Developmental pathways in colon cancer

Abstract

A hallmark of cancer is reactivation/alteration of pathways that control cellular differentiation during developmental processes. Evidence indicates that WNT, Notch, BMP and Hedgehog pathways have a role in normal epithelial cell differentiation, and that alterations in these pathways accompany establishment of the tumorigenic state. Interestingly, there is recent evidence that these pathways are intertwined at the molecular level, and these nodes of intersection may provide opportunities for effective targeted therapies. This review will highlight the role of the WNT, Notch, BMP and Hedgehog pathways in colon cancer.

Introduction

Alterations in genes that control developmental processes during embryogenesis and organogenesis are recognized as hallmarks of cancer. Pathways such as wingless-related integration site (WNT), Hedgehog (HH), Notch and bone morphogenic protein (BMP) are well-characterized in the developing embryo for establishing cell position, body pattern segmentation, polarity and cell fate decisions.¹ It is not surprising that several of these signaling pathways are altered in oncogenic processes.¹ This was perhaps first recognized in the hematopoietic system, in which overexpression of various Hox genes is associated with a variety of leukemias.²

Similarly, epithelial cancers also exhibit alterations in genetic pathways more commonly associated with embryogenesis.¹ This review will highlight four of these pathways, WNT, Notch, Hedgehog and BMP, and will discuss their role in colon cancer and their possible utility as therapeutic targets.

Colon Cancer

With an average of 50,000 deaths per year, colorectal cancer has emerged as the second leading cause of cancer death in the United States and worldwide.^3-5 Early diagnosis and surgical intervention, along with combination chemotherapy, has resulted in improved outcomes.⁶ However, there are few effective strategies to treat colon cancer once first-line approaches have been exhausted. Improved understanding of the cellular basis for colon cancer and the role that signaling pathways such as WNT (wingless-related integration site), BMP (bone morphogenic protein), Notch and HH (hedgehog) have in the establishment and maintenance of the tumorigenic state will be critical for the development of novel therapeutics.^7-9 The advent of small-molecule inhibitors for targeting these pathways and their success in other diseases, either as single agents or in combination therapy, provides a rationale for exploiting these pathways as potential targets in the treatment of colon cancer.¹⁰

The unit of structure in the normal colon is the crypt of Lieberkuhn, which is composed of colon stem cells, transit amplifying cells and terminally differentiated goblet cells, enterocytes and endocrine cells.¹¹ Each normal crypt is comprised of about 2,000 cells.¹² Similar to the crypt of the small intestine, less differentiated cells reside in the bottom, and terminally differentiated cells reside near the top. Thus, there is a developmental hierarchy from bottom (undifferentiated) to top (differentiated) in the colon crypts, which consists of a stem/progenitor compartment, a proliferative zone and a differentiated compartment (Fig. 1). These cells continuously cycle from undifferentiated in the bottom of the crypt through the terminally differentiated cells at the top.^11,12 This process is controlled by the colon stems cells and the microenvironment. The colon epithelium is replaced nearly every 5 d.

Figure 1. Signaling in the colon crypt. Illustration indicating the overall structure of the human colon crypt and the gradient of WNT, Notch, BMP and HH signaling. Also indicated are the stem, proliferative and differentiative zones of the crypt, illustrating how colon epithelial cells differentiate from bottom of the crypt to top.

The architectural structure of the colon is reflected by a gradient of WNT, HH, BMP and Notch signaling (Fig. 1).^1,8 Starting at the base of the crypt unit, Notch signaling is highest in the stem cell compartment and decreases as cells move upwards through the proliferative areas and into the differentiative areas. WNT signaling is similar, with the highest levels of expression being in the earliest stages of the proliferative compartment and tailing off in the differentiative compartment.¹³ BMP signaling is active in the differentiated compartment, and despite the presence of BMP protein, it is relatively inactive in early compartments in the base of the crypt due to the presence of the BMP inhibitor Noggin.^14,15 HH expression primarily occurs in the differentiated compartment. Thus, these gradients of developmentally regulated signaling pathways serve to establish the pattern of stem cell/self-renewal, proliferation and differentiation that comprise the colon architecture.

The normal colon has two distinct pools of stem cells, which together make up the total population of 16 stem cells.^11,12,16 The contribution of each pool to the total is not known. The first stem cell pool is localized in the crypt base and can be characterized by high LGR5 expression and is largely comprised of a proliferating population.^17,18 The next pool is nearby in the +4 position of the colon base (four cells away from the base of the crypt) and consists of relatively quiescent or dormant cells. This second pool exhibits high expression of BMI-1 and telomerase reverse transcriptase (TERT).¹¹

Nearly 70% of colon cancers arise from adenomatous polyps; masses of cells that emerge from the intestinal mucusa and project into the colon lumen.^6,19 Of the three broad types of polyps, juvenile (harmatomous), hyperplastic and adenomatous, only the adenomatous polyps are associated with the emergence of sporadic colon adenocarcinoma.⁶ Not all adenomatous polyps will give rise to colon tumors, and these can be further subdivided based on gross or histologic appearance. Visualization and classification of polyps are the basis for early detection and prevention of colon cancer through colonoscopy or sigmoidoscopy.⁶

Adenocarcinomas of the colon can take up to 10 y to completely emerge. This sequence was first proposed by Fearon and Volgelstein,²⁰ building upon the Knudson two-hit hypothesis,^21,22 to be a stepwise accumulation of specific genetic mutation. Activating mutations in the Ras oncogene, and inactivating mutations in the APC and/or p53 tumor suppressors are frequently found in the majority of sporadic colon tumors.^19,23-25 In part, due to a better understanding of the function of APC in WNT signaling, data has emerged that alteration of developmentally regulated pathways, which function in normal colon formation, have a role in colon carcinogeneis.^8,26

Pathways

WNT

In the canonical WNT pathway, WNT ligand binds to the Frizzled receptor and the LRP5/6 coreceptor.^26,27 This results in stabilization of the Disheveled gene, which leads to cytoplasmic accumulation and subsequent nuclear translocation of β-catenin. It is involvement of β-catenin that defines canonical WNT signaling as compared with alternative WNT/Frizzled signaling pathways.^26,28

The pool of β-catenin is regulated through phosphorylation by glycogen synthase kinase 3 (GSK3), in which phosphorylated β-catenin is targeted for proteosomal degradation.²⁶ Together, with Axin and APC, phosphorylated β-catenin is shunted to the proteasome. In the presence of WNT ligand, GSK3 activity is inhibited by cytoplasmic Disheveled, thus stabilizing β-catenin for translocation to nucleus. Stabilized, free β-catenin that translocates to the nucleus associates with TCF/LEF to regulate transcription. Noncanonical WNT signaling also uses the Frizzled receptor, but involves an increasing list of co-receptors that includes Cripto and Ror2.²⁹ This alternative pathway typically involves PLC and PKC signaling.³⁰ In colon cancer, WNT target genes include c-myc³¹ TCF-1,³² LEF1,^33,34 c-jun,³⁵ MMP-7,^36,37 CD44,³⁸ VEGF,³⁹ Jagged1⁴⁰ and BMP4.⁴¹ Many of these gene targets are likely tissue and/or cell type-specific, but Axin2 is widely used as a general indicator of WNT pathway activation.⁴² For a comprehensive list of WNT target genes, see the WNT homepage at www.wnt.stanford.edu.

Mutations in the WNT pathway cause colon cancer through constitutive activation of the nuclear β-catenin/TCF transcription factor complex.^13,43 The most well-documented WNT pathway mutation in colon cancer is loss of the APC tumor suppressor gene.^11,13,26,43 APC normally functions as part of a complex that also contains GSK-3β and Axin. This complex destabilizes β-catenin through its phosporylation by GSK-3β and its subsequent degradation by the ubiquitin/proteasome pathway. Nearly 90% of sporadic colon cancers contain loss of function at both alleles of APC, resulting in constitutive stabilization of β-catenin and activation of WNT pathway genes, namely TCF, which are required for colon crypt maintenance.⁴³ This results in inappropriate proliferation that presents as colon polyps. Interestingly, point mutations in β-catenin have been identified in the approximately 15% of sporadic colon cancers that bear wild-type alleles of APC.⁴⁴ These mutations in β-catenin render it insensitive to de-stabilization by the Axin/GSK-3β/APC complex and result in constitutive WNT signaling.^43,44 However, mutations in both APC and β-catenin have yet to be identified within the same tumor.⁴³

Improved strategies for high-throughput screening have driven the development of WNT inhibitors.^45-47 Clinicaltrials.gov lists compounds that are undergoing clinical testing for WNT pathway inhibition, including LGK974 (Novartis, NCT01351103), CWP232201 (JW Pharmaceutical, NCT01398462) and PRI724 (Prism BioLab, NCT01302405). Despite the fact that nearly 80% of colon cancers bear activating mutations in the WNT pathway, the PRI724 compound is the only one of these in which colorectal cancer is included in the study. In addition to synthetic inhibitors, monoclonal antibodies and natural products are also being tested as WNT inhibitors. OTSA011 is a mAb against Frizzled homolog 10 (FZD10) that is currently being tested (Centre Leon Berard/Oncotherapy Science, NCT01469975). An interesting trial that examined resveratrol, a plant-derived natural phenol, in colon cancer, was recently concluded.⁴⁸ Reseveratrol has been implicated in colon cancer prevention.⁴⁹ Based on reports that some of resveratrol’s effects could be attributed to WNT pathway inhibition, a small cohort of colorectal and normal patients was given low-dose, plant-purified resveratrol (NCT00256334).⁴⁸ However, no effect on WNT signaling was observed in colorectal cancer, although WNT was inhibited in normal colon mucosa.

Nonsteroidal anti-inflamatory drugs have been shown to have some effect in inhibiting WNT signaling. This would include aspirin, indomethacin and celecoxib.^50,51 These Cox-2 inihibitors are thought to inhibit WNT by suppressing prostaglandin E2 (PGE2) production, since PGE2 has been shown to promote WNT signaling. A connection between WNT and mTOR led to the observation that the mTOR inhibitor everolimus could reduce the number of colon polyps and cancer mortality in a mouse model.⁵²

Interestingly, inhibition of the WNT pathway may not be the only approach. Activation of non-canonical WNT signaling by HDAC inhibitors has been reported to inhibit the growth of colon cancer cells in the presence APC mutations.⁵³ Given the importance of the WNT pathway in several diseases, in addition to colon cancer, the development of effective inhibitory strategies should remain a priority.

Notch signaling

The Notch signaling pathway in humans consists of four receptors, Notch-1, -2, -3, -4 and at least five ligands, Jagged-1, Jagged-2, Delta-1, Delta-3 and Delta-4.⁵⁴ The Notch pathway is highly conserved, with homologs in species ranging from worms through Man.⁵⁵ In the canonical Notch pathway, ligand interaction with receptor results in a cascade of proteolytic cleavages mediated first by a metalloprotease, and second by a γ-secretase activity that is made up of at least the presenilin, nicastrin and Aph proteins.⁵⁴ These cleavage steps result in release of a constitutively active intracytoplasmic Notch (ICN) fragment that is then translocated to the nucleus, where it associates with CBF-1 and MAML-1 as part of a larger transcription complex.⁵⁶ The net effect of ICN is to switch transcriptional complexes of CBF-1 from repression to activation.⁵⁷ The precise signaling differences between different Notch-receptor and Notch-ligand pairing is not well-understood. While they all seem to go through the same pathway, there is evidence that different receptor-ligand parings yield distinct biological outcomes.⁵⁸ In addition, glycosylation of the Notch receptors adds an additional layer of complexity when considering biochemical and biologic outcomes from specific Notch receptor-Notch ligand pairings.^58-60 Notch signaling is terminated by CDK8-mediated phosphorylation of a PEST domain on the ICN. This then targets ICN for proteosomal degredation and allows the cells to be responsive to new Notch signals.^54-56

Notch-1, the most widely studied Notch receptor, was first identified from a t(7;9) translocation in a subset of T-cell acute leukemia (T-ALL), which fuses the cytoplasmic portion of Notch-1 to the T-cell receptor β-locus, resulting in constitutive activation of Notch-1.⁶¹ Subsequent studies confirmed that constitutive Notch-1 was likely the causative agent of this rare subset of T-ALL.⁶² More recent work has demonstrated that nearly 80% of all T-ALL bear activating mutations in the Notch-1 receptor, demonstrating a role for Notch-1 as an oncogene within the T-lineage.⁶³ On the other hand, Notch-1 functions as a tumor suppressor in chronic myelomonocytic leukemia (CMML), in which inactivating mutations of Notch-1 have been identified.⁶⁴ Within the B-lineage, enforced expression of constitutively active Notch results in the death of B-lineage acute leukemia (B-ALL) cell lines, indicating a tumor-suppressive role for Notch.⁶⁵ Thus, even within the hematopoietic system, Notch-1 can function as either to repress or promote tumorigenesis, dependent upon cell type.

In solid tumors, the role for Notch-1 is less well-characterized, but evidence for a dual oncogene/tumor suppressor role has been reported.⁶⁶ Oncogenic Notch has been reported in breast epithelial tumors and is thought to have a role in tamoxifen resistance.^67,68 On the other hand, some studies have reported a potential role for Notch as a tumor suppressor. Inactivating mutations in Notch-1 have been reported in squamous cell carcinoma, suggesting that the role Notch signaling may have in tumor biology is likely cell- and tissue type-specific.^69-71 The best characterized role for Notch as a tumor suppressor is in skin keratinocytes.⁷² The presence of functional Notch is required to protect keratinocytes from chemical or UV-mediated transformation.^73,74

The most well-documented molecular target of Notch-1 signaling is Hes-1.⁷⁵ Transcriptional activation of Hes or HEY family genes appears universal in nearly all Notch systems studied.⁷⁵ Other Notch targets appear to be more cell type-specific. c-Myc, p27^Kip1, p57, Akt, p53 and PTEN are just a few Notch target genes that have also been implicated in tumorigenesis, independently of Notch.^66,74

The role of Notch signaling in normal intestinal development has been well-documented and is the subject of several excellent reviews,^{1,11,12,76-78} and hence will not be extensively discussed here. However, the contribution of Notch signaling to colon cancer is not as well-characterized. Expression of the Notch-1 receptor and its target Hes-1 was reported to increase with increasing tumor grade in a gene array analysis consisting of 10% colon mucosa, 15% colonic polyps, 55% primary colon cancers and 13% liver metastasis.⁷⁹ Notch-2 levels were not increased, nor were levels of Jagged-1 or Delta-3 ligand. Expression of the Notch antagonist, Numb, was decreased in advanced colon cancers. These data suggest a general activation of Notch-1 signaling in colon cancer. Another group used in situ hybridization to examine the Notch pathway, and concluded that Notch signaling was active in colon tumors, but did not find a correlation between Hes-1 gene expression and survival among colon cancer patients.⁸⁰

Interestingly, investigators have reported that treatment of colon cancer cell lines with oxaliplatin, 5-FU or SN-38 (irinotecan) upregulates the γ-secretase complex and results in increased levels of cleaved activated Notch. They went on to demonstrate that treatment with a γ-secretase inhibitor would render the cell lines more sensitive to chemotherapeutic treatment.⁷⁹ In a recent clinical trial, however, the γ-secretase inhibitor RO4929097 was tested as single agent in metastatic colorectal cancer with little to no effect.⁸¹ There are currently two additional trials testing the efficacy of RO4929097 in colorectal cancer in combination with other chemotherapeutic drugs (NCT01198535 and NCT01270438).

Other recent research used a colorectal cancer explant model to evaluate the effectiveness of the γ-secreatese inhibitor PF-03084014 in combination with irinotecan.⁸² In this model, tumor fragments isolated from patient material were explanted into nude mice, which were then treated with PF-03084014, irinotecan or both. PF-03084014 plus irinotecan was more effective than γ-secretase inhibitor or irinotecan administered as single agents. This same group also demonstrated that for tumors with elevated Notch expression, PF-03084014 plus irinotecan treatment resulted in reduced tumor recurrence. The bulk of these effects were in ALDH⁺ tumor cells, which is a subpopulation enriched for CIC (cancer initiating cell) activity.^12,82 Presently, there is a phase one clinical trial with PF-03084014 (NCT00878189, Pfizer).

BMP

Bone morphogenetic proteins (BMP), first identified for their role in controlling bone formation, are members of the TGF β superfamily.⁸³ BMPs bind to the BMP receptors, BMPRI or BMPRII. Both of these are serine-threonine kinase receptors. BMP binding to BMPRII results in phosphorylation of BMPRI, which subsequently phosphorylates Smad1, Smad5 and Smad8. These then associate with Smad4, resulting in activation and nuclear localization.⁸⁴ Extracellular molecules such as Noggin can regulate BMP signaling by sequestering BMP away from the BMPRI and BMPRII.⁸⁴

In colon cancer, mutations in Smad4 or BMPRI have been shown to be responsible for juvenile polyposis.¹⁴ In sporadic colon cancer, loss of phosphorylation of Smad1, Smad5 and Smad8 has been observed in 70% of cancers.⁸⁵ Loss of Smad4 or loss or BMPRII is the likely mechanistic basis for loss of BMP signaling in sporadic colon cancers. However, because studies have indicated that loss of BMP signaling in sporadic colon cancers correlates with tumor grade, it is likely that this is not an initiating event (as it is in juvenile polyposis), but rather contributes to tumor progression.¹⁴

It is estimated that nearly 80% of colon cancers bear mutations in the TGF β family signaling pathway.¹⁴ Either BMPRI or BMPRII have been reported to be mutated in more than 70% of cases.⁸⁶ About 20–30% of colorectal cancer cases bear mutations in Smad4.^87,88 There is increasing evidence in sporadic colon cancers (as compared with JP) that mutations affecting BMP signaling corroborate with activated WNT to drive colon cancers, particularly in later stages.¹⁴ It has been reported that Smad4 levels are predictive of outcome/prognosis in stage II colorectal cancers. Those with low Smad4 levels only had a median survival of 1.7 y compared with greater than 9 y for patients with a high Smad4 level.^89,90 These clinical observations are consistent with a recent report that signaling through Smad4 can inhibit colon cancer progression by reducing the expression of β-catenin. BMP signaling has also been reported to promote the growth of colon carcinomas.⁹¹

Hedgehog

The Hedgehog (HH) pathway derives its unusual name from the phenotype of hedgehog loss in Drosophila; larvae take on a curled, bristly appearance that may remind some of a hedgehog.¹ In humans, there are three HH proteins, Sonic HH, Indian HH and Desert HH. Sonic HH is the most well-studied isoform.^1,92 HH is synthesized as a 45 kDa precursor that is self-cleaved into a C- and N-terminal peptides. The role of the C-terminal peptide is unknown, but the N terminal forms the active HH ligand.⁹³ Diffusible HH can bind to its receptor, Patched, which then de-represses the membrane-bound protein Smoothened (Smo). This last step results in activation and release of Gli transcription factors and their subsequent nuclear translocation.¹ Vertebrates have three Gli proteins. Gli1 will result in activation of HH target genes, while Gli3 is a repressor of signaling. Gli2 serves a dual role, with both repressive and activator functions.¹

Genes regulated by HH signaling include Myc, Bcl-2 and the Notch ligand, Jagged2.⁹⁴ Also induced by HH signaling are the stem cell-associated proteins LGR5, CD133 and CD44, as well as transcription factors that regulate epithelial to mesenchymal transition (EMT) such as Snail, Slug and Twist.⁹⁴

Mutations that result in activation of HH signaling are the driver mutations in basal cell carcinomas, for which there are now targeted therapies.⁹⁵ Pharmaceuticals targeting the HH pathway have also been used in some instances of multiple myeloma (MM) and chronic myeloid leukemia (CML).⁹⁶

Evidence from mouse models indicates that HH may cooperate with activated WNT to drive lethality in colon cells.⁹⁷ This suggests that HH inhibitors may be an interesting target to consider in colon cancer. The best known example of a HH inhibitor is the plant-derived steroid cyclopamine.¹ This alkaloid binds to Smo, resulting in its inactivation. Cycloplanine has been shown to have anti-tumorigenic activity in a variety of laboratory oncogenic settings. With the recent success of HH pathway inhibitors in the treatment of basal cell carcinoma [e.g., visodegib/erivedge (Roche/Genentech)], HH has moved more to the forefront of a potential targeted therapy in cancer.⁹⁶ Clinicaltrials.gov lists 34 trials recruiting for seven different inhibitors. Only one, LEQ-506 Novartis, lists colon cancer among the tumors to be tested. HH inhibitors are also being tested in combination with Notch inhibitors in advanced breast cancer and sarcoma.⁸

Communication Between Pathways

Figure 2 summarizes the linear process of signaling for the WNT, Notch, BMP and HH pathways. However, there is an increasing body of evidence from a variety of tissues that these developmental pathways exhibit cross-talk or share molecular points (nodes) of intersection.^1,29 The first report of such cross-talk was in fruit flies, in which it was shown that wingless (the fly homolog of WNT) signaling could be regulated by Notch via a mechanism in which disheveled would bind to the cytoplasmic tail of Notch.⁹⁸ WNT signals can also control Gli3 from the HH pathway.⁹⁹ HH can antagonize WNT signaling in the colon.^100,101 Likewise, HH has been reported to control the expression of the Notch ligand Jagged2, whereas WNT/β-catenin can control Jagged1.^102,103 Hes-1 can be activated by both Notch and HH signaling.^104-106 BMP and WNT appear to be interconnected via the PI3k/Akt pathway.¹⁰⁷ TGF β/Smad signaling promotes EMT through WNT, Ras, HH and Notch.¹⁰⁸ In APC mutant mice, Notch signaling is required for the development of colon polyps and subsequent cancer.^97,109 Thus, there is interplay between these pathways, and alterations in one could have potential effects on others.

Figure 2. Summary of Notch, WNT HH and BMP signaling pathways. Illustrated are the major components of each of the Notch, WNT, HH and BMP pathways, starting with ligands expressed by the signaling cell, receptors expressed on the receiving cell and cytoplasmic signaling intermediates and transcriptional effectors. In this illustration, signal transduction for each pathway travels from the top of the figure to the bottom.

Kwon et al. recently reported physical interaction between β-catenin and the cytoplasmic tail of membrane bound Notch.¹¹⁰ Only the active pool of β-catenin protein was capable of binding to Notch. They provided evidence that this interaction would result in degradation of β-catenin protein and is one mechanism by which WNT signaling is modulated.

In addition to cross-talk, these various developmental pathways can also have an impact on cell signaling pathways such as PI3K/Akt and Ras/Raf/Mek/Erk.^111-114 In one recent paper, it was demonstrated that Ras signaling is enhanced through increased stabilization of Ras in colon cancers that bear mutated APC (in the WNT pathway). The authors presented data that stabilization of Ras was controlled by altered WNT activity regulating recruitment of Ras to the proteosome.¹¹⁵ Given that Ras inhibitors have performed poorly in clinical trials; this paper emphasizes the need for examining and identifying molecular connections between pathways. Interactions between WNT and Ras have also been reported in lung cancer.¹¹⁶ Hedgehog and Ras have been reported to be interconnected in colon cancer.¹¹⁷ Other work has implicated interactions between PTEN/PI3K/Akt signaling and BMP in colon cancers.^112,118 Insulin-like growth factor-1 (IGF1) has been reported to promote the growth of colon cancer cells via a β-catenin-dependent mechanism.¹¹⁹ Targeting IGF1R with a specific monoclonal antibody or inhibition of Akt reduced colon tumor growth.¹²⁰ Connections between HH and p53 have also been proposed, further illustrating the complex interconnectivity between signaling pathways.^120,121

Summary

WNT, Notch, BMP and HH represent fundamental pathways that regulate development. In the past decade, a role for these pathways in oncogenesis has emerged: driving the development of targeted therapies. Research suggests that these pathways do not act in isolation, but are interconnected such that alterations in one lead to alterations in another. Understanding how signaling pathways are interconnected is critical to the development of successful targeted therapies. By focusing on the molecular points of intersection, it may be possible to develop more efficient strategies for therapy.

Recent studies into tumor heterogeneity illustrate the complexity of cancer at the molecular level.^122-125 These recent studies have indicated that the accumulation of mutations contributing to tumor heterogeneity may not be mere bystanders, but may have an active role in establishing a unique biologic phenotype that make each cancer individual.¹²⁶ By understanding the interplay between pathways such as WNT, Notch, HH and BMP we can hope to develop global strategies to effectively target a broad array of cancers.