (range 15–95) were evaluated. Most of the study population was Caucasian (80.0%), and non-Hispanic (74.9%). Two-hundred ten individuals had normal colonoscopies, while 190 had colorectal neoplasia (78 colorectal cancer and 112 colonic adenomas). The mean follow up time was 38.2 ± 12.9 months. Clinical characteristics of the study subjects are described in Table 1.

Table 1.

Clinical characteristics of study subjects (N = 400)

Clinical characteristic	All subjects (N = 400)	No colorectal neoplasia (N = 210)	Colorectal neoplasia (N = 190)
Gender
Female (%)	221 (55.3)	127 (60.5)	94 (49.5)
Male (%)	179 (44.8)	83 (39.5)	96 (50.5)
Age in years (mean ± SD); range	60.7 ± 14.1; 15–95	58.2 ± 14.6; 15–88	63.3 ± 13.1; 21–95
Follow up time (months ± SD)	38.2 ± 12.9	37.1 ± 11.9	35.2 ± 12.1
Racea
Caucasians (%)	287 (80.0)	154 (77.4)	133 (83.1)
African-Americans (%)	38 (10.6)	22 (11.1)	16 (10.0)
Other (%)	34 (9.5)	23 (11.6)	11 (6.9)
Ethnicity
Non-Hispanic (%)	304 (76.0)	148 (70.5)	156 (82.1)
Hispanic (%)	96 (24.0)	62 (29.5)	34 (18.4)

^aRace not self-identified in 41 subjects.

Prevalence of LOI of IGF2 in different age strata

The prevalence of LOI of IGF2 varied from 0% to 28.6% with the highest prevalence in the 15–21 age-category and the lowest prevalence in the 30–39 year age-group (Table 2). There was no statistically significant difference in the prevalence of LOI of IGF2 according to age-strata in the complete study population (p trend = 0.34), the no colorectal neoplastic group (p trend = 0.77) or the colorectal neoplastic group (p trend 0.17). A separate analysis of age as a dichotomous variable (<60 vs ≥60 years) revealed lack of statistical association of LOI of IGF2 with age (21.5% vs. 21.9%, respectively, p = 0.92). Also, there was no statistically significant association of LOI of IGF2 with age among the various racially and ethnically defined groups (data not shown).

Table 2.

Prevalence of IGF2 imprinting status according to specified age categories

Age strata Years	All subjectsa (N = 400)		No colorectal neoplasia Sb (N = 210)		Colorectal neoplasiac (N = 190)
	Normal N = 315 N (%)	LOI N = 85 N (%)	Normal N = 179 N (%)	LOI N = 31 N (%)	Normal N = 136 N (%)	LOI N = 54 N (%)
15–21	5 (71.4)	2 (28.6)	4 (66.7)	2 (33.3)	1 (100.0)	0 (0.0)
22–29	4 (80.0)	1 (20.0)	3 (75.0)	1 (25.0)	1 (100.0)	0 (0.0)
30–39	12 (92.3)	1 (7.7)	9 (100.0)	0 (0.0)	3 (75.0)	1 (25.0)
40–49	38 (74.5)	13 (25.5)	23 (85.2)	4 (14.8)	15 (62.5)	9 (37.5)
50–59	79 (79.8)	20 (20.2)	46 (83.6)	9 (16.4)	33 (75.0)	11 (25.0)
60–69	88 (73.3)	32 (26.7)	55 (83.3)	11 (16.7)	33 (61.1)	21 (38.9)
≥70	89 (84.8)	16 (15.2)	39 (90.7)	4 (9.3)	50 (80.7)	12 (19.3)

Normal, normal imprinting; LOI, loss of imprinting; p values for the trend

^ap = 0.34,

^bp = 0.49,

^cp = 0.17.

Temporal stability of genomic imprinting

Forty-nine patients (12.5%) had determination of imprinting status on PBL at baseline and again at a mean follow up of 38.2 ±12.9 months. At baseline, seven patients had LOI of IGF2 and 42 had normal imprinting in PBL (Table 3). All but three patients had identical imprinting PBL status on follow up (kappa 0.79, p < 0.001). All subjects in the no colorectal neoplasia cancer group had identical PBL imprinting status at follow up, representing 100.0% agreement (kappa 1.0, p < 0.001). However, for 3 out of 23 individuals with colorectal neoplasia (adenomas), imprinting status was not concordant over time; at baseline they had normal imprinting status, whereas at follow-up they had LOI of IGF2 (87% agreement, kappa 0.68).

Table 3.

Baseline and follow up imprinting status according to baseline colorectal neoplasia status (N = 49)

	Baseline		Follow up		Kappaa
	Normal N (%)	LOI N (%)	Normal N (%)	LOI N (%)	(p value)
Total Study Population	42 (85.7)	7 (14.3)	39 (79.6)	10 (20.4)	0.79 (<0.001)
No Colorectal Neoplasia	24 (92.3)	2 (7.7)	24 (92.3)	2 (7.7)	1.00 (<0.001)
Colorectal Neoplasia	18 (78.3)	5 (21.7)	15 (65.2)	8 (34.8)	0.68 (0.0003)

Normal, normal imprinting; LOI, loss of imprinting;

^aKappa values <0.4 = lack of agreement; between 0.4–0.6 = poor agreement; between 0.6–0.8 = good agreement; >0.8 = excellent agreement.

Discussion

Previous investigation has shown that LOI of IGF2 is independently associated with both personal and family history of colorectal neoplasia.^12,13,19,20 Also, we have demonstrated that LOI of IGF2 is present throughout the colorectal mucosa and appears to be a constitutional phenomenon present in both peripheral blood lymphocytes and colonic mucosa of affected individuals.^12,13 Furthermore, the discovery of LOI in the PBL of individuals who expressed LOI of IGF2 in colonic tissue suggests that abnormal imprinting may predispose to colorectal neoplasia.

Evaluation of LOI of IGF2 in PBL as a biomarker for colorectal cancer risk requires assessment of this phenomenon in different age groups, including children, since an inherited epigenetic marker should be present from birth. In the present study, we showed that LOI of IGF2 occurred in children (15 to 20 years) and adults (21–95 years), and was not associated with age. Hence, LOI of IGF2 does not appear to be an age-related acquired phenomenon. However, determination of imprinting status in newborns and young children is needed to confirm our present results. Also, there was no statistically significant association of LOI of IGF2 with age among various racially and ethnically defined groups.

Before concluding that LOI of IGF2 is a reliable risk factor for colorectal cancer the temporal stability of the imprinting status of the IGF2 gene needs assessment. If LOI of IGF2 is a risk factor for colorectal neoplasia, its presence should precede the development of disease and remain constant. Alternatively, if LOI is an acquired phenomenon from aging or environmental exposure, LOI status may change over time, altering the individual's risk of disease accordingly. Hence, imprinting was reevaluated in the PBL of patients who were longitudinally followed for 1 to 3 years after initial baseline assessment. We observed that imprinting status in PBL was stable in most individuals examined, with only 3/49 (6.1%) patients exhibiting a change from retention of imprinting to LOI. Of note, all three patients with change to LOI had colorectal neoplasia at baseline. Unfortunately, reconfirmation of baseline imprinting status in these three persons was not possible due to lack of additional baseline sample. However, all samples at baseline and follow up were analyzed using the same methodology and repeated twice with internal controls making inaccurate results unlikely.

An important observation in this investigation is the presence of LOI of IGF2 in the PBL of non-CRN individuals. It has been proposed that LOI of IGF2 may be a constitutional epigenetic alteration occurring early in normal tissue and blood cells and predisposing to malignancy.²² A murine model of LOI of IFG2 confirmed that LOI in mice increased the IGF2 protein expression level and promoted the occurrence of intestinal tumors suggesting that the altered maturation of non-neoplastic tissue might be affecting cancer risk.¹⁴ Based on these and other data, Feinberg et al. suggest an epigenetic progenitor model for cancer.²² They propose that various cancers might have a common basis in a polyclonal epigenetic disruption of stem/progenitor cells which are then further altered by gatekeeper mutations of oncogenes and tumor suppressor genes.¹⁴ Also, LOI of IGF2 has been shown to be present in tissue specimens of common human malignancies including ovarian, lung, liver, colon and breast.²¹ Individuals with LOI of IGF2 in PBL may be at increase risk for developing colorectal and other non-colonic cancers. Prospective studies of individuals with and without LOI of IGF2 are warranted to test this hypothesis.

This is the first study, to our knowledge, of the assessment of temporal stability of a human imprinted gene. Although our findings support a possible inherited colorectal cancer risk marker several study limitations should be acknowledged. First, this investigation had a relatively small number of informative persons in some age strata, including children (n = 7). Nonetheless, LOI of IGF2 was present in all age-strata, supporting the hypothesis that LOI of IGF2 is not age-related or acquired but rather an inherited or congenital phenomenon. Second, the follow up time reported was relatively short and inferences about temporal stability of imprinting status are limited. Our data, however, demonstrated that most individuals had identical imprinted status over the time period examined. Finally, ideally enough biological specimens would have been available to repeat imprinting assays in all patients to confirm baseline results. However, identical methods were used on baseline and follow up blood samples and multiple controls were performed in every assay minimizing the likelihood of laboratory error.

In summary, this is the first report evaluating LOI of IGF2 among various age groups including children, and examining the temporal stability of imprinting of the IGF2 gene (or any human imprinted gene) in PBL. Our results add support to the concept that LOI of IGF2 is inherited or congenitally acquired and buttress the validity of this alteration as a biomarker for colorectal neoplasia risk. Further analysis should focus on longer follow up times, larger cohorts, younger age children and prospective determination of neoplasia among patients with known imprinting status at baseline.

Materials and Methods

Study design, population and data collection

We conducted a cross-sectional study of consecutive patients referred to three academic centers (Johns Hopkins Hospital, Cleveland Clinic Florida and University of Puerto Rico Cancer Center). Eligible subjects were those having a colonoscopy for any medical indication, children ≥15 years of age, and adults, who signed informed consent/assent. Individuals with familial adenomatous polyposis or hereditary non-polyposis colorectal cancer were excluded. For this analysis we included individuals that have been recruited for our previous investigation.¹² Clinical, demographic and exposure information were obtained with the use of validated questionnaires. A peripheral blood sample (30 ml) was obtained from every participant at baseline and 12–36 months later. Colonoscopic examinations were performed at baseline with a standard Olympus colonoscope by several endoscopists. The colonoscopic examination and pathology reports were used to determine histological diagnosis of current and/or previous colonic findings. Based on these data patients were classified as having no history of colorectal neoplasia or a history of colorectal neoplasia (adenomas or colorectal cancer). The study was approved by the Institutional Review Boards of each academic center.

Laboratory analysis

DNA and RNA preparation

Lymphocytes were separated from blood (30 ml) from every patient with Accuspin tubes (Sigma) using Ficoll-Paque Plus (Amersham Pharmacia Biotech) and centrifuged at 400 g at room temperature for 30 minutes. The lymphocyte layer was collected and washed once with PBS. The isolated lymphocyte pellets were immediately stored at −80°C until the assays were performed. DNA extraction was performed as previously described.¹⁷ RNA was extracted from lymphocytes with the RNeasy Mini Kit (Qiagen) shortly before reverse transcription (RT). Each RNA sample was quantified by spectrophotometry using a nanodrop (Thermo Scientific).

Screening for informative samples

The IGF2 exon 9 ApaI poly-morphism site was detected using PCR. The primer sequences are as follows: IGF2apFw: 5′-GCT GAG AAA CAA TTG GCA AAA-3′ and IGF2apRev: 5′-TGC ATG GAT TTT GGT TTT CA-3′. Genomic DNA (gDNA) was denatured at 94°C for 6 min, amplified by 35 cycles of 94°C for 45 sec, 60°C min for 1 min and 72°C for 1 min. PCR products were digested overnight with endonuclease ApaI at 30°C (10 unites of enzyme for 17 μl PCR product and 3.0 μl 10X buffer in a final volume of 30 μl). Digestion products were electrophoresed on 2% agarose and visualized with ethidium bromide. The allele A (not digested by ApaI) shows one band of 335 bp and allele G (digested by ApaI) shows two bands of 204 and 131 pb (Fig. 1A). Samples of heterozygous A/G at the ApaI polymorphism were considered as informative and selected to synthesize cDNA.¹⁸

Figure 1.

(A) ApaI digestion of PCR products amplified from gDNA in ten representative studied cases. Six cases lines 1, 3, 6, 8, 9 and 10 exhibited homozygosis for allele G which contains ApaI site and two bands (204 and 131 bp) are observed after ApaI digestion. Two cases, lines 2 and 5 exhibited homozygosis for allele A which lack the ApaI site and only one band (335 bp) is observed after ApaI digestion. Two cases, lines 4 and 7 exhibited heterozygosity for the allele A and allele G. (B) Electrophoresis of RT-PCR products representative samples. The allele G contains Apa I site and two bands (204 and 131 bp) are observed after ApaI digestion. The allele A lack the ApaI site and only one band (335 bp) is observed after ApaI digestion. Lines 2 and 3 show cases with normal imprinting which express only allele G. Line 5 show a cases with normal imprinting that only express allele A. Cases 1 and 4 show cases with biallelic expression or LOI of IGF2.

RNA reverse transcription and IGF2 imprinting assay

Total RNA was extracted from heterozygous samples using the RNeasy kit (Qiagen) following manufacturer instructions. To avoid any possible gDNA contamination, the RNA was treated with Turbo DNA-free kit (Ambion) for 40 min at 37°C, the DNase-Inactivator Reagent was added and incubated for 2 min at room temperature, then centrifuged at 10,000 RPM for 2 min, and the supernatant (RNA) was collected. For each reaction a total of 2 μg of RNA was used. The cDNA was synthesized by Retroscript kit (Ambion). Briefly, the RNA was mixed with, 4 μl of random decamer and water to a volume of 24 μl, and 72°C for 3 min. Later, 8 μl of 10 mM dNTP, 2 μl of Rnase inhibitor, 2 μl of 100 U/μl MMLV and 8 μl of 10X buffer, were added to a final reaction volume of 40 μl. Reaction was mixed and incubated for 1 hour at 44°C. Reverse transcriptase was inactivated through heating for 10 min at 92°C. Five μl of synthesized cDNA was amplified by PCR reaction (35 cycles) using the primers and conditions described above for IGF2 genotyping. Amplified PCR products were digested overnight with endonuclease ApaI at 30°C (10 unites of enzyme for 17 μl PCR product and 3.0 μl 10X buffer in a final volume of 30 μl). Digested products were electrophoresed on 2% agarose and visualized with ethidium bromide. The allele A (not digested by ApaI) shows one band of 335 bp and allele G (digested by ApaI) shows two bands of 204 and 131 pb. LOI of IGF2 was assigned when a three band pattern was observed; allele A one fragment of 335 bp fragment and allele G two bands 204 and 131 bp; determination of LOI status was based on the number of bands observed (Fig. 1B). All reactions at baseline and at follow up were done in duplicate with presence and absence of reverse transcriptase on identical RNA samples to exclude any possibility of DNA contamination.¹⁸ Pair samples (baseline and follow up) were run separately first and then run in the same batch to guarantee identical laboratory methods.

Statistical analysis

The primary outcome was presence of LOI evaluated in PBL (dichotomized as positive or negative LOI). The covariates evaluated included: gender (women—baseline category), age [mean ± standard deviation (SD)], ethnicity (Hispanic vs Non-Hispanic—baseline category), race (African American vs Caucasian—baseline category), and colorectal neoplasia (past or present history of adenoma and/or colorectal cancer vs none—baseline category).

To determine agreement in LOI status between baseline and follow-up samples, the Kappa statistic was calculated producing measurement of agreement with values of 0 to 1 (values <0.4 lack of agreement, between 0.4–0.6 poor agreement, between 0.6–0.8 good agreement and >0.8 excellent agreement). To test whether the prevalence of LOI of IGF2 differed across age categories, we divided age into 8 categories, calculated the prevalence of LOI in the baseline blood specimen, and used a nonparametric test for the trend (nptrend). Statistical analysis was performed using STATA 9.0 software (Stata Corporation, Texas, USA).