Soluble IGF 2 Receptor Rescues ApcMin / + Intestinal Adenoma Progression Induced by Igf 2 Loss of Imprinting

The potent growth-promoting activity of insulin-like growth factor-II (IGF-II) is highly regulated during development but frequently up-regulated in tumors. Increased expression of the normally monoallelic (paternally expressed) mouse (Igf2) and human (IGF2) genes modify progression of intestinal adenoma in the Apc mouse and correlate with a high relative risk of human colorectal cancer susceptibility, respectively. We examined the functional consequence of Igf2 allelic dosage (null, monoallelic, and biallelic) on intestinal adenoma development in the Apc by breeding with mice with either disruption of Igf2 paternal allele or H19 maternal allele and used these models to evaluate an IGF-II–specific therapeutic intervention. Increased allelic Igf2 expression led to elongation of intestinal crypts, increased adenoma growth independent of systemic growth, and increased adenoma nuclear B-catenin staining. By introducing a transgene expressing a soluble form of the full-length IGF-II/mannose 6-phosphate receptor (sIGF2R) in the intestine, which acts as a specific inhibitor of IGF-II ligand bioavailability (ligand trap), we show rescue of the Igf2-dependent intestinal and adenoma phenotype. This evidence shows the functional potency of allelic dosage of an epigenetically regulated gene in cancer and supports the application of an IGF-II ligand–specific therapeutic intervention in colorectal cancer. (Cancer Res 2006; 66(4): 1940-8)


Introduction
New molecular markers of colorectal cancer risk and treatment are being identified, and one such common marker is epigenetic modification and loss of imprinting (LOI) of insulin-like growth factor 2 (IGF2), leading to increased ligand supply (1)(2)(3)(4).LOI of IGF2 is a commonly observed epigenetic abnormality in a wide range of other solid tumors in both human and murine models and is implicated as a potent modifier of cell growth and survival (5)(6)(7)(8)(9)(10).Recent evidence suggests that f10% of the human population may have LOI of IGF2 in peripheral blood lymphocytes and normal colonic tissue, but in cases of colorectal cancer, the correlation with biallelic expression increases to 30%, with an odds ratio of >3.5 (11).There is little information concerning the intestinal, adenoma, or carcinoma phenotype that is generated by only a 2-fold increase in allelic dosage of IGF2.Two mechanisms seem to account for the LOI and are the basis of molecular tests: hypomethylation of a differentially methylated region (DMR) close to the IGF2 coding region and hypermethylation of a CTCF binding region and DMR upstream of the noncoding H19 promoter, resulting in disruption of an enhancer boundary (1,3).In these circumstances, alterations in IGF2 imprinting may simply reflect a bystander effect of global modification of DNA methylation that, in some circumstances, may be age related (12).However, direct functional information in the Apc Min/+ mouse, a model of human intestinal polyposis, has shown that IGF-II supply directly modifies intestinal growth and promotes adenoma progression when continuously expressed using a bovine keratin 10 promoter driven transgene (5), and that biallelic expression of Igf2 can increase adenoma burden (13).
IGF-II is a potent embryonic and tumor growth factor that signals via the IGF-I receptor (IGF1R) through the Ras/mitogenactivated protein kinase, phosphatidylinositol 3-kinase/Akt/FOXO, and S6K/mammalian target of rapamycin (mTOR) signaling pathways to modify cell proliferation, cell survival, gene expression, and cell growth (8,9,14).IGF-I, the related ligand, accounts for postnatal growth and persists throughout adult life, with levels in the high reference range being correlated with increased risk of prostate and breast cancer (15).IGF-II can activate signaling via the IGF1R, isoform A of the insulin receptor, and chimeric forms of the IR and IGF1R (16).Specific IGF1R kinase inhibitors, therapeutic antibodies to IGF1R and to both ligands, have been developed and seem to inhibit colorectal cancer cell growth (17,18).Unlike IGF1R and IGF-binding proteins (IGFBP), IGF2R is a type I membrane protein that specifically binds IGF-II with high affinity (K d = 10 À10 mol/L) but does not lead to signal transduction (19).Rather, IGF2R functions to sequester IGF-II for internalization and degradation in the prelysosomal compartment and reflects its other main function as a mannose 6-phosphate receptor (19,20).
Epigenetic modification and mutations of the IGF-II signaling system occur in human colorectal tumors (21).Supply of IGF-II is frequently up-regulated, and serial analysis of gene expression has shown IGF2 as a commonly overexpressed gene in human colorectal cancer cell lines and tumors (22), and inactivating mutations of IGF2R and loss of heterozygosity are often detected in hereditary nonpolyposis colorectal cancer-derived tumors (23).Moreover, the proposed tumor suppressor function of IGF2R may also implicate other ligands, such as latent transforming growth factor-h1 (TGF-h1), which is activated once bound to the mannose 6-phosphate binding sites of the receptor (24).IGF-II supply is also a potent modifier of colorectal cancer cell line growth (21,25,26) and seems to be expressed in early colorectal adenoma (27).
In the mouse, Igf2 and Igf2r are reciprocally imprinted and expressed from the paternal and maternal alleles, respectively (28)(29)(30).When detected with in situ hybridization, Igf2 mRNA is normally detected during post-implantation development, with expression restricted to the paternal inherited allele in later development and in the adult, except in the exchange tissues of the brain (30).Disruption of the paternal allele (Igf2 +m/Àp ) results in proportional dwarfism first detected between E9.5 to E11, when quantified by either cell number or weight, respectively (31,32).LOI of Igf2 following maternal allele deletion of H19 and a CTCF boundary element (DH19 Àm/+p ) results in proportionate overgrowth that can be completely rescued by combination with Igf2 +m/Àp , suggesting that loss of H19 generates no other Igf2independent effects (33).
Here, we quantify the specific functional consequences of the Igf2 gene dosage in combination with the Apc Min/+ model using Igf2 Àm/Àp (homozygous null), Igf2 +m/Àp (maternal allele expressed alone), Igf2 +m/+p (wild-type with paternal expressed allele intact), and DH19 Àm/+p (biallelic expression).We then tested whether the phenotypes observed are due to increased ligand supply by the introduction of a bovine keratin 10-driven transgene expressing a soluble full-length mouse Igf2r (K10DIgf2r), previously shown to act as a ligand trap and to limit IGF-II bioavailability (23,(34)(35)(36)(37).

Materials and Methods
Mice and genotyping.All experimental breeding was approved by local ethics committee and done under license from the U.K. Home Office.Igf2 +m/-p , Apc Min/+ , and DH19 Àm/+p breeding pairs were obtained from Christopher Graham, Agiris Efstratiadis, Andrew Silver, and Shirley Tilghman and generated as described (30,33,38).Mice were housed, fed, and maintained as described previously in non-specific pathogen-free conditions (5).All mouse lines tested positive with PCR for stool Helicobacter species.Breeding schedules were as described (Table 1).DNA extraction from ear clips from 10-day-old mice and PCRs were done as previously described (5,38), except for H19, which were genotyped using MutH19-For CTAGAGCTCGCTGATCAGCCT, MutH19-Rev GACAGTGG-GAGTGGCACCTT, WtH19-For CCATCTTCATGGCCAACTCT, and WtH19-Rev AATGGGGAAACAGAGTCACG (annealing at 62jC).
Intestinal morphometry, adenoma scoring, and immunostaining.All animals were dissected at postnatal day 120, 1 hour after injection of bromodeoxyuridine (BrdUrd, 100 Ag/g body weight).Small intestines and colons were opened, cleaned, and fixed in neutral buffered formalin as described (5).Whole mounts examined under a dissecting microscope were used to determine surface area and score adenoma number and size in the proximal, middle, and distal third of the small intestine, total colon, and cecum.H&E sections were blinded to genotype and used to quantify crypt depth (nuclei number), villus length (nuclei number), and pathologic grading.Colon epithelial tumors were confirmed by examination of sections.Counts of vertically aligned epithelial nuclei were obtained from 10 representative and intact crypts ( from the base of crypt to the cryptvillus junction) and associated villi ( from the crypt-villus junction to the tip of the villus) located in either the proximal, middle, and distal thirds of the small intestine, and 1 cm from the anorectal margin of the distal colon ( five male mice per genotype).Pathologic grading was done blinded to genotype by (M.P., an experienced human and rodent gastrointestinal pathologist) and scored using established criteria in the mouse (39).Intestines from at least five mice per genotype were immunostained as described (5), except antigen retrieval was done by boiling in 10 mmol/L citric acid (5-20 minutes).Primary antibodies were against IGF1Rh (3033, New England Biolabs, Beverly, MA), Ser 473 phospho-Akt (9277, New England Biolabs), Ser 256 phospho-FKHR (9461, New England Biolabs), E-cadherin (108, Santa Cruz Biotechnology, Santa Cruz, CA), Villin (7672, Santa Cruz Biotechnology), MCM2 (9839, Santa Cruz Biotechnology), Tyr 632 phospho-IRS1 (17196, Santa Cruz Biotechnology), h-catenin (610154, Transduction, Lexington, KY) using BEAT blockade (Zymed, South San Francisco, CA), Lysosyme (DAKO, Carpinteria, CA), Muc2 (15334, Santa Cruz Biotechnology), and anti-BrdUrd (Roche, Indianapolis, IN).The proportion of epithelial cell nuclei stained with MCM2 was quantified using color digital images of at least 10 orientated crypt and villi using Image (NIH) software.The results expressed as a ratio of the number of nuclei from the base of the crypt to the level, where intense staining stops, relative to the total crypt-villus epithelial nuclei height.
5V -FAM-TCCGCTCTGAGAGTCCTTTATACTCTGGCC-TAMRA-3V were used for real-time product detection.For the K10DIgf2r transgene, primers JHForward and Reverse JHK10 5V -TCCCTTCTCTCCTTCTTACTAGT-3V , with the 3V end corresponding to the inserted SpeI site in the deleted transmembrane region.JHTaq was also used to detect transgene expression, and RNA from an independent nontransgenic line was used as a negative control.Reactions were done for 40 cycles using Qiagen Quantitect reagents (204343) and following manufacturer's instructions (38,40).Reaction conditions were initial denaturing, 94jC; annealing, 56jC; and extension for all experiments, 76jC.Additional positive and negative controls were cDNA from Igf2 +m/+p and Igf2 Àm/Àp mouse embryonic fibroblasts (passage 1) and plasmids with Igf2r and K10DIgf2r cDNA.Target gene quantification was done relative to Gapdh controls (normalized to 100,000 copies) using calibration reactions.Statistical analysis.Comparisons between genotypes were displayed as box plots and used nonparametric statistics throughout (Mann Whitney, two tailed).Expected and observed breeding outcomes used v 2 with Yates correction.Calculations used Minitab release 14.

Results
Allelic dosage of Igf 2 modifies intestinal growth and the crypt-villus axis.To quantify allelic dosage of Igf2 on the Apc Min/+ phenotype, we first did coisogenic (129/B6) and isogenic crosses (B6/B6) of inbred mouse lines to control for straindependent modifiers of Apc Min/+ (41).Igf2 +m/Àp (null), Igf2 +m/+p (monoallelic), and DH19 Àm/+p (biallelic) were crossed with inbred Apc Min/+ (Table 1).Increased allelic expression of Igf2 resulted in increased intestinal growth by 120 days, as judged by overall surface area and crypt cell number (Fig. 1A-B; Supplementary Fig. S1; Supplementary Table S1).The effect of allelic dose did not seem uniform along the length of the small intestine and seemed more prominent in the proximal small intestine (Fig. 1A).Elongation of the crypt and villus proliferative zone was quantified using staining with an anti-MCM2 antibody that labels cells undergoing DNA replication (Fig. 1C-D, MCM2).Significant expansion of the proliferative zone was detected with increasing Igf2 allelic dosage and was confirmed by anti-BrdUrd staining (data not shown).Differentiation of the small intestine was assessed using antibodies to E-cadherin, villin, MUC2, and lysosyme and only showed subtle attenuation of villin staining in DH19 Àm/+p , which seemed independent of altered intracellular distribution of h-catenin (Fig. 1D; Supplementary Fig. S2).
Igf 2 expression modifies intestinal signaling.The distribution of IGF1R seemed independent of Igf2 dosage (Fig. 1D).Specific ligand activation of the IGF1R was probed using a Tyr 632 phospho-insulin receptor substrate-1 (phospho-IRS1) antibody, which detected signal in smooth muscle, intervillus stroma, and in a gradient extending from epithelial cells of the crypt to the villi (Fig. 1D).Little cytoplasmic labelling for phospho-IRS1 is visible in epithelial cells in sections from Igf2 +m/Àp , except in the smooth muscle and stromal layers, presumably due to circulating IGF-I, and suggests that epithelial cell signal may be Igf2 dependent (Fig. 1D).Similar distribution of cytoplasmic staining was detected with antibodies to phospho-Ser 473 of Akt and phospho-Ser 256 of FKHR (not shown).Quantitative RT-PCR using Taqman probes from control 120-day-old, 129, B6, and 129/B6 (data not shown) mice showed that Igf2 mRNA is expressed in adult small intestine and colon at lower levels than in the heart, kidney, forestomach, and Apc Min/+ adenoma, the latter confirming results from in situ hybridization (ref.5; Fig. 2A).This evidence is supported by others (13) and refutes previous assumptions that Igf2 expression is switched off in all adult mouse tissues.
Moreover, Igf2 mRNA is detectable in heart tissue at high levels, even in Igf2 +m/Àp , indicating either loss of imprinting of Igf2 in the heart, or high basal maternal allele expression (Fig. 2A).Igf2 mRNA expression is globally increased in DH19 Àm/+p but is not a direct 2-fold effect in most tissues at 120 days (Fig. 2A).There seems to be differences following comparison of Igf2 expression between mouse strains (129 and B6) without disrupted Igf2 supply.The magnitude of the increased expression in wild-type 129 compared with B6 is similar to comparison of wild-type B6 to DH19 Àm/+p (B6/B6).Moreover, endogenous Igf2r expression on the 129 strain also seems higher than 129/B6 or B6/B6 expression (Fig. 2B), irrespective of expression of the K10DIgf2r/+ transgene (below, Fig. 2C-D).Overall, these observations suggest that continuous supply of ligand modifies crypt cell progenitor maturation independent of Apc Min/+ .
Allelic dosage of Igf 2 modifies Apc Min/+ intestinal adenoma progression and dysplasia.Increased allelic Igf2 dosage increased the total number of intestinal adenoma when genetically combined with Apc Min/+ detected using dissection microscope examination of fixed intestinal whole mounts (mean F SD): 10.4 F 2.4 Igf2 +m/Àp , Apc Min/+ (null) versus 26.8 F 7.4 Igf2 +m/+p , Apc Min/+ (monoallelic), P = 0.0002 (129/B6) and 40.4 F 22 Igf2 +m/+p , Apc Min/+ (monoallelic) versus 58.8 F 15.6 DH19 Àm/+p , Apc Min/+ (biallelic), P = 0.0097 (B6/B6).However, when adenoma were normalized relative to growth of the small intestine and colon surface area, a significant decrease in total adenoma number was only detected in the small intestine of Igf2 +m/Àp , Apc Min/+ (129/B6, P = 0.0002), confirming our previous data obtained with lines contaminated with unknown Apc Min/+ modifier alleles (ref.5; Fig. 3A, normalized to intestinal growth).In comparison, a significant increase in normalized adenoma burden was only detected in DH19 Àm/+p , Apc Min/+ (B6/B6) colon but not small intestine (Fig. 3B, normalized to colon growth).Inbreeding of Igf2 +m/Àp against the parent C57Bl6 line was done for five generations, before we attempted to generate complete Igf2 Àm/Àp , Apc Min/+ (homozygous Igf2 null mice) but failed to observe correct Mendelian segregation at birth (Supplementary Table S2).One Igf2 Àm/Àp , Apc Min/+ (B6/B6) mouse survived into adulthood and had an adenoma burden equivalent to Igf2 +m/Àp , Apc Min/+ controls [129/B6, 0.58 versus 0.54 F 0.12 (mean F SD) adenoma cm À2 ].Distribution of small intestine adenoma size was similar between genotypes, except in DH19 Àm/+p , Apc Min/+ , where a disproportionate number seemed larger compared with wildtype Apc Min/+ (Fig. 3C).Examination of adenoma sections, scored independently of genotype, revealed an increase in the proportion of adenoma containing at least a single focus of high-grade versus low-grade dysplasia in a mid-adenoma section of DH19 Àm/+p , Apc Min/+ (68 of 85) versus Apc Min/+ controls (12 of 22), confirming an effect of Igf2 LOI on intestinal adenoma progression (Supplementary Fig. S3).Features of carcinoma in situ and stromal reaction were also detected in colonic adenoma from DH19 Àm/+p , Apc Min/+ , but no metastases were observed (Supplementary Fig. S3; ref. 39).Overall, this suggests that Igf2 has predominant effects on adenoma progression when compared at a single time point (120 days), which seem to become more marked with increasing time. 3Here, we cannot exclude modifier effects on adenoma initiation during crypt fission events in early intestinal development (42).A, ratio of villus to crypt epithelial nuclei number (height) in the proximal, middle, and distal thirds of the small intestine.Box, interquartile range; vertical line, minimum to maximum value excluding outliers; horizontal line, median; dot, mean (n = 50 crypts and villi per genotype).B, height of distal colonic crypts expressed as nuclei number (n = 50 crypts per genotype).C, ratio of the number of MCM2-positive nuclei (height) from the crypt base up to where signal stops in the villi, relative to the total number of crypt and villi epithelial cell nuclei (proximal small intestine).D, representative small intestinal sections from 120-day-old mice stained with antibodies to IGF1-R, phospho-IRS1, E-cadherin, h-catenin, and MCM2 (brown/black staining).Note the elongation of biallelic Igf2 small intestinal crypts and level of MCM2 staining (white horizontal lines), differential labelling of epithelial cells with phospho-IRS1 in Igf2 +m/Àp (note signal in smooth muscle but absence from epithelial cells) and K10DIgf2r/ + transgene intestine (see text), and no alteration in h-catenin distribution, with nuclear staining only in Paneth cells at the base of the crypts (black arrows ).Bar, 100 Am.Genotypes for null Igf2 expression Igf2 +m/Àp (À), monoallelic expression Igf2 +m/+p (+), and biallelic expression DH19 Àm/+p (++), with (+) and without (À) expression of a sIGF2R transgene K10DIgf2r/ +.NS, not significant.
Expression of a soluble IGF2R transgene rescues Igf2 phenotypic effects in the intestine and in Apc Min/+ adenoma.To confirm that effects of Igf2 LOI were due to increased ligand supply, we evaluated a soluble form of IGF2R as a ligand-specific trap (36).In two independent bovine keratin 10 promoterdriven transgenic lines, sIGF2R reduced Igf2-dependent growth of intestinal tissues that expressed the transgene (129JS2, K10DIgf2r/+; refs.34,38).Heterozygotes of the highest expressing transgenic line (Kipps), expressed the transgene in the stomach, small intestine, colon, and adenoma, as detected by a transgene specific Taqman quantitative RT-PCR (Fig. 2C-D).In all cases, the transgene was transmitted from a 129 background in coisogenic crosses with B6 (129/B6).No circulating serum protein was detected with an IGF2R ELISA (data not shown), and endogenous Igf2r expression was not significantly affected by transgene expression (Fig. 2B).Crossing K10DIgf2r/+ with Apc Min/+ (129/B6) resulted in a significant reduction of colonic crypt depth, MCM2 and phospho-IRS1 staining, and adenoma number to levels comparable with Igf2 +m/Àp (Fig. 1B-D and Fig. 3A-C; Supplementary Fig. S1 and Table S1).In particular, significantly less adenoma formed in the distal small intestine compared with Apc Min/+ (Igf2, wild type) littermate controls (P = 0.0003; Fig. 3A).A second series of genetic crosses were used to evaluate Igf2-dependent and Igf2-independent effects of sIGF2R supply (Table 1).Combination of Igf2 +m/Àp , Apc Min/+ (B6/B6) with the transgene K10DIgf2r/+ (129/129, n = 6) resulted in no further reduction of adenoma number in the small intestine compared with Igf2 +m/Àp , Apc Min/+ controls (n = 3) and suggests that there is no significant Igf2-independent activity of sIGF2R (Fig. 4A).Combination of DH19 Àm/+p , Apc Min/+ with the transgene also led to suppression of small intestinal MCM2 labeling (Fig. 1C), small intestine (Fig. 4B) and colonic (Fig. 4C) adenoma number, to levels equivalent to the K10DIgf2r/+, Apc Min/+ controls despite limited mouse numbers, and confirmed the Igf2-dependent activity of K10DIgf2r/+ as a specific inhibitor to the consequences of increased ligand supply generated by Igf2 LOI.Moreover, the effects of transgene expression are independent of overall growth effects in the mouse, as determined by intestinal surface area and whole body weight (Supplementary Table S1).
Biallelic Igf2 expression is associated with increased nuclear B-catenin staining.Unlike wild-type Apc Min/+ small intestine and intestinal adenoma, where nuclear h-catenin was only detected in Paneth cells of the crypt (Fig. 1D), frequent presence of large regions of cytoplasmic and nuclear h-catenin were detected in sections through DH19 Àm/+p , Apc Min/+ adenoma all sizes (23 of 28 adenoma from mid-adenoma sections has at least one obvious focus, n = 5 mice) compared with littermate Apc Min/+ controls (3 of 12, n = 5 mice; Fig. 4D).Staining of sections and Western blotting with E-cadherin antibodies failed to show significant down-regulation of E-cadherin in DH19 Àm/+p , Apc Min/+ and suggests that the appearance of nuclear h-catenin is independent of E-cadherin activity (data not shown).In all adenoma, more extensive phospho-IRS1 labeling was also evident in areas, where there was nuclear h-catenin in DH19 Àm/+p , Apc Min/+ .Phospho-IRS1 staining seemed reduced in parallel processed specimens and was associated with a reduced extent of nuclear h-catenin staining in K10DIgf2r/+, DH19 Àm/+p , Apc Min/+ adenoma, although persistence of staining was observed in adenoma that did only form in the proximal small intestine, presumably because of the differential extent of transgene expression in this region (Fig. 4D).

Discussion
A number of genetic mouse models with tumor susceptibility have been used to show the loss of Igf2 expression and its Figure 3. Analysis of Igf2 allelic supply on Apc Min/+ intestinal adenoma number and size.A, number of adenoma (corrected for surface area, cm À2 ) for the proximal (P), middle (M ), and distal (D ) small intestine for the different genotypes of both 129/B6 coisogenic and B6/B6 isogenic crosses (see Table 1).Genotypes for null Igf2 expression Igf2 +m/Àp (À), monoallelic expression Igf2 +m/+p (+), and biallelic expression DH19 Àm/+p (++), with (+) and without (À) expression of a sIGF2R transgene (K10DIgf2r/ +).B, number of adenoma (corrected for surface area, cm À2 ) for the colon as for (B ).C, adenoma size (mm) percentage distributions for all mice counted in (B) separated for B6/129 and B6/B6 crosses (see inset ).Significant increase in adenoma size is detected in DH19 Àm/+p (++).
consequence to the progression of the tumor phenotype (5-8, 43, 44).In these circumstances, tumor size and frequency decrease and frequently reactivation of the normally silenced maternal allele of Igf2 can occur (LOI).The mechanism of reactivation is presumed to be selection of epigenetically modified alleles that modify tumor cell survival, suggesting that activation of Igf2 expression is essential for tumor progression in the mouse.Overexpression of either Igf2 or Igf1r using transgenes results in local and systemic tumor formation when expressed at high levels alone and as a potent tumor promoter when combined with tumor-susceptible genetic models (5,45,46).However, the nonphysiologic expression in these circumstances limits the interpretation of these experiments.
Here, we show that biallelic expression of Igf2 following germ line disruption of an imprinting control region upstream of H19 results in marked intestinal phenotypic effects that seem independent of the generalized overgrowth phenotype (f130% of wildtype body weight).We then evaluated the consequences of this biallelic expression of Igf2 on the development of the Apc Min/+ intestinal adenoma phenotype and show that growth and differentiation effects seem to develop with increasing Igf2 expression.In particular, we found that comparison of the expression of a single paternal allele of Igf2 had significant effects on adenoma number as shown previously, whereas expression of both Igf2 alleles resulted in no significant alteration in adenoma number in the small intestine, but an alteration in adenoma differentiation with associated high-grade dysplasia, and an increase in colonic adenoma number (5).Similar differential dose response effects of the IGF1R receptor in the RIPTag model have been described, with initial expression modifying tumor proliferation and apoptosis and higher expression associated with invasion and tumor metastasis (9).Results of a similar nature to those Figure 4. Analysis of Igf2-dependent and Igf2 -independent activity of sIGF2R on Apc Min/+ intestinal adenoma.Total number of adenoma (corrected for surface area, cm À2 ) for littermates at postnatal day 120 in the small intestine following combination of genotypes Igf2 +m/Àp , K10DIgf2r/+, Apc Min/+ (A) and DH19 Àm/+p , K10DIgf2r/+, Apc Min/+ (B, small intestine; C, colon).D, small intestinal adenoma stained with h-catenin, phospho-IRS1, and MCM2.Note the nuclear h-catenin in DH19 Àm/+p , Apc Min/+ (++) and significant reduction of both epithelial nuclear h-catenin and phospho-IRS1 by expression of sIGF2R.Bar, 100 Am.

Cancer Research
Cancer Res 2006; 66: (4).February 15, 2006 reported here have been obtained recently, and both are relevant to the observation that humans that develop colorectal cancer seem to frequently have systemic loss of imprinting leading to biallelic Igf2 expression (2,13).The data from Sakatani et al. (13) differ from those reported here, as we did not detect a significant increase in small intestinal adenoma number (when normalized for intestinal surface area) as a result of biallelic Igf2 expression presumably because of the effects of strain-dependent Apc Min/+ modifiers (13,41).In this regard, although mRNA expression may not correlate with ligand supply, we note that global increases in relative Igf2 expression were observed in wild-type 129 compared with B6 strain, suggesting that crosses containing 129 may have increased relative Igf2 supply.In the context of DH19 Àm/+p , Apc Min/ + , although Igf2 expression increases compared with Apc Min/+ B6 controls, the eventual level of expression is only equivalent to that detected in the wild-type 129 line and may contribute to the differences in adenoma number between strains.However, the relative strain expression of Igf2r matches that of Igf2, and total adenoma number are lower in 129/B6 crosses than B6/B6, suggesting a more complex situation (see Figs. 2 and 3).
We directly tested the dose-dependent effect of Igf2 allelic supply by expressing a soluble IGF2R expressing transgene, previously shown to inhibit overgrowth effects in the intestine generated as a consequence of IGF-II overexpression (34,42,47).Our results show rescue of the bialleic Igf2 intestinal phenotype when combined with the Apc Min/+ mouse, an effect which we also determined not to be independent of Igf2 in parallel crosses with Igf2 +m/Àp .This evidence suggests that the phenotypic effects observed are due to increased supply of ligand, as soluble IGF2R acts to limit IGF-II bioavailability.Previous studies in cell lines and tumors derived from them support the tumor suppressor function of membranebound and soluble IGF2R (35)(36)(37)48).In some circumstances, excess supply of IGF2R may lead to increased conversion of latent TGF-h1 to active TGF-h1, but the results of genetic crosses fail to show any significant Igf2-independent activity in the small intestine and colon.
We propose that the reason that we observed an increase in adenoma number in the colon and adenoma growth in the small intestine was as a direct result of Igf2 acting as a progression factor subsequent to the loss of initiating Apc tumor suppressor function (5,8).The mechanism may be due to a combination of modification of cell survival, proliferation, growth, and differentiation.In addition, the mechanisms that underlie the modifier effects of Igf2 allelic dosage on the intestinal adenoma phenotype implicate potential direct interactions between the Igf2 and Wnt gene signaling systems.To date, evidence points towards modification of the bioavailability of h-catenin and phosphorylated substrates of Akt.One obvious target may be glycogen synthase kinase 3 (GSK3h), an Akt substrate and key modifier of Wnt signaling via phosphorylation of h-catenin.However, structural constraints exerted by Apc/Axin complexes negate functional interactions with Akt (49).Akt activation as a consequence of overexpression IGF1R mediated signaling seems to down-regulate the supply of E-cadherin, which also regulates cellular h-catenin, although we found no evidence to support this effect as a mechanism of nuclear h-catenin accumulation in adenoma (9,50).Previous work suggests that IGF-II expression modifies colorectal cell differentiation and the nuclear localization of h-catenin (25,26).In addition, expression of activated Akt can increase expression of luciferase reporter gene for h-catenin/ Tcf (51).Recently, direct binding of h-catenin to FOXO transcription factors, phosphorylated and depleted from nuclei by IGF1R-and IR-mediated signaling, accounted for the increased activation of FOXO target genes (e.g., p27 Kip1 ) by increased nuclear h-catenin (52).Intriguingly, this evidence suggests that activated h-catenin would generate cell cycle arrest and apoptosis via direct activation of FOXO target genes, an effect that is inhibited and tumor promoting in the presence of IGF-II and Akt activation.If true, the marked effect of reduction in IGF-II supply in the context of activated h-catenin, as shown here in the context of Apc Min/+ , may be accounted for by h-catenin specifically sensitizing these cells to cell cycle arrest and cell death (52).Disruption of the FOXO target gene p27 Kip1 increases progression of Apc Min/+ , but not Smad3, intestinal adenoma, and is consistent with this hypothesis (53).Moreover, components of the IGF signaling pathway, such as mTOR and FOXO, may actually increase the rate of Apc loss of heterozygosity and increase adenoma-initiating events (54).
IGF-II ligand availability also seems to be modified by the h-catenin/Tcf-mediated overexpression of matrix metalloproteinase-7.Release of enzyme into the extracellular space then results in cleavage of IGFBP and increased local bioavailability of IGF-II ligand (55).The inhibitory consequences of Mmp7 disruption on Apc Min/+ adenoma is also consistent with this hypothesis but does not explain the apparent increase in Igfbp5 expression in adenoma detected using expression arrays (56,57).More importantly, the 2-fold effect of Igf2 gene dosage would be normally omitted from array gene expression analysis and underscores the bias against potentially haploinsufficient gene phenotypes.
The potency of allelic dosage of Igf2 is evident from the alteration in phenotype of the intestine crypt in the mouse, and this evidence supports the significant correlation between LOI of Igf2 and the risk of developing sporadic colorectal cancer.Moreover, we have applied this validated genetic model to test potential therapeutic molecules that target this system.We tested the effects of a specific IGF-II intervention based on a soluble form of an endogenous receptor that normally functions to sequester IGF-II and is classed as a tumor suppressor.Similar use of ligand trap to other growth factors have been investigated in intestinal cancer (e.g., anti-vascular endothelial growth factor antibodies, soluble IGF1R, and soluble Noggin; refs.[58][59][60].A number of alternative approaches have also been developed to directly inhibit IGF ligand-mediated signaling, including IGF1R kinase inhibitors and antibodies directed at ligands and IGF1R (18,58).We now provide genetic evidence that the consequences of excess supply of IGF-II ligand can be rescued by a specific high-affinity soluble IGF2 receptor, and this supports the rationale application of this novel IGF-II-targeted therapy for cancer.

Figure 1 .
Figure1.Morphometric analysis of Igf2 allelic expression and intestinal growth.A, ratio of villus to crypt epithelial nuclei number (height) in the proximal, middle, and distal thirds of the small intestine.Box, interquartile range; vertical line, minimum to maximum value excluding outliers; horizontal line, median; dot, mean (n = 50 crypts and villi per genotype).B, height of distal colonic crypts expressed as nuclei number (n = 50 crypts per genotype).C, ratio of the number of MCM2-positive nuclei (height) from the crypt base up to where signal stops in the villi, relative to the total number of crypt and villi epithelial cell nuclei (proximal small intestine).D, representative small intestinal sections from 120-day-old mice stained with antibodies to IGF1-R, phospho-IRS1, E-cadherin, h-catenin, and MCM2 (brown/black staining).Note the elongation of biallelic Igf2 small intestinal crypts and level of MCM2 staining (white horizontal lines), differential labelling of epithelial cells with phospho-IRS1 in Igf2 +m/Àp (note signal in smooth muscle but absence from epithelial cells) and K10DIgf2r/ + transgene intestine (see text), and no alteration in h-catenin distribution, with nuclear staining only in Paneth cells at the base of the crypts (black arrows ).Bar, 100 Am.Genotypes for null Igf2 expression Igf2 +m/Àp (À), monoallelic expression Igf2 +m/+p (+), and biallelic expression DH19 Àm/+p (++), with (+) and without (À) expression of a sIGF2R transgene K10DIgf2r/ +.NS, not significant.