The capacity to model cancer within the mouse has advanced significantly in recent years. Perhaps the most notable technical gains have been in the development of techniques that allow the temporal and spatial control of gene expression, so that it is now possible to regulate target genes in the tissue of choice and at a given time [Maddison and Clarke (2005) J. Pathol. 205, 181–193; Shaw and Clarke (2007) DNA Repair 6, 1403–1412; Marsh and Clarke (2007) Expert Rev. Anticancer Ther. 7, 519–531]. We have used these approaches to study tumorigenesis in the murine intestine. Loss of function of the tumour-suppressor gene Apc (adenomatous polyposis coli) has been associated with the development of both human and murine neoplasia, principally those of the intestinal epithelium. However, as Apc has been implicated in multiple cellular functions, the precise mechanisms underlying these associations remain somewhat unclear. I review here the use of an inducible strategy to co-ordinately delete genes from the adult murine epithelium. This approach has allowed a characterization of the direct consequences of inactivation of gene function. For Apc, these include failure in the differentiation programme, failure to migrate, aberrant proliferation and the aberrant induction of apoptosis. Transcriptome analysis of this model has also identified potential new targets for therapeutic intervention, such as Sparc (secreted protein acidic and rich in cysteine), deficiency of which, we have now shown, suppresses adenoma formation. Finally, we have been able to address how other genes modulate the consequences of Apc loss. Thus we show that there is little effect following loss of cyclin D1, Tcf-1 and p53, but that there are marked differences following loss of either c-Myc or Mbd2. The models therefore allow us to define the earliest events associated with carcinogenesis in the intestine.
The crypt–villus axis of the murine small intestine is becoming increasingly well defined both in terms of its genetics and cellular physiology. The structure is composed of enterocytes, paneth cells, goblet cells and enteroendocrine cells. The paneth cells are localized to the base of the crypt, just below a zone recognized to contain the intestinal stem cell. Above this area, there is an area of high cell turnover, recognized as the transit-amplifying zone. Cells are born in the base of the crypt, in a process assumed to reflect asynchronous division of the stem cell. Most of the cells then migrate along the length of the crypt–villus, before being shed into the intestinal lumen. The exception to this is the Paneth cell, which migrates down to the base of the crypt.
Over recent years, there has been a renewed surge of interest in both crypt physiology and genotype–phenotype relationships that govern the crypt–villus architecture. This has been driven both by the availability of new in vivo tools and the realization that the crypt–villus biology axis is determined by a key set of molecular pathways, including the Wnt pathway . The net result of this is that the small intestine has emerged as an ideal experimental system for investigating both the genetic and epigenetic control over normal and diseased physiology.
The most notable technical gains in modelling cancer within the mouse have perhaps been in the development of techniques that allow the temporal and spatial control of gene expression, so that it is now possible to regulate target genes in the tissue of choice and at a given time [2–4]. We have used these approaches to study tumorigenesis in the murine intestine.
Apc (adenomatous polyposis coli)
The human syndrome FAP (familial adenomatous polyposis) is characterized by the spontaneous development of numerous polyps within the colon, of which some progress to carcinoma. Mutation of the Apc tumour-suppressor gene was found to underlie this syndrome, and has subsequently also been found to be mutated or epigenetically repressed in most of the sporadic colorectal tumours . These observations established the Wnt pathway as a critical mediator of tumorigenesis within the intestine, as the best-characterized role for Apc is in repressing Wnt signalling by mediating the destruction of β-catenin. Such a role provides a ready mechanism for tumour suppression as, in the absence of Apc, Wnt-mediated transcription of a range of proliferation-associated genes is thought to become de-regulated .
The first mouse model of human FAP was developed in William Dove's laboratory . This model, termed the ApcMin mouse, emerged out of a chemical mutagenesis screen and is characterized by germline mutation in Apc. Mice heterozygous for this mutation develop multiple polyps within both the small and large intestine. These polyps are characterized by a second hit at the Apc locus, leading to loss of function of Apc. This model has proven to be a very useful model of the human disease and has been used in a wide range of experiments to test both genetic predisposition and therapeutic regimes. For example, we have used this approach to probe the role of PPAR-δ (peroxisome-proliferator-activated receptor-δ)  and aspirin  in adenoma formation. Although a potent model of human disease, the ApcMin mouse is somewhat limited in its ability to model the very first molecular and physiological events following inactivation of Apc. To address this limitation, we and others have developed conditional approaches that allow inducible gene regulation within the adult small intestine. Our strategy was to use the Cre–Lox system, in conjunction with a LoxP-flanked Apc allele . The transgene used to deliver Cre expression was placed under the Cyp1A1 (cytochrome P450 subfamily A1) promoter, the expression of which is inducible following exposure to a range of agents, including β-naphthoflavone . Following induction by intraperitoneal injection of β-naphthoflavone, Cre is induced within the base of the crypt and drives recombination of any LoxP-flanked alleles with near 100% efficiency. Induction of the transgene also occurs within a number of other cell types, most notably within hepatocytes and the kidney . In the intestine, those cells that have undergone Cre-mediated recombination include the stem cell population, such that the entire crypt–villus architecture becomes populated by cells bearing the desired recombined alleles. Using a surrogate marker of recombination, such as the Rosa26 LacZ reporter, we have shown that, as long as the introduced mutation is not deleterious, these recombined crypts remain stable within the intestine for at least a year following induction .
By breeding mice that carried both the Cre transgene and two LoxP-flanked Apc alleles, we were able to assess the immediate effects of loss of Apc function within the crypt–villus . These include a rapid change in histological appearance of enterocytes with the crypt architecture grossly disrupted, an immediate entry of cells into S-phase as noted by BrdU (bromodeoxyuridine) incorporation, and loss of most of the differentiated cell types including enteroendocrine cells and goblet cells. The co-ordinate entry of nearly all Apc-deficient cells into S-phase provides one ready mechanism by which loss of a single gene can impose a pro-tumorigenic phenotype. However, this apparent deregulation of cell proliferation was not matched by an increase in mitotic figures, and indeed we observe both a marked induction of cell death and a loss of control over nuclear volume within at least a subset of Apc-deficient cells . These observations are consistent with the hypothesis that Apc is required for the normal execution of division, possibly through a role in spindle formation and function. Thus activation of the Wnt pathway following loss of Apc appears to immediately drive cells into proliferation, but deficiency of Apc also compromises cell division. This scenario provides a ready mechanism for the selection of subsequent mutations that will allow Apc-deficient cells to proliferate, and indeed the levels of cell death we observe immediately following removal of Apc are noticeably lower than those scored in fully developed adenomas.
Our immediate analysis of Apc deficiency also revealed other pro-tumorigenic mechanisms. Most notably, we observed a complete block on cell migration along the length of the crypt–villus . Enterocytes are normally born within the base of the crypt and then rapidly migrate along the length of the crypt–villus axis before being shed into the gut lumen. Clearly, if a cell (other than a stem cell) is destined to become the founder of an adenoma, it must somehow avoid this rapid ‘escalator’. Our results provide an immediate mechanism for this, such that Apc-deficient cells remain static within the distorted crypt, so avoiding the normal shedding mechanism.
Analysis of downstream components and potential modifiers of the Wnt pathway
The approach we have used delivers high efficiency and co-ordinated loss of gene function [10,12,13] and therefore offers an ideal platform for transcriptome analysis. We have therefore performed affymetrix array-based studies [10,15] to identify the major consequences of Wnt activation in the intestine. For those genes that we observe to be direct Wnt targets, we have begun to address their role in adenoma formation and indeed to validate them as potential novel targets for therapeutic intervention. For example, we identified extracellular matrix protein Sparc (secreted protein acidic and rich in cysteine) to be up-regulated following deletion of Apc. We therefore crossed ApcMin mice on to a Sparc-deficient background and observed an attenuation of adenoma formation .
Remarkably, our transcriptome analysis has shown that many purported Wnt target genes do not become deregulated immediately following loss of Apc in the intestine. Indeed, the dominant pattern that we observe is that those genes identified in the literature as Wnt targets tend to become deregulated at the stage of adenoma formation (K.R. Reed, O.S. Sansom, J. Wilkins and A.R. Clarke, unpublished work). We interpret this to show either that the intestine deregulates a specific subset of Wnt targets or that many genes considered direct targets actually become deregulated as a secondary consequence. We have advanced this argument most fully for cyclin D1, where we have directly tested the cyclin D1 dependency of the immediate Apc phenotype. These studies have suggested that cyclin D1 is important for the delayed consequences of Apc deficiency and so is consistent with the notion that cyclin D1 is not an immediate target of the Wnt pathway, at least in the intestinal crypt .
We have performed similar genetic experiments in relation to three other genes that are predicted to strongly influence the Wnt pathway. p53 is a potent tumour suppressor that has been associated with the later stages of human colorectal disease, and it has also been proposed to act in a regulatory Wnt capacity. Its association with adenoma formation is, however, somewhat less clear, with p53 deficiency reported to have either no effect or relatively little effect in the ApcMin background [18,19]. Remarkably, we find that p53 deficiency has very little effect on the immediate consequences of Apc loss, which supports the notion that p53 is largely redundant at the earliest stages of adenoma formation (K.R. Reed, O.S. Sansom, J. Wilkins and A.R. Clarke, unpublished work). We see a similar lack of effect for deficiency of Tcf-1 in the intestine. This is in marked contrast with the synergy we have previously observed in the mammary gland between loss of Apc and Tcf-1, which converts a metaplastic phenotype into one of rapid adenocarcinoma formation .
In contrast with the above observations, we do observe a major effect following loss of the Wnt target c-Myc. This area, however, remains somewhat controversial. We have found that a single deficiency of c-Myc compromises enterocyte physiology, such that the cells within the crypt become smaller and less metabolically active . Using the Cyp1A1–Cre strategy, conditional deletion of c-Myc eventually leads to the loss of c-Myc-deficient cells, which become replaced by wild-type crypts after 14 days. This repopulation of the intestine seems to occur by competitive expansion of the healthier, c-Myc proficient crypts. In contrast with our results, Andreas Trumpp's laboratory  has reported that c-Myc deficiency is well tolerated in the adult intestine, although they do see clear loss of crypts in younger animals. The reason for this difference remains unclear, but may well relate to the experimental systems being used. For example, through differences in the pattern of cre-mediated excision.
As we have shown that c-Myc deficiency is tolerated for up to 8 days, this has afforded a time window in which we have been able to test the c-Myc dependency of the immediate phenotypes of Apc loss. Remarkably, this has revealed nearly all the aspects of the phenotype to be c-Myc-dependent, identifying this as a key mediator of Wnt tumorigenesis in the intestine .
Epigenetic modulation of tumorigenesis in the intestine
It is increasingly clear that epigenetic processes also influence tumorigenesis in the intestine. In the mouse, this was first clearly demonstrated by the attenuation of adenoma formation within the ApcMin mouse on a DNMT1 (DNA methyltransferase 1) mutant background, in which DNA methylation is perturbed . We have again used the ApcMin mouse and conditionally mutant Apc mice to further probe this epigenetic dependency. In pursuing this goal, we have focused on the mbd (methyl binding domain) proteins and a related protein, Kaiso. This family of proteins serves to interpret the methylation signals in DNA and to recruit chromatin remodelling complexes . Their predominant role is therefore to repress gene expression. In terms of cancer predisposition, the hypothesis being pursued is that the absence of these proteins will reduce the likelihood of epigenetic inactivation of key tumour-suppressor genes. Consistent with this hypothesis, we have shown that deficiency of one family member, Mbd2, specifically and dramatically reduces adenoma burden in the ApcMin mouse [25,26]. More recently, we have shown that Mbd2 plays a normal role in controlling correct spatial gene expression in the intestine . We have also probed the mechanism underlying adenoma repression using the conditional Apc mutant background, which has revealed a role for Mbd2 in mediating feedback repression of the Wnt pathway. Very similar data have been obtained for the related protein Kaiso , identifying both these genes as excellent therapeutic targets, given that constitutive deficiency of each leads to a very mild or absent phenotype.
In conclusion, we have used both conditional and constitutive models of Apc deficiency to probe the basic physiology of both the normal and diseased intestine, with a specific focus on the control over the earliest stages of adenoma formation. These models are proving particularly potent in defining the mechanisms that underlie adenoma formation and in identifying those genes that either augment or repress this process. The ultimate goal of this strategy is to identify and validate new therapeutic targets for disease. Towards this goal, we have already identified several promising targets, including both mediators of the epigenetic programme, such as Kaiso and Mbd2, and direct mediators of the Wnt pathway, such as c-Myc.
Cancer: A Focus Topic at Life Sciences 2007, held at SECC Glasgow, U.K., 9–12 July 2007. Edited by D. Gillespie (Beatson Institute, Glasgow, U.K.) and H.M. Wallace (Aberdeen, U.K.).