Increased expression of the insulin-like growth factor-II gene in Wilms' tumor is not dependent on loss of genomic imprinting or loss of heterozygosity.

Loss of imprinting of insulin-like growth factor-II gene (IGF2) and/or loss of heterozygosity at the 11p15 loci have been postulated to be responsible for IGF2 overexpression in Wilms' tumor. In order to delineate the mechanism of IGF2 overexpression in Wilms' tumors, we have genotyped the 11p15-11p13 chromosomal region and determined allelic expression of IGF2 and H19 in both tumor tissue and in normal adjacent kidney tissue from 40 patients with Wilms' tumor. In five of the eight subjects informative for the ApaI IGF2 polymorphism, loss of imprinting of IGF2 was observed in both normal and tumor tissues. A significant increase (>5-fold) in IGF2 expression in tumor tissues compared to the normal adjacent kidney tissue was observed regardless of the IGF2 imprinting or the chromosome 11p15 heterozygosity status. In each case, the overexpression of IGF2 in the tumors was accompanied by activation of all four IGF2 promoters. Our data indicate that alterations of IGF2 imprinting occurred in normal adjacent kidney tissue before tumorigenesis and that the IGF2 overexpression in Wilms' tumor tissue occurs through a loss of heterozygosity- or loss of imprinting-independent process.

Loss of imprinting of insulin-like growth factor-II gene (IGF2) and/or loss of heterozygosity at the 11p15 loci have been postulated to be responsible for IGF2 overexpression in Wilms' tumor. In order to delineate the mechanism of IGF2 overexpression in Wilms' tumors, we have genotyped the 11p15-11p13 chromosomal region and determined allelic expression of IGF2 and H19 in both tumor tissue and in normal adjacent kidney tissue from 40 patients with Wilms' tumor. In five of the eight subjects informative for the ApaI IGF2 polymorphism, loss of imprinting of IGF2 was observed in both normal and tumor tissues. A significant increase (>5fold) in IGF2 expression in tumor tissues compared to the normal adjacent kidney tissue was observed regardless of the IGF2 imprinting or the chromosome 11p15 heterozygosity status. In each case, the overexpression of IGF2 in the tumors was accompanied by activation of all four IGF2 promoters. Our data indicate that alterations of IGF2 imprinting occurred in normal adjacent kidney tissue before tumorigenesis and that the IGF2 overexpression in Wilms' tumor tissue occurs through a loss of heterozygosity-or loss of imprinting-independent process.
Wilms' tumor, one of the most common childhood tumors (1:10000, 8% of childhood tumors), is associated with abnormalities in the short arm of human chromosome 11 (1). Somatic chromosomal events leading to the loss of constitutional heterozygosity (LOH) 1 of this chromosomal region occur in 30 -57% of Wilms' tumors (1)(2)(3). Among the cluster of genes residing at this chromosomal region that have been implicated in tumorigenesis are IGF2 and H19. These genes are located within 200 kilobases of each other on chromosome 11p15.5, and both genes are imprinted. IGF2, spanning ϳ30 kilobases at this region, codes for a 67-amino acid mitogenic peptide that plays an important role in fetal growth and development (4). In most normal human tissues except for liver and CNS, IGF2 is transcribed exclusively from the paternally derived allele (5,6). A number of malignant tumors, including Wilms' tumor, have been shown to contain abnormally high levels of IGF2 mRNA, and loss of IGF2 imprinting, where both paternal and maternal alleles are transcribed, has been observed in many of these tumors (7,8). The regulation of IGF2 expression in Wilms' tumor has been studied extensively as a model system for examining mechanisms of genomic imprinting and loss of imprinting.
In an earlier study, we quantitated the allelic and promoterspecific expression of IGF2 in Wilms' tumor, comparing tumors in which imprinting was lost (LOI) with those in which imprinting was maintained (MOI, maintenance of imprinting). We demonstrated that while the normal imprinting of IGF2 is promoter-specific and may fluctuate during early development, all four IGF2 promoters are expressed biallelically when there is LOI in Wilms' tumors (9). Coordinate LOI of all four IGF2 promoters in tumor has also reported by others (10,11). IGF2 may act as an autocrine or paracrine growth factor when it is produced by tumors, potentially maintaining and enhancing tumor growth (12,13). It has been suggested that the increased expression of IGF2 observed in some of these tumors is associated with, or caused by, the relaxation of genomic imprinting and the subsequent biallelic expression of the gene. In theory, biallelic IGF2 expression should lead to increased mRNA and protein synthesis.
H19 is a gene which is highly expressed during development in a number of fetal tissues but not in most adult tissues (14 -16). It is thought that H19 exerts its effects as an RNA molecule and not as a protein, since there is no significant open reading frame in the gene (17). In contrast to IGF2, only the maternal allele of the H19 gene is transcribed (18). Alterations in H19 imprinting have also been reported in Wilms' tumor (7,8). Interestingly, loss of imprinting of IGF2 has been linked to reduced expression and abnormal methylation of H19 in Wilms' tumors (19,20). A similar linkage of overexpression and LOI of IGF2 has been found in children with IGF2-induced overgrowth disorder (21).
In this report we have compared the allelic expression of IGF2 and H19 in 40 sets of Wilms' tumor specimens. In order to ascertain the role of LOI and LOH in the overproduction of IGF2 in Wilms' tumor, we quantitated IGF2 mRNA levels in the two groups of Wilms' tumors: LOH and non-LOH (LOI and MOI groups). IGF2 expression was measured in tumor and in normal adjacent kidney tissue (NAT) from the same patient. We demonstrate that LOI can be observed in some NAT samples and that IGF2 expression is increased in all categories of Wilms' tumor tissues compared to NAT, suggesting that mechanisms other than LOI and LOH with consequent biallelic expression of IGF2 can result in increased IGF2 mRNA abundance in Wilms' tumors.

MATERIALS AND METHODS
Human Tissue and Nucleic Acid Preparation-Forty pairs of fresh frozen Wilms' tumors and their normal adjacent kidney tissues were obtained from the Cross Cancer Institute Northern Alberta Cancer Program. Total nucleic acid (TNA, containing RNA and DNA) was extracted from fresh frozen tissues as described previously (9). From the TNA preparations, RNA and DNA were prepared using RNA-DNA STAT (Tel-Test, Friendswood, TX). RNA was also prepared by simply removing DNA from TNA using DNase I treatment (9). cDNA was synthesized by RNA reverse transcription and diluted 10-fold before PCR amplification as described previously (9).
Measurement of Relative mRNA Levels by PCR Analysis-To quantitate relative mRNA levels of IGF2 and H19 in tumors and normal tissues, we simultaneously amplified cDNAs derived from IGF2, H19, and 18 S rRNA in a multiplex PCR reaction. Variation among PCR reactions and sampling was therefore minimized. We used a mixture of two sets of primers of IGF2 (p23-p24) and H19 (p11-p12) at a 0.08 M final concentration, and an internal control 18 S rRNA (18SF and 18SR) at a 0.04 M concentration. The IGF2 and H19 primers were designed to amplify DNA crossing intron-exon borders to differentiate genomic DNA and cDNA products. Amplification and analysis of PCR products were performed as described above, omitting the restriction digestion.
Analysis of Promoter Usage-To compare IGF2 promoter usage in tumor tissues versus NAT, we utilized a multiplex PCR (9). Tumor and NAT cDNAs were amplified with four promoter-specific 5Ј primers (p17, p18, p19, and p20) and a common 3Ј primer (p22) end-labeled with [␥-32 P]ATP. In the PCR reaction, the four promoter-specific 5Ј primers compete with each other for the one end-labeled 3Ј primer, thereby reflecting the relative abundance of the promoter-specific derived transcripts (31). The PCR conditions were the same as those used for quantitation analysis.  Table I shows the DNA polymorphisms found within these five genes in the 40 pairs of Wilms' tumor and normal adjacent tissues. Heterozygous specimens were informative for further analysis of loss of heterozygosity and imprinting status. Among the eight DNA markers, the most informative site was the TH-CATT repeat site at which 65% (26/40) of cases were shown to be heterozygous, and the least informative site was INS-PstI, where only 20% (8/40) were informative (Table II). We observed eight paired specimens to be heterozygous for the ApaI polymorphism of IGF2, and eight pairs for the H19-AluI polymorphism, which were thus informative for further examination of genomic imprinting ( Table I).
Loss of Heterozygosity Involving Deletion-Duplication in Tumorigenesis-To determine the presence or absence of LOH at chromosome 11p, we compared genotypes in the tumors with those in their paired normal tissues. When LOH was observed in one polymorphic site, all other sites, if informative, also showed LOH, suggesting that the LOH in the Wilms' tumor DNA extended from the TH through the WT1 locus except for patient 10, in which LOH did not involve 11p13 (WT1); these findings are consistent with previous studies (2,32,33). Considering all of the cases which were informative for these eight DNA markers, 14/37 (37.8%) of Wilms' tumors demonstrated LOH for the short arm of chromosome 11. Because of the strong association of LOH of these genes across the region from 11p15.5 to 11p13, we classified tissues showing LOH in any of the eight DNA markers into the LOH group. There were 14 cases of LOH and 23 cases of maintenance of heterozygosity (MOH); three cases could not be classified due to a lack of informative sites available in any of these eight DNA markers (Table I).
Gene expression in tumor demonstrating loss of heterozygosity was quantified by densitometric analysis. The average ratio of the density of the remaining allele bands in eight tumors with LOH for H19-AluI and five tumors with LOH for WT1-HinfI to that of the two allele bands in their corresponding normal tissues was 0.96 Ϯ 0.07 and 1.24 Ϯ 0.13, respectively, suggesting that the loss of one allele was accompanied by the duplication of the other allele (isodisomy). To confirm the presence of LOH with deletion-duplication in tumors as previously reported (34), we used end-labeled primers and an internal control 18 S ribosomal RNA to assess gene duplication in the LOH samples at TH-CATT and IGF2-ApaI sites (Fig. 3). The normalized ratio revealed that the intensity of the band of the one remaining allele in the tumor was equal to the total density of the two alleles seen in normal adjacent tissue DNA (p ϭ 0.84), confirming the presence of a deletion-duplication in LOH (35).
Loss of Imprinting of IGF2 Predisposes to Tumorigenesis-Allelic expression of IGF2 and H19 was investigated using primers flanking the polymorphic IGF2-ApaI (p9, p10) and H19-AluI (p11, p12) sites, respectively (Fig. 4). Among eight informative cases for IGF2, monoallelic expression of IGF2 was found in two paired tissues (patients 15 and 20), while the other six tumors showed biallelic expression of IGF2 (LOI). Interestingly, LOI was also observed in five of the six normal tissues adjacent to the tumors that showed LOI for IGF2 (Table  III). To exclude possible contamination of the cDNA samples with genomic DNA, cDNAs from these samples were subjected to PCR using primers spanning intron 8 of IGF2 (p23 and p24, see Fig. 4, IGF2). Reverse transcription PCR with these primers produced only the 115-bp fragments characteristic of the cDNA without any detectable band derived from genomic DNA (408 bases, data not shown). Tumor tissue contamination of the NAT was not considered likely because the genotyping data of LOH patients clearly showed that the normal tissues differed from tumor tissues (for polymorphism heterozygosity analysis see Table I). In addition, the analysis of IGF2 abundance argues against the possibility of tissue cross-contamination in our specimens (see below). Thus, these data indicate that LOI frequently occurs in tissue adjacent to Wilms' tumors, suggesting the possibility that LOI leads to a proclivity for tumorigenesis.
As shown in Table III, of the eight tissue pairs informative for the H19-AluI polymorphism, only one tumor (patient 15) displayed biallelic expression of H19; there was maintenance of IGF2 imprinting in this case. H19 was monoallelically expressed in all remaining tumors and in all normal adjacent tissues including subjects 30, 31, and 32, where there was a loss of IGF2 imprinting, suggesting a lack of association or concordance between IGF2 and H19 genomic imprinting.
Overexpression of IGF2 in Wilms' Tumors-We next examined the relative abundance of IGF2 and H19 transcripts in the 40 pairs of Wilms' tumors and adjacent normal tissues using quantitative reverse transcription PCR. To exclude interference from genomic DNA contamination during PCR transcription quantitation, we used specific primers which cross exon/ intron borders for each gene. The PCR products derived from cDNA and DNA are therefore of different sizes (Fig. 4). We have optimized our PCR conditions to amplify simultaneously the transcripts of IGF2, H19, and 18 S ribosomal RNA (as an internal control) in a quantitative manner in the same tube and subsequently evaluate them in the same gel lane (Fig. 5).
As shown in Table III, a significantly increased abundance of IGF2 mRNA was observed in 40 tumors regardless of their imprinting or heterozygosity status (average 5.26 Ϯ 0.76-fold increase; p Ͻ 0.0001) compared to their NAT. No consistent changes in H19 gene expression could be discerned. Twentyeight of the 40 tumors showed decreased (Յ1.0-fold) expression of H19 and 12 displayed an increase (Ͼ1.0 fold) in H19 abundance when compared to their normal adjacent tissue. The average H19 abundance in Wilms' tumors was 90% relative to the normal tissues, but there was no significant difference between them (p Ͼ 0.05).
To determine the effects of LOH and LOI on the expression of IGF2 and H19 in Wilms' tumor, we analyzed the quantitation data by subgroups according to allelic expression analysis (Table III) and genotype data (Table I). Tumors which demonstrated LOH had a significant increase of IGF2 expression (6.74 Ϯ 1.66-fold compared to normal tissues) than did tumors which maintained heterozygosity (4.04 Ϯ 0.51-fold compared to normal tissues), indicating that LOH and isodisomic reduplication contributed to the overexpression of IGF2 in tumors. There was a significant decrease in H19 abundance in Wilms' tumors with LOH (29%, p Ͻ 0.05) and in those tumors which demonstrated loss of IGF2 imprinting (26%, p Ͻ 0.05) when compared with their normal adjacent tissues. No consistent changes in H19 gene expression could be discerned in the other MOH tumor group, confirming previous work which showed that the decrease of H19 in Wilm's tumors was related to LOH and LOI of IGF2 (19,20,36). However, we did not observe any relationship between the increase in IGF2 or the decrease in H19 in Wilms' tumors with their LOI/MOI imprinting status. In summary, these data suggest that the overexpression of IGF2 in Wilms' tumor cannot be directly or solely attributable to LOI or LOH mechanisms.
Activation of IGF2 Promoters in Tumorigenesis-The IGF2 gene has four promoters (37). Previously we observed that transcripts from promoters P1-P4 are derived from both alleles in Wilms' tumors with LOI. In contrast, in tumors with MOI, transcripts from P2-P4 originate from only one allele, while P1 transcripts were from one or both alleles (9). To see which promoter contributes to the increase of IGF2 expression in Wilms' tumors, we examined the promoter usage of IGF2 using the competitive PCR with end-labeled primers (Fig. 4, IGF2). Using this methodology, transcript products from only promoters P3 and P4 were observed in tumors or in normal tissues. Transcripts derived from promoters P1 and P2 were not detectable in either normal or malignant of tissues (Fig. 6A), confirming that P3 and P4 were dominant promoters in both normal and malignant kidney tissues (9,37). The expression of P3derived IGF2 transcripts was 6.42 Ϯ 1.0-fold higher in tumors than in the normal tissues, while expression from P4 was 3.48 Ϯ 0.83-fold higher in tumors. A similar trend was observed in tumors with either LOI (patients 11, 30, and 31)   (patients 15 and 20) (Fig. 6A).
To focus on whether promoter P1 and P2 were activated in Wilms' tumors, we examined each promoter by a more sensitive PCR analysis using single primer sets. As we have previously shown, IGF2 transcripts derived from P1 and P2 were observed in Wilms' tumors, but these transcripts could not be detected in normal adjacent tissues (Fig. 6B). The contribution of promoters P1 and P2 to the overexpression of IGF2 in Wilms' tumors was much less than the contribution from promoters P3 and P4. DISCUSSION IGF-II is a mitogenic peptide which is overexpressed in many types of tumors, and it has been speculated that the peptide can act in an autocrine or paracrine manner to initiate and/or maintain oncogenesis and tumor growth (12,13). The recent observation that IGF2 imprinting is lost in a majority of Wilms' tumors appears to have provided one potential mechanism for the increased abundance of IGF2 in such tumors since biallelic gene expression would presumably lead to twice the amount of mRNA as would be transcribed from an imprinted gene (7,8,35,38). In this study, we have characterized 40 sets of Wilms' tumors and their normal adjacent tissues. We confirm earlier findings that LOH is accompanied by reduplication of the remaining (presumably paternal) allele and by increased expression of the normally maternally imprinted IGF2 (1, 39). Moreover, we have shown that the abundance of IGF2 was Ͼ5 times greater in the tumor tissue than in the normal kidney tissue. This increased IGF2 expression occurred in tumor tissue irrespective of imprinting or heterozygosity status. Even NATs which demonstrated loss of IGF2 imprinting contained less IGF2 than did the tumor tissue which also manifested LOI, suggesting that loss of IGF2 imprinting may not directly alter IGF2 gene expression (or mRNA abundance) in Wilms' tumor.
Instead, it appears that other mechanisms not associated with imprinting must contribute to the overexpression of IGF2.
In Wilms' tumors, LOH was restricted to the short arm of FIG. 4. Diagram of the PCR assay used to determine the mRNA abundance level and imprinting status of IGF2 (top) and H19 (bottom) genes. The location of the oligonucleotide primers is indicated by the numbered arrowheads. The PCR products of DNA or RNA from appropriate primers are shown. The numbered boxes represent the gene exons. The location of the four promoters of the IGF2 gene are noted above the boxes. ApaI and AluI are polymorphism sites used for determining the imprinting status of the IGF2 and H19 genes, respectively. End-labeled primers for promoter usage analysis are indicated by asterisks. chromosome 11 (2). LOH can be detected at 11p15 to 13 in 30 -57% of informative samples of Wilms' tumor (1)(2)(3). In this study, we identified LOH in 37.8% (14 of 37) informative tumors, and we found linkage of LOH at 11p13 to 11p15. Due to the limited number of samples which were informative at each polymorphic site, the frequency of LOH that we determined at each polymorphic site might not reflect the exact incidence of LOH at these loci in Wilms' tumors. For example, the apparent low LOH rate at INS-PstI site (1/8, 12.5%) was probably due to the relatively few patients who were informative (i.e. heterozygous) at this site because another polymorphic site (HphI) of the INS gene displayed a higher incidence of LOH (5/11, 45.5%). LOH was restricted to the Wilms' tumors and was not observed in any of normal adjacent tissues. By contrast, LOI of IGF2 did occur in both tumor and normal tissues.
LOI of IGF2 has been reported in a variety of cancers including Wilms' tumor (7), rhabdomyosarcoma (40), uterine leiomyosarcoma (29), lung cancer (41), breast cancer (42), and testicular cancer (43). In the present study we confirmed that LOI of IGF2 occurs in Wilms' tumor, but surprisingly, we also observed LOI of IGF2 in the adjacent normal tissues, suggesting that LOI can occur prior to tumorigenesis. There are several reasons to believe that the normal tissues displaying LOI of IGF2 were not contaminated by their adjacent tumor tissues. First, LOH occurred only in the tumor tissues and not in any of the adjacent normal tissues; second, there was a significant difference in the abundance of IGF2 and of H19 transcripts between the normal and the malignant tissues, even when both tissues demonstrated LOI of IGF2; and finally, there were also significant differences in the expression of the WT1, DNA methyltransferase, and glypican-3 genes when comparing the normal tissues to the tumor tissues (data not shown).
In previous studies, increased IGF2 expression was observed in Wilms' tumors demonstrating LOI or LOH, but no statistically significant difference could be discerned because of the wide range of expression of IGF2 among the tumors in each group (20,36). By examining 40 paired Wilms' tumors/NAT  A, transcripts derived from promoters P1 to P4 from competition among the four promoters examined by a multiplex PCR assay using four promoter-specific 5Ј primers (p17, p18, p19, and p20) and one common 3Ј primer (p22) 32 P-end labeled (see Fig. 4). Only transcripts from P3 and P4 were observed. B, transcripts from each promoter in Wilms' tumor and their adjacent normal tissues of patient 15 separately examined by a standard PCR assay using a single set of corresponding primers with a 18 S primer set, showing activation of four promoters of IGF2 in Wilms' tumor. The same primers were used as above except for amplification of promoter P2 (p17-p21). T, Wilms' tumor; N, normal adjacent kidney. samples, we were able to observe a highly significant increase of IGF2. Furthermore, our data also revealed a significant increase in the expression of IGF2 in tumor groups which maintained heterozygosity as well as imprinting. In each case, the overexpression of IGF2 in the malignant tissue was due to activation of all four promoters of IGF2, regardless of their LOI/MOI status. In our previous study (9), we demonstrated that, in every tumor that demonstrates LOI of IGF2, expression from promoter P1 as well as from the normally imprinted promoters P2-P4 was biallelic, but we did not examine promoter-specific gene expression in the NAT samples.
LOH at chromosome 11p15 in Wilms' tumor is a result of a loss of the maternal chromosome (39,44). Since H19 is paternally imprinted, LOH should result in elimination of expression of H19. Previous studies have shown little or very low expression of H19 in tumors with LOH, although these studies did not report the measurement of H19 in paired normal tissues (20,36). We observed reduced (71% of the NAT) expression of H19 in 14 tumor specimens with LOH (Table III), suggesting that while LOH may have been responsible for the reduction in the abundance of H19 RNA in the tumor, some expression of H19 from the paternal chromosome (loss of H19 imprinting) may also have occurred. Similar results have previously been reported (19). The association of LOI of IGF2 with a downregulation of H19 expression has been reported by others (19,20). However, our data showed that the abundance of IGF2 transcripts was not related to H19 expression in Wilms' tumor. While numerous studies have shown that H19 may control IGF2 imprinting (20,36,45), we could not confirm that H19 regulates the abundance of IGF2 mRNA in Wilms' tumor.
It is unclear what mechanism(s) is involved in enhancing the expression of IGF2 in Wilms' tumors since it appears that this activation is independent of LOH or LOI. Differential DNA methylation has been proposed as a candidate for epigenetic marking of the parental alleles. Recently, it has been shown that genes at the 11p15 locus were hypermethylated regardless of the specificity of the methylation pattern at the IGF2/H19 locus (36). We recently showed that treating cultured human and mouse astrocyte cells with DNA demethylating agents resulted in an increase in IGF2 expression by activating the normally imprinted maternal allele (30). In mutant mice that were deficient in DNA methyltransferase activity, the expression of IGF2 gene was undetectable, whereas the normally silent paternal allele of the H19 gene was activated, suggesting that a normal level of DNA methylation was required for controlling differential expression of the paternal and maternal alleles of imprinted genes (46). In agreement with these results, we found a significant increase in the level of DNA methyltransferase RNA in these Wilms' tumors comparing their normal adjacent tissues, suggesting a similar pattern of methylation in Wilm's tumor (data not shown).
A number of putative tumor suppresser genes, such as WT1 (47), p53 (48), igf2r (49), and GPC3 (50) can regulate IGF2 mRNA or IGF-II protein levels. WT1 interacts directly with the IGF2 promoter and represses gene expression during nephrogenesis (47). Since mutations in the WT1 gene are frequently found in Wilms' tumors (51), it has been suggested that the functional loss of WT1 transcriptional repressor activity may result in increased synthesis of IGF2 in the tumor (47). Thus, it is possible that WT1 may contribute to the unregulated growth of Wilms' tumor through transcriptional activation of the IGF2 gene.
A number of studies have focused on the expression of IGF2 and H19 in Wilms' tumor and have examined the roles of LOH and LOI in tumorigenesis (2,7,20,36). Our findings support the previous observations indicating that LOH, LOI, and al-tered expression of IGF2 and H19 occur in Wilms' tumors, but our data show that overexpression of IGF2 was a consistent common feature for Wilms' tumor regardless of heterozygosity or imprinting status. The elucidation of the mechanism activating IGF2 gene overexpression may help us to understand the general molecular basis for Wilms' tumor oncogenesis.