- Top of page
- Complex microenvironment of the intestinal crypt
- Mouse models
Sequences of molecular events that initiate and advance the progression of human colorectal cancer (CRC) are becoming clearer. Accepting that these events, once they are in place, accumulate over time, rapid disease progression might be expected. Yet CRC usually develops slowly over decades. Emerging insights suggest that the tumor cell microenvironment encompassing fibroblasts and endothelial and immune cells dictate when, whether, and how malignancies progress. Signaling pathways that affect the microenvironment and the inflammatory response seem to play a central role in CRC. Indeed, some of these pathways directly regulate the stem/progenitor cell niche at the base of the crypt; it now appears that the survival and growth of neoplastic cells often relies upon their subverted engagement of these pathways. Spurned on by the use of gene manipulation technologies in the mouse, dissecting and recapitulating these complex molecular interactions between the tumor and its microenvironment in the gastrointestinal (GI) tract is a reality. In parallel, our ability to isolate and grow GI stem cells in vitro enables us, for the first time, to complement reductionist in vitro findings with complex in vivo observations. Surprisingly, data suggest that the large number of signaling pathways underpinning the reciprocal interaction between the neoplastic epithelium and its microenvironment converge on a small number of common transcription factors. Here, we review the separate and interactive roles of NFκB, Stat3, and Myb, transcription factors commonly overexpressed or excessively activated in CRC. They confer molecular links between inflammation, stroma, the stem cell niche, and neoplastic cell growth.
- Top of page
- Complex microenvironment of the intestinal crypt
- Mouse models
A functional link between inflammation, the tumor microenvironment, and cancer progression is now accepted. Historically, gastrointestinal (GI) cancers were among the first, where compelling associations between chronic inflammation, the tumor microenvironment, and progression had been noted. Such associations were based on the elevated risk for development of CRC associated with long-term inflammatory bowel disease (IBD),1 and more generally with autoimmune disorders affecting the GI tract.2 Evidence indicates that the persistent cycles of tissue damage and repair lead to molecular events that drive precursors lesions to cancer.3 However, pharmaceutical intervention with non-steroidal anti-inflammatory drugs in patients with chronic inflammation reduced colorectal cancer (CRC) risk not only in the general population,4,5 but also in those individuals with genetic predispositions for this malignancy.6 Here, prostaglandins provide a molecular rationale and therapeutic target assumed to be cyclooxygenase (Cox)-2, the rate-limiting enzyme for prostaglandin synthesis, although constitutively-active Cox-1 might also play a role.7,8 Excessive Cox-2 activity promotes CRC development by protecting neoplastic cells from apoptosis and enhancing angiogenesis.
Other mechanisms are now being uncovered by which various subsets of innate immune cells supplement the tumor microenvironment with cytokines, chemokines, and other mediators that promote malignancies. On the other side of the equation, the recognition that T-cell-mediated antitumor immunity also impacts on CRC has transformed our thinking. With the revelation that the status of immune cell infiltration and the T-cell repertoire at the edge of and within the centre of tumors is every part as predictive of patient outcome as classic histological and lymph node staging,9 the idea that the host immune system would have such a profound effect on patient outcome reinforces the view that CRC is more than malignant epithelial cells alone.
While much of the concept of tumor-promoting inflammation (and antitumor immunity) emerged from observations on patients, the power of genetics of the laboratory mouse now provides a tool to confirm links that underpin epidemiological associations, as well as to extend concepts formulated from the use of cell lines in the laboratory. The molecular precision of these models enables us to not only understand the impact of a particular gene mutation, but also to explore its function in a particular cell type. Importantly, it affords an exciting opportunity where genetic interactions underpinning CRC can be reconstructed in a mammalian model. Here, we focus on recent mouse models that are helping to define the roles played by cancer-promoting inflammation and stromal components of the tumor microenvironment, and how these activities might be integrated by a limited number of transcription factors in neoplastic cells.
- Top of page
- Complex microenvironment of the intestinal crypt
- Mouse models
The mouse has served cancer research more than any other organism, and for decades, spontaneously arising mutations and derivation of the corresponding strains enabled alignment of phenotypes with genotypes. Affordable genome-wide sequencing can now be expedited to saturate the entire genome with unbiased and random mutations using chemical or (retro-) viral approaches. These forward genetic screens are complemented by hypothesis-driven reverse genetics brought about by our practically limitless opportunities for “DNA surgery” that is facilitated by homologous recombination technology. The latter has enabled the identification of gene functions that, although required for the formation of the crypt niche, their expression is rather confined to non-epithelial cells comprising the niche. Historically, single deletions of genes with functions attributed to the immune response, including the interleukin (IL)-encoding IL-220 and IL-10 genes,21 were among the first non-epithelial genes found in mice to predispose to CRC alone or in combination with other gene deletions.22,23 Subsequent studies identified and demarcated roles for components of the innate immune system, including molecules with either extracellular (e.g. Toll-like receptors [TLR]) or intracellular sensing function (i.e. inflammasome), or associated with signaling components and/or effector function.24–27 Collectively, these studies suggest that non-epithelial components have tumor-promoting roles when unabated inflammation occurs in the same milieu as the epithelial cells that harbor (somatic) neoplastic mutations, conceptionally shifting the homeostatic niche to a pro-neoplastic microenvironment.
A key contribution to mouse models of intestinal tumorigenesis has been the isolation of the Min mouse (C57BL/6J-ApcMin/J strain) by William Dove and his colleagues in 1990.28 This strain arose from a random ethylnitrosourea (ENU) mutagenesis screen, and was initially identified by the onset of anemia. It was subsequently recognized as a paralog for the familial adenomatous polyposis (FAP) syndrome.29 It is noteworthy that, although this mutation is in all cells, Apcmin/+ mice develop adenomas predominantly in the proximal SI, and to a lesser extent in the colon. Akin to the mechanism initiating tumor formation in FAP patients with germline inactivation mutations in one APC allele, aberrant activation of canonical Wnt signaling occurs in response to spontaneous loss of heterozygosity (LOH) of the remaining Apc allele through somatic recombination,30 and this triggers GI polyposis.
Although much of the underlying biology is assumed to be similar, the prevalence of ApcMin mouse tumors in the SI sets this model apart from CRC that occurs in FAP patients. Indeed, the protein encoded by the ApcMin allele carries a more severe truncation mutation than Apc proteins arising from the “hot spot” non-sense mutations in humans. Consequently, a number of murine models have been designed to encode less truncated forms of Apc.31,32 Extensive comparisons of the corresponding allelic series suggest the existence of an intermediate level of aberrant Wnt signaling that promotes tumorigenesis most effectively, consistent with observations in humans that suggest a relationship between the primary and secondary mutations in this pathway.33,34
Triggered by the observation that tumor burden in ApcMin mice is dependent on modifier loci, which (as in the case of Pla2g2a) play a central function in inflammatory cells, numerous studies have now documented compounding effects from mutations in molecules that are associated with inflammation. Notably, tumor burden is reduced in ApcMin mice lacking the TLR-associated-signaling molecule MyD88, or deletion of inflammatory cytokines that signal through gp130. Conversely, induction of experimental colitis (with the associated cytokine storm arising from excessive infiltration of innate immune cells) exacerbates tumor load in ApcMin mice. Similarly, (mucin) muc2 ablation, which leads to impairment of the protective activity afforded by the mucous barrier, also increases tumor formation in ApcMin mice, with a shift of tumor location from the SI to the colon. Interestingly, Pla2g2a expression suppresses tumor formation in Muc-2-deficient mice.35 The connection with inflammation is extended in ApcMin mice to situations where compounding mutations are involved with the inflammatory response; these include the induction of inflammatory cytokines in response to ablation of the detoxifying enzyme glutathione S-transferase, Cox-2 or the prostaglandin receptor, EP2.36,37 Furthermore, the absence of Fas/Fas ligand interaction modulates inflammation and promotes a tumor-permissive environment,38 as does infection with enterotoxic bacteria in ApcMin mice via excessive IL-17 production and induction of the Th17 subset of lymphocytes, which is markedly reduced by IL-17A deletion.39 It is noted here that the presence of a global Apc mutation has systemic effects on the immune system. Thus, ApcMin mice suffer a progressive collapse of their hematopoietic (e.g. splenomegaly and stem cell deficits)40 and immune41,42 systems occurring before or in parallel with GI adenoma initiation. These observations imply that the inherent collapse of the immune system in ApcMin mice aids the development of adenomas.
Over the past decade, epithelial-restricted conditional Apc mutants and those expressing a constitutively-active form of β-catenin have enabled the field to more precisely model the acquisition of activating somatic mutations that underpin the majority of sporadic human CRC. Cre-recombinase driver strains allow for directed tropism of these mutations. For instance, deletion of Apc throughout the SI, using a naphthaflavone-sensitive Cyp1a1 : Cre transgene,43 resulted in devastating epithelial ablation due to Myc-dependent exhaustion of proliferating cells. Other Cre trangenes that restrict conditional or inducible recombinase activity to the transient amplifying and differentiated epithelial compartment (e.g. Fabp : Cre) produced milder effects. Meanwhile, the patchy ablation of Apc via Cre activity driven by Bmi1 and Lgr5 loci, that are active in the slowly- (quiescent) and highly-proliferating ISC compartment, respectively,18,19 resulted in the formation of tubular adenomas similar to those observed in ApcMin mice. Furthermore, confining Cre activity to both the ISC and the transient amplifying compartment using the regulatory elements of the villin (vil) or the cdx2 gene44–46 also mediated tumor formation. Note that these two transgenes drive recombination at a far higher frequency than the presumed, much rarer events that occur in sporadic human CRC. These differences raise the issue of potential field effects that might enhance tumor initiation. To address this concern, the use of Cre alleles, such as A33Cre, has been employed; these can be manipulated to drive recombination in a minority of colonic stem cells.47
The temporal control over inducible Cre drivers also sparked efforts to replicate aspects of the sequential accumulation of mutations that is believed to be part of the molecular journey that underpins tumor progression in humans. The timing and length of induction of either Cre-transgene expression (i.e. Cyp1a1 : Cre) or Cre (fusion-) protein activity in response to the administration of tamoxifen (i.e. CreErT2) or the progesterone analog RU486 (i.e. CrePR2) have been exploited in various lineage-tracing experiments to functionally dissect the homeostatic turnover of the intestinal epithelium.48 Experimental control over the duration of Cre activity in TgN (Cyp1a1 : Cre) mice allowed the targeting of Paneth cells,49 while Apc inactivation in response to the short induction of Cre activity induced adenoma formation in Lgr5ErT2Apcfl, but not in TgN (Cyp1a1 : Cre) Apcfl mice.50 Similarly, extended oral administration of tamoxifen conferred extensive recombination throughout the entire intestine in TgN (vil : Cre) R26lacZ mice, while the exposure of A33CrePR2mybfl/fl mice to RU486 initiated recombination in the rectum; progressive recombination towards the SI occurred only after several weeks of Cre activity.47 Thus, the cellular distribution of the Cre transgenes, along with the agent and administration route employed to activate the recombinase, enables temporal and spatial fine-tuning of mutations (Fig. 2).
Figure 2. Range of different gastrointestinal (GI) tropic gene expression patterns have been documented, and these have formed the basis for focused gene expression studies and knock out systems. (a) Genes display different expression levels across the intestines from stomach to anus. (b) Using this information and molecular approaches, genes encoding recombinases, such as Cre or Flp, have been either coupled or knocked into the regions adjacent to the promoters of these GI genes. Level of activation of expression of the Cre, and whether it is constitutively active or constructed as a fusion with a steroid receptor ligand binding domain of the estrogen receptor or progesterone receptor, provides different strategies for inducing gene recombination. Using fusions to fluorescent proteins further allows tagging the expressed Cre, as well as lineage tracing. Additionally, the route of steroid or chemical inducer administration either intravenously or in the mouse chow affects the kinetics of induction. (c) We find that adding RU486 and tamoxifen in the chow initially favors recombination in the rectum. Prolonged exposure in the chow for up to 4 weeks leads to almost complete recombination across the intestines, as shown by expression of β-galactosidase catalysis of X-gal (blue–green) in a Rosa 26 reporter mouse.
Download figure to PowerPoint
Mice have also been used to reconcile the finding that aberrant activation of the WNT pathway also occurs in approximately 10% of sporadic CRC through somatic mutation of CTNNB1. The Cre-mediated excision of exon 3, encoding the phosphorylation residues that mark β-catenin for proteosomal degradation, induces widespread tumor formation.51 Significantly, these are the very residues that are commonly subject to mutation in human CTNNB1, as well as its murine homolog, ctnnb1, in mice exposed to the colonotropic alkylating agent, azoxymethane (AOM).52 Indeed, a robust model for colitis-associated colon (CAC) cancer in rodents arises from an experimental setting, whereby the mutagenic AOM insult is followed by consecutive administration of the luminal irritant and toxin, dextran sulfate sodium (DSS). This strategy elicits cycles of mucosal damage, followed by tissue repair.53 During the early response to DSS, the colon undergoes a massive wave of apoptosis, resulting in impaired epithelial barrier function that enables commensal microbes to activate resident macrophages to release inflammatory cytokines, such as IL-1, tumor necrosis factor-α (TNFα), and IL-6. Accordingly, the CAC model is exquisitely sensitive to genetic and pharmacological interventions that affect and/or modulate the innate immune response. The emerging picture suggests that the immune cells that infiltrate the wounded epithelium and provide the signals that collectively promote an orchestrated wound healing response is subverted in the few cells where prior exposure to mutagens has induced oncogenic DNA damage. Thus, the overexpression of heparanase, which is a frequent observation in CRC and is believed to facilitate the release of sequestered heparin-binding growth factors, promotes chronic inflammation and cell growth to exacerbate CAC-associated tumorigenesis in a TNFα-dependent manner.54
Recent evidence indicates that some players in the host pathogen response, such as MyD88 and components of the inflammasome, might act in two ways to promote an inflammatory response, as well as being central in ensuring a homeostatic outcome for the continuous physiological renewal of the intestinal mucosa.26,55 This raises a potentially complex therapeutic challenge, whereby the same (sets of) factors and pathways might be engaged during homeostatic renewal, and in pathogenesis of colitis, as well as functionally connecting the microenvironment to neoplastic cell growth. Similarly, infiltrating adaptive immune cells might play a dual role in conferring an antitumor immune response, as well as regulating the epithelial response during mucosal inflammation.56 It has been argued that the DSS-based CAC model might not accurately mimic the Th2-biased immune cell response characteristic of ulcerative colitis. It will therefore be interesting to explore the extent by which the above findings are also applicable to a model where AOM is combined with the haptene oxazolone to trigger a NKT-cell dependent IL-13 response.57 Indeed, the predominant T-cell subtype associated with the inflammatory response might affect aberrant β-catenin activation in colonic adenomas of AOM-challenged mice in the Th1-mediated 2,4,6-trinitrobenzene sulfonic acid colitis model.58
AOM challenge has become the preferred experimental strategy in mice to mimic aberrantly-activated WNT signaling in sporadic human CRC; AOM biases disease in the SI, observed in Apc mutant mice, to the colon. Combining the two approaches has provided insights into its molecular etiology.59–63 It has also demonstrated a threshold effect on compounding mutations within the canonical Wnt-signaling cascade. Nevertheless, the mode of action of AOM in Apcmin mice remains unclear.64 It might reflect an apparent fine tuning of Wnt signaling required for adenoma formation,65 but the cellular differences between the SI and colon also still pose ambiguities.
Our capacity to genetically remove individual molecules from the mouse complements the reductionist dissection of signaling pathways in vitro that has been a mainstay of cancer research in building of our knowledge of pathway cross-talk and feedback relevant to CRC.66 Ultimately, many of these outcomes can be reconciled with transcriptional events in neoplastic cells, which further endow, maintain, or reinforce cancer hallmarks. In GI malignancies, this boils down to a small set of transcription factors that take the role of signaling nodes through which many pathways communicate. Here, we confine our attention to the three transcription factors NFκB, Stat3, and Myb, because of their involvement in CRC cells during tumor initiation and promotion in mouse models, as well as their excessive activation and/or expression in human CRC. Although these factors are activated by distinct mechanisms, including permissive transcription elongation (Myb), phosphorylation (Stat3, NFκB), proteolysis (NFκB), and nuclear translocation, they share many target genes and cooperate with each other. However, as the majority of sporadic CRC are based on dysregulated canonical Wnt signaling as the most likely tumor initiating event, the contribution of NFκB, Stat3, and Myb needs to be assessed against a backdrop of gene activation by the β-catenin/TCF4/Lef1 complex.
Although excessive NFκB activation is frequently detected in the form of elevated cytosolic and nuclear staining of p65RelA in primary and metastatic disease,67,68 attempts to find mutations in components of the canonical NFκB signaling (via IκB kinases [IKK]) pathway that might explain its persistent activation in CRC have been unrewarding. Although the active NFκB complex traditionally comprises p50 and p65RelA, lesser roles have also been attributed to c-Rel-, RelB-, and p52-mediating inflammatory responses. The two tiers of NFκB regulation include retention of the p50/p65RelA complex in the cytoplasm (when bound to IκBα), and the phosphorylation state of p65RelA in the nucleus. In response to pro-inflammatory signals, IκBα becomes phosphorylated by IKK, thereby enabling ubiquitin-mediated IκBα proteolysis. However, newly-synthesized IκBα might also enter the nucleus to retrieve active NFκB. Accordingly, it has been proposed that IκB family members might be exploited therapeutically to inhibit DNA binding of the p50/p65RelA complex.69 Alternatively, transactivation-defective p50 homodimers can compete with transactivation (domain)-proficient p50/p65RelA heterodimers and p50 homodimers for binding to the same cognate enhancer elements.70
Using TgN (vil : Cre)-mediated ablation of the IKK encoding IKKβ gene, Egan and colleagues demonstrated negligible effects of NFκB signaling on SI morphology or differentiation. However, they did observe that radiation or exposure to lipopolysaccharide71 caused a substantially enhanced apoptosis response. These data suggest that while epithelial NFκB plays a minor role under homeostasis, its function is required for epithelial repair; most particularly, this is the case during CAC.
Employing the aforementioned CAC model, Greten et al.72 showed that the production of pro-inflammatory cytokines by infiltrating myeloid cells was partly responsible for tumor growth. This depended on NFκB activation in non-epithelial cells, as ablation of IKKβ in myeloid cells reduced the number and size of colonic tumors. In contrast, IKKβ deletion in the intestinal epithelium conferred by (TgN) vil : Cre only reduced tumor numbers that were attributed to the reduced survival of neoplastic cells in the face of deficient NFκB signaling. These studies highlight the interplay between the tumor microenvironment and the intestinal epithelium more generally, and how NFκB activity across the two compartments functionally links inflammation to CRC.73,74
Stat3 is a latent transcription factor activated in response to cytokines and growth factors. In the GI tract, the latter are primarily comprised of the IL-6 cytokine family alongside the receptors for c-Met and EGFR ligands, as well as the tyrosine kinase c-Src.75 Receptor binding of IL-6 (or IL-11) triggers dimerization of the shared receptor subunit gp130, and subsequent activation of the Stat3 and Ras/extracellular signal-regulated kinase pathways.76 As binding of Socs3 to the activated gp130 complex results in its proteosomal degradation, tissue-specific Socs3 ablation in mice amplifies ligand-dependent gp130 signaling, while the tyrosine-to-phenylalanine substitution in the corresponding gp130Y757F knock-in in mutant mice destroys Socs3 binding; this results in excessive activation of Stat3 (and Stat1).76
Excessive activation of Stat3 is a recurring observation in a majority of epithelial malignancies,77 including CRC where tyrosine-phosphorylated Stat3 has been identified in half of all biopsies. As observed with many other solid maligancies, this activation has been noted most at the tumor margins and in peritumoral lymphocyes, and this has been associated with adverse clinical outcome and reduced survival.78 Akin to NFκB, there is no genetic evidence for constitutively-activating mutations within Stat3 itself, nor for tumor-specific locus amplification. However, a variety of (hemopoietic) malignancies harbor activating mutations of Jak2,79 and in-frame deletion mutations in GP130 that trigger ligand-independent activation of Stat3 in hepatocellular carcinomas.80 Excessive Stat3 activation can also arise from impairment mutations in, and epigenetic silencing, of genes encoding negative regulatory proteins, including Socs3.81 However, in the majority of cancers, excessive Stat3 activation reflects oversupply of soluble ligands in the tumor microenvironment. In particular, IL-6 has been put forward as the molecular component released by non-stem cancer cells to allow their conversion to cancer stem cells, and thereby maintain a dynamic equilibrium between these two tumor intrinsic cell types.82
Stat3 upregulates proteins of the Bcl-2 pro-survival family. In epithelial cells, it also induces other proteins that indirectly suppress apoptosis, such as the chaperone protein Hsp70, the C-type lectin-type RegIIIβ, and survivin,83 which are all overexpressed in CRC and IBD. The latter proteins not only suppress apoptosis, but might also promote cell cycle progression through binding to Cdc2. Stat3 also promotes the G1/S phase transition of the cell cycle more directly through the transcriptional induction of cyclinB1, cdc2, c-myc, and cyclinD1, and repression of the cell cycle inhibitor p21.83 As a third tumor-intrinsic property, Stat3 induces expression of the angiogenic factors, VEGF and HIF1α.83 Thus, excessive activation of Stat3 correlates with tumor invasion and metastasis in a variety of cancers.
In the absence of epithelial Stat3 expression, the CAC model yields reduced tumor formation. Conversely, excessive Stat3 activation, through epithelial-specific Socs3 ablation or introduction of the Socs3-binding deficient gp130Y757F mutation, results in increased multiplicity and size of these tubular adenomas.84,85 Administration of hyper-IL-6 (a fusion protein between IL-6 and soluble IL-6Rα), but not of IL-6, also increased tumor burden in CAC-challenged mice,85 suggesting that the extent of membrane-bound IL-6Rα, rather than gp130, limit the tumor-promoting response. Consistent with these observations, we found functional redundancy between IL-6 and IL-11, and that both cytokines conferred Stat3-dependent, epithelial resistance to apoptosis and colitis.84 Genetic deficiency for the ligand binding IL-11Rα subunit in the CAC model significantly abrogates colonic tumor formation in gp130Y757F mice, while systemic reduction of Stat3 expression in gp130Y757FStat3+/− mice also reduced their susceptibility to colon tumorigenesis in the CAC model (Ernst et al., unpubl. observ., 2011). Furthermore, intestinal tumor burden is reduced in ApcMin mice lacking IL-6, and in ApcMin mice that are also haplo-insufficient for IL-11Rα or Stat3. However, IL-11 administration protected against radiation-induced mucositis, suggesting that IL-11 signaling might play a physiological role in the maintenance of intestinal epithelial integrity.
Notwithstanding the central role played by excessive epithelial Stat3 signaling for the promotion of intestinal tumorigenesis, it has been recently suggested that this might also be part of an epigenetic switch mechanism that initiates tumor formation from non-transformed cells, rather than solely-expanding neoplastic cells that have arisen after exposure to mutagens.86 In breast cancer cells, this switch comprises excessive Stat3 activation as part of a feed-forward-signaling loop that also comprises c-Src and NFκB, both of which are excessively activated in intestinal tumors. Evidence in support for an autocrine/paracrine amplification loop that arises from the capacity of pStat3 to induce its own transcription and that of IL-6 and IL-1187 has now also been described for other solid malignancies. However, it remains unclear as to why epithelial Stat3 ablation in ApcMin mice results in a more invasive behavior of the few remaining lesions; these are more invasive when compared to age-matched Stat3 proficient animals (Ernst et al., unpubl. observ., 2011).
We and others have reviewed the evidence for Myb in CRC elsewhere.88,89 In brief, this includes the observation of the overexpression of MYB mRNA and protein88,90,91 in the majority (∼ 80%) of CRC, and in the evidence that this overexpression is of prognostic significance, being associated with metastasis.92 Mouse studies93 and human biopsy investigations have allowed the evaluation of premalignant adenomas90 to show the elevation of Myb in these. This indicates that increased Myb is a relatively early event. Myb is also required for proliferation of CRC cell lines, and is associated with perturbed differentiation and cell survival in vitro.88
The development of mouse models has been very helpful in the exploration of Myb function. Although embryonically lethal, global KO mice still allowed fetal transplant studies of the GI to be performed. The results show that Myb is essential for colonic crypt formation.94 More recent data indicate that Myb is required at diploid levels for the timely development of adenomas in ApcMin mice,93 and for the expression of genes considered to be Wnt targets, such as Myc,93 and Lgr5 (Cheasley et al., in press), and for recovery following radiation damage.89 Colons of Myb-/- mice fail to express Bcl-2,94 while CRC shows elevated Bcl-2 concordant with Myb overexpression.91
The recognition that a series of ENU-induced mouse mutants, initially identified for their defective development of blood cells, had impaired Myb function also provided an unexpected prospect to investigate this gene in the GI tract.89 Indeed, the very concept of using hypomorphic mutants is sometimes neglected in mouse studies. In fly studies, they often occupy centre stage, because hypomorphs might be viable, whereas classic KO animals might not be. Thus, defective, rather than absent, gene function can be investigated in adult animals.
Of particular relevance to this review is the observation that Cox-2 can be regulated by Myb alone95 or in partnership with the Wnt pathway in CRC.93Cox-2 is of particular interest in the context of CRC; when it is ablated in ApcMin mice, adenoma formation is substantially reduced, and survival extended.7 Myb also appears to regulate Bcl2,91 and perhaps BclXL,92 in CRC, as well as Grp78 an endoplasmic reticulum stress response gene.96 Grp78 is elevated in human CRC97–99 and in IBD, as well as various mutant mouse models following mucosal damage or enhanced pro-inflammatory stimuli.100–102 Like Bcl2, BclXL, and Cox-2, Grp-78 affords a survival advantage to cells, as well as protection against cytotoxic insult.
The regulation of the Myb gene is predominantly at the level of transcription; more specifically, transcriptional elongation.103 Notably, an attenuation region within the first intron is the principal determinant of whether mRNA is generated; this region is subject to mutations in Wnt-activated and mismatch repair-deficient CRC cell lines and primary tumors.103 No mutations in Myb coding exons have been reported, although occasional examples of amplification in CRC cell lines exist.88 The role of Myb in stroma has not been specifically investigated in the GI tract, but such a role is clearly important in hemopoiesis.104
Pathways working together
There is abundant evidence for an intimate link between inflammation-associated hyperactivation of NFκB and pStat3, including the coincident presence of NFκB, Stat3, and of Myb binding sites in the regulatory elements of many pro-survival genes. NFκB and Stat3-mediated signaling also converge on the epithelial–mesenchymal transition (EMT) process. Thus, IL-6-mediated Stat3 activation promotes EMT through the transcriptional induction of the E-cadherin repressor snail, while activation of NFκB promotes post-translational stabilization of the Snail protein.105 However, Stat3 signaling prolongs nuclear retention of canonically-activated NFκB through RelA/p50 acetylation and associated interference with its nuclear export.106
Meanwhile, unphosphorylated Stat3 can compete with IKKβ for binding to, and activation of, unphosphorylated NFκB, to trigger transcription of target genes independent of their binding sites for NFκB and/or Stat3. Both transcription factors can also act in a hierarchical fashion as part of a feed-forward loop, whereby NFκB induction of the RNA binding protein Lin28 blocks processing of the let-7 microRNA, and thereby derepresses the transcription of IL-6.86 Epistatic interaction also exists between aberrantly-activated Stat3 and Wnt/β-catenin pathways, for instance, based on the observation that tumors in the CAC-challenged gp130Y757F mice harbor activating mutations in β-catenin, and that gp130Y757FApcMin mice show increased tumor multiplicity, while enterocyte-specific Stat3 ablation reduced tumor incidence in ApcMin mice. While these two pathways share a common transcriptional response of Myc and cyclinD1 and other proliferative target genes, IL-11 administration and excessive Stat3 activation also facilitate survival of epithelial cells, conferring them with the capacity to repopulate the intestine after radiation damage. Stat3 seems to increase the pool of “stem” cells susceptible to tumor-inducing mutation, including LOH in ApcMin mice akin to the role of IL-6–Stat3 signaling in maintaining a dynamic equilibrium between stem and non-stem cancer cells.
Findings from our laboratories also suggest that Myb is required to maintain the stem cell pool and allow self-renewal through its action on the Lgr5 promoter (Cheasley et al., in press), although Myb expression in myeloid cells might be indirectly blocked by Stat3.107 Similarly, NFκB activation might be upstream of Myb in some instances, as NFκB regulates the transcriptional activation of Myb in proliferating hemopoietic cell lines108,109 and in the intestinal epithelium in response to radiation damage.110 Akin to Stat3, Myb can also regulate Myc expression alone or in partnership with the β-catenin/TCF4 complex;93 these mechanisms appear to extend to the regulation of Cox-2.
Compelling evidence now suggests that NFκB, Stat3, and Myb all can induce the expression of Cox-2 and other genes important in CRC (Fig. 3). It is tempting to speculate that such transcriptional hard wiring might ensure continuous activation of key genes in a “passing on the baton” like manner, despite the temporal and spatially restricted manner by which each of these transcription factors is active within the tumor and /or its microenvironment. For Cox-2, such a model might underpin the observations that the change in tissue location of its expression from stromal cells to the transforming epithelial cells is important for the process of CRC progression, where Cox-2 can stimulate proliferation and angiogenesis. For example, Ishikawa and colleagues inactivated the Cox-2 locus in myeloid, endothelial, and epithelial cells using tissue-specific TgN (LysM : Cre), TgN (VECad : CreERT2), and TgN (Vil : Cre) mice, respectively, in response to DSS challenge; only deletion in stromal, not epithelial components, influenced the resulting colitis.111 While CAC can be induced independently of epithelial Cox-1 or Cox-2, Cox-2 expression in myeloid cells is clearly important,111 and is elevated in the stroma in IL-10-deficient mice.112 Consistent, therefore, with the observation that Cox-2 contributes at various points in the progression of CRC, Cox-2 is regulated by multiple pathways,113 including canonical Wnt signaling,114 in concert with other factors, including Myb,93 NFκB,115,116 and Stat3.117
Figure 3. Interplay between transcription factors NFκB, pStat3, Myb, and activated Wnt signaling leads to the production of many genes involved in inflammation, proliferation, protection from apoptosis, and induction of angiogenesis. External stimuli by inflammatory cytokines and mediators leads to transcription of shared target genes. For instance, cyclooxygenase-2 (Cox-2) is induced by all pathways in different components of the intestinal crypt niche. IL, interleukin.
Download figure to PowerPoint
Using mouse GI cells in vitro
Embedded in the crypt niche are highly-proliferative cells that express the intestinal stem cell marker, Lgr5. This surface receptor for the Wnt-enhancing ligand R-spondin118 is regulated in part by the β-catenin/TCF418 and Ascl2119 components of the Wnt-signaling cascade and also by Myb (Cheasley et al., in press). Combining defined culture conditions with the ability to isolate Lgr5eGFP-positive cells confirmed that a single ISC could form crypt villus-like structures in vitro that comprise enterocytes, goblet, enteroendocrine, and Paneth cells.120 While this pioneering work implies that these purely epithelial organoids arise in the absence of a stromal niche, it does not exclude that these stem cells in their in situ environment are susceptible to the influences conferred by extracellular matrix components and soluble factors that stimulate or inhibit key crypt-signaling pathways and regulate anoikis. Interestingly, it appears that the frequency of organoid formation in vitro is increased if Lgr5-expressing cells are cultured in the presence of Paneth cells.121 This reflects the topographic arrangement within the crypt, where Lgr5-expressing cells are interspersed between Paneth cells, and is consistent with the observation that blockade of monocyte cytokine CSF-1 receptor signaling results in Paneth cell loss and a concomitant reduction of Lgr5 expression.122 It should not be ignored that Paneth cells serve in the immune system's first line of defense, as well as being immediately intercalated in the stem cell niche.
Like most epithelial cells, crypt cells have a preference to aggregate and respond to soluble and extracellular matrix-derived signals. It remains to be established whether adding back other cell types from the niche environment influences the capacity to grow organoid cultures from Lgr5-expressing cells. The ability to grow such organoids (Fig. 4) now affords opportunities to explore the role of various signaling pathways by culturing primary stem cells from mutant mice120,123 and CRC-initiating cells.124,125 Expanding the latter in immunocompromised mice126 has already started to provide novel insights in understanding intestinal biology and to allow investigators to address the enormous complexity of host–cancer interplay as it impacts upon the neoplastic target cells for transformation and progression to fully invasive CRC.
Figure 4. Ability to recapitulate some of the features of the intestinal crypt has been a major advance for gastrointestinal biologists. (a) Small intestine crypt is the engine room that produces a vast volume of epithelial cells. Key cell in the crypt niche is the lysozyme+ve Paneth cell (red arrow). With the advent of specialized culture methodologies, single Lgr5+ve cells or crypt fragments (b) can be exploited to generate “mini guts” or “organoids” that contain all the epithelial elements of the crypt (c). Optimal culturing success depends upon the presence of Paneth cells in the crypt fragments, but when single Lgr5+ve cells are employed, Paneth cells arise early in the organoids and can be identified by lysozyme staining or as granular cells (red arrow).
Download figure to PowerPoint
Conclusions and future directions
In conclusion, we have attempted to show the utility in studying CRC as a complex entity that embraces the epithelial tumor, along with an array of other tissue elements that collectively constitute the tumor microenvironment. The development of tissue-specific, inducible mouse mutants now allows for the detailed molecular dissection of the disease process. The combination of these mutants enables us to start rebuilding the interactions that most certainly occur in vivo. With technical advances, including live cell in vivo imaging technologies, in vivo cell ablation strategies, and miniaturized mouse colonoscopies, we can now monitor and control early events in the genesis of adenomas without killing the mice (Fig. 5). As in humans, the latter device provides the opportunity to introduce therapeutic interventions and to collect tissue biopsies. However, the ability to reproducibly isolate and grow intestinal stem cells and to form organoids is likely to enable us to conditionally modify their genomes by inducing Cre activity in vitro and to complement observations of corresponding mutations in vivo. We predict that these and other future studies will further cement the concept illustrated here that a small set of transcription factors, which act as common signaling nodes, will ultimately determine if and when the homeostatic process is subverted to support tumor progression and development of metastatic CRC.
Figure 5. Recent technological advances now allow the imaging of adenomas in the mouse rectum and distal colon. (a) Rigid colonoscopy successfully assists the detection of inflammatory disease in the colon (not shown), or (b) adenomas from mice such as that with Gp130Y/Y mutations following multiple azoxymethane exposures. This combination of genetic lesion and colon carcinogen recapitulates the human disease with the added utility that the mouse colorectal cancer is induced over a relatively short period of time.
Download figure to PowerPoint