WNT signalling events near the cell membrane and their pharmacological targeting for the treatment of cancer

Abstract

WNT signalling is an essential signalling pathway for all multicellular animals. Although first described more than 30 years ago, new components and regulators of the pathway are still being discovered. Considering its importance in both embryonic development and adult homeostasis, it is not surprising that this pathway is often deregulated in human diseases such as cancer. Recently, it became clear that in addition to cytoplasmic components such as β‐catenin, other, membrane‐bound or extracellular, components of the WNT pathway are also altered in cancer. This review gives an overview of the recent discoveries on WNT signalling events near the cell membrane. Furthermore, membrane‐associated components of the WNT pathway, which are more accessible for therapeutic intervention, as well therapeutic approaches that already target those components will be discussed. In this way, we hope to stimulate the development of effective anti‐cancer therapies that target this fascinating pathway.

Linked Articles

This article is part of a themed section on WNT Signalling: Mechanisms and Therapeutic Opportunities. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v174.24/issuetoc

Abbreviations

APC

adenomatosis polyposis coli

CRD

cysteine‐rich domain

DKK protein

Dickkopf protein

DVL

Dishevelled

FZD

Frizzled

GSK3

glycogen synthase kinase 3

LGR

leucine‐rich repeat‐containing GPCR

LRP

LDL‐receptor related protein

RNF43

ring finger protein 43

sFRP

secreted Frizzled related protein

SOST

sclerostin

Swim

Secreted wingless interacting molecule

TCF

T‐cell factor

TSPAN12

tetraspanin‐12

wg

wingless

WIF

WNT inhibitory factor

Wise

WNT modulator in surface ectoderm

ZNRF3

zinc and ring finger 3

Tables of Links

TARGETS
GPCRs a	Enzymes b
FZD	CK1, casein kinase 1
GPR124, ADGRA2	GSK3B
LGR4
LGR5
LGR6

LIGANDS
sFRP‐1	WNT‐1
WNT‐1	WNT‐3a
sFRP‐2	WNT‐7a
sFRP‐4	WNT‐8a
sFRP‐5
SOST

These Tables list key protein targets and ligands in this article that are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Southan et al., 2016), and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 (^a,bAlexander et al., 2015a,b).

WNT signalling

Originally named Int‐1, WNT 1 was identified as a preferential insertion site for the mouse mammary tumour virus in virally induced breast cancers (Nusse and Varmus, 1982). Shortly after, WNT1 was found to be the mammalian homologue of a Drosophila gene named wingless (wg) (Nüsslein‐Volhard and Wieschaus, 1980). In wg mutant fly embryos, major embryonic defects including the absence of proper wing development were observed. Similar phenotypes were found in flies with defects in the genes encoding zeste‐white (glycogen synthase kinase 3; GSK3), Dishevelled (DVL) and armadillo (β‐catenin). Subsequently, epistasis experiments revealed that these genes encode the core proteins of a developmental signalling cascade (Siegfried et al., 1992; Noordermeer et al., 1994; Peifer et al., 1994), which was named the WNT signalling pathway, as an amalgam of wg and Int‐1 (Cabrera et al., 1987). Around the same time, another protein, adenomatosis polyposis coli (APC), was found to interact with β‐catenin (Rubinfeld et al., 1993; Su et al., 1993). As APC mutations had previously been associated with a hereditary predisposition for colon cancer (Kinzler et al., 1991; Nishisho et al., 1991), this was the first time that the WNT pathway was directly linked with human disease (Clevers and Nusse, 2012).

While WNT genes have not been identified in the genomes of unicellular organisms, these genes are essential in even the simplest multicellular animals suggesting that WNT signalling might have been essential during this early step of evolution on our planet (Srivastava et al., 2008; Petersen and Reddien, 2009; Clevers and Nusse, 2012; Ryan et al., 2013). In line with this hypothesis is an evolutionarily widespread function of WNT signalling during embryonic development, which is to break symmetry of the developing early embryo. WNT signalling regulates axillary patterning in a widespread range of multicellular organisms ranging from sponges to humans (Petersen and Reddien, 2009; Holstein et al., 2011). In larger multicellular organisms, the functions of WNT signalling extend to tissue renewal and maintenance. During tissue turnover and upon injury or damage, WNT signalling regulates the re‐establishment of the normal body or tissue patterning and therefore plays an essential role throughout life (Clevers et al., 2014).

WNT signalling has since become a relevant therapeutic target for human disease, as the pathway is frequently deregulated in cancer and plays a pivotal role in development (which is illustrated by the diverse range of human diseases associated with congenital WNT pathway defects, as summarized by Clevers and Nusse (2012)). Like most mammalian genomes, the human genome harbours 19 different WNT genes. Recently, a wide range of small molecules and biological compounds that inhibit, block or somehow affect the WNT pathway have been developed and are currently being explored as therapeutics (Clevers and Nusse, 2012). Additionally, an ever‐growing number of publications reveals the complexity of WNT secretion and modification, WNT/Frizzled (FZD) interactions and signal amplification or inhibition by molecules such as R‐spondin (RSPO), DVL and the E3 ligases ring finger protein 43 (RNF43) and zinc and ring finger 3 (ZNRF3) (Hao et al., 2012; Koo et al., 2012; Lau et al., 2012; Jiang et al., 2015). This complex network of WNT signalling regulators creates a window of opportunity for therapeutic applications.

It has been known for a long time that mutations in cytoplasmic components of WNT signalling, such as APC or β‐catenin, can lead to cancer. However, in recent years, it has become clear that membrane‐bound or extracellular components of the WNT pathway are also frequently altered in this disease. These components are more accessible for therapeutic interference due to their cellular localization. For these reasons, this review aims to give an overview of the recent discoveries on WNT signalling events near the cell membrane. Furthermore, we discuss the therapeutic approaches that might be – or already are being – investigated to treat cancer by interfering with WNT signalling at the membrane.

β‐catenin‐dependent WNT signalling

This review focuses on β‐catenin‐dependent WNT signalling, also called “canonical” WNT signalling, a signalling pathway where the binding of WNT to its receptors leads to β‐catenin stabilization. However, there are several β‐catenin‐independent WNT pathways, including the WNT/Ca²⁺ pathway and the planar cell polarity pathway (Seifert and Mlodzik, 2007; Wang and Nathans, 2007). Although highly important during development and homeostasis (Seifert and Mlodzik, 2007; Wang and Nathans, 2007), these “non‐canonical” pathways are beyond the scope of this review. WNT signalling plays a crucial role during embryogenesis and also remains essential for adult tissue homeostasis throughout adult life. β‐catenin‐dependent WNT signalling has been shown to be involved in various cellular functions such as cell proliferation, cell differentiation, apoptosis, cell migration and polarity.

WNT proteins are 40 kDa cysteine‐rich proteins, which require lipid modification (cysteine palmitoylation) to function and bind to their receptors named FZDs (Tanaka et al., 2002; Willert et al., 2003; Takada et al., 2006). FZD‐proteins are seven‐pass transmembrane receptors with an extracellular N‐terminal cysteine‐rich domain (CRD) to which WNT binds (Bhanot et al., 1996; Dann et al., 2001; Janda et al., 2012). LDL receptor related protein 5/6 (LRP5/6 or Arrow in Drosophila) serve as co‐receptors and are required for WNT/FZD‐induced WNT signalling. (Pinson et al., 2000; Tamai et al., 2000; Wehrli et al., 2000). When bound to the FZD/LRP complex, WNT induces a molecular chain reaction, described in detail elsewhere (Barker and Clevers, 2006; Clevers, 2006; Clevers and Nusse, 2012). In short, WNT binding induces accumulation of β‐catenin in the cytoplasm, which results in its translocation to the nucleus. In the cytoplasm, β‐catenin is bound to a complex composed of APC, Axin, and the kinases GSK3 and casein kinase 1. In the absence of WNT signalling, this so‐called destruction complex recruits the E3 ligase β‐transducin containing protein. This ligase subsequently ubiquinates β‐catenin and thus targets it for degradation by the proteasome. In the presence of WNT however, the complex no longer recruits this E3 ligase, thereby preventing β‐catenin's ubiquitination (Li et al., 2012). Now, β‐catenin can stay bound to the destruction complex, saturating the complex and allowing unbound β‐catenin to accumulate and translocate to the nucleus. Nuclear β‐catenin displaces Groucho protein to form a complex with DNA‐p (T‐cell factor (TCF)/lymphoid enhancer factor) transcription factors and induces transcription of WNT target genes (Behrens et al., 1996; Molenaar et al., 1996; Van de Wetering et al., 1997; Cavallo et al., 1998; Roose et al., 1998). Transcriptional output of WNT/β‐catenin varies with cell type, although there are some ‘universal’ WNT/Tcf targets such as Axin2 (Jho et al., 2002) and SP5 (Weidinger et al., 2005). Therefore, while the initiating steps of WNT signal transduction are universal, the target gene response is highly context‐dependent.

Only recently, it has become clear that WNT signalling via GSK3 might not only exert its effect via transcriptional regulation as described above. Niehrs and colleagues showed that in mitosis, WNT functions independent of transcription. They showed that – in the absence of WNT – GSK3 actively ubiquinates mitotic effectors, thereby flagging them for degradation (Huang et al., 2015). This newly discovered WNT signalling pathway was named WNT/stabilization of protein signalling. Although beyond the scope of this review, this pathway – although independent of β‐catenin, does require GSK3. As such, it differs from the above‐mentioned β‐catenin‐independent pathways as it seems to follow the initial signalling events of canonical WNT signalling, to later divert from this by signalling without involvement of β‐catenin.

WNT is post‐translationally glycosylated and lipid‐modified

WNT proteins are cysteine‐rich, 350 to 400 amino acids long and contain an N‐terminal signal peptide for secretion. Smolich et al. (1993) showed that WNT proteins were N‐glycosylated at one or more amino acids. The function of this modification is still unknown, but there are implications that glycosylation is involved in the extracellular spreading of WNT, which might involve interactions with heparan sulphate proteoglycans (Tanaka et al., 2002; Vincent and Dubois, 2002; Eaton, 2006).

The crystal structure of WNT unveiled conserved cysteine‐residues forming intermolecular disulphide bonds in both the C and N‐terminal domain of the WNT molecule (Janda et al., 2012). These bonds proved to be essential for the proper functioning of WNT, as mutations in the conserved cysteines at the N‐terminal resulted in loss of WNT activity. Disruption of any of these disulphide bonds induced the formation of WNT oligomers through intramolecular disulphide bonding (MacDonald et al., 2014). In addition, lipid modifications of WNT proteins have been shown to be crucial for its function. In 2003, Willert et al. were the first to purify WNT protein, murine WNT3a, which revealed an addition of palmitate to what seemed to be Cys⁷⁷ (Willert et al., 2003). Later, this modification was also found on Wg and WNT8 of Drosophila, suggesting functional relevance (Willert et al., 2003; Zhai et al., 2004). Indeed, mutation of the conserved cysteine to alanine resulted in WNT that could still be secreted, but showed little or no signalling activity (Willert et al., 2003; Galli et al., 2007; Komekado et al., 2007). However, later work showed that this essential lipid modification was actually not found at the conserved cysteine, but at a different amino acid of the WNT peptide (Janda et al., 2012). Crystallization studies of WNT8 interacting with FZD revealed that the cysteine corresponding to Cys⁷⁷ in WNT3a was involved in a sulphide bond that is conserved throughout all WNT subtypes, and therefore could not carry this modification. In contrast, palmitoleic acid is added to the Ser¹⁸⁷ (Ser²⁰⁹ in WNT3a) that is conserved in all WNT subtypes. This conserved lipid modification appears to be required for WNT secretion. If prevented, WNT accumulates in the ER (Takada et al., 2006). The enzyme responsible for lipid modifications of WNT is porcupine, a multipass transmembrane ER protein that contains an O‐acyl transferase domain. Porcupine was identified as a component of the WNT pathway more than twenty years ago (Riggleman et al., 1990; van den Heuvel et al., 1993), and not much later was identified as the enzyme responsible for these post‐translational modifications of WNT (Kadowaki et al., 1996; Hoffman, 2000). Inhibition of this protein by IWP2 or other low MW compounds is widely used in vitro to inhibit WNT signalling (Chen et al., 2009; Koo et al., 2012).

It was further shown that the palmitoleic acid on WNT actually serves as one of the two binding sites for FZD. The other binding site for FZD is characterized by multiple hydrophobic amino acids (Janda et al., 2012). Conservation of the amino acid sequence in these two regions probably explains the cross‐reactivity between different subtypes of WNTs and FZDs, to which we will return later. The two domains of WNT that bind to FZD can be compared to a thumb and index finger interacting with the CRD (Figure 1).

Figure 1.

A graphical representation of Wnt ligand interacting with Fz8. The palmitoleic acid that is post‐translationally added to the Wnt molecule, is essential for the proper function of the ligand. The Wnt protein adopts a shape that can be seen as a thumb and index finger that interact with the Fz receptor. (Figure courtesy of Claudia Janda, Garcia lab, Stanford University).

WNT trafficking

Before WNT is secreted at the membrane, it needs to translocate from the Golgi to the cell surface. Multiple trafficking routes for WNT have been suggested, which include essential roles for a protein named WNTless (Bänziger et al., 2006; Bartscherer et al., 2006; Goodman et al., 2006) and the retromer, a protein complex (Seaman, 2012). Although beyond the scope of this review, many studies support crucial roles for these proteins, opening up possibilities to affect WNT signalling by interfering with WNT trafficking (Bänziger et al., 2006; Goodman et al., 2006; MacDonald et al., 2010; Coombs et al., 2010).

Extracellular WNT

How does WNT travel from one cell to the other to reach its target cell following translation, protein modification and delivery at the cell membrane? For a long time, it has been claimed that WNT is both a short‐ and long‐term morphogen, as evidence exists for both (van den Heuvel et al., 1989; Zecca et al., 1996; Neumann and Cohen, 1997; Thorpe et al., 1997; Goldstein et al., 2006). However, recent work in Drosophila argues against the requirement of a long‐range secretory WNT‐gradient in the wing imaginal disc by showing that the use of membrane tethered WNT delayed, but did not alter early development (Alexandre et al., 2014). The same study also showed, however, that later during fly development, long‐range WNT gradients become indispensable. It would be very interesting to see if this claim holds true for other developmental stages that are assumed to require a long‐range WNT gradient.

In addition to the distance that WNT travels to reach its recipient cell, the mode of travel for extracellular WNT has remained unclear. The lipidation of WNT impairs its diffusion through extracellular space. Indeed, there have been reports showing that purified WNT shows increased activity when packaged into liposomes, where lipids are shielded from the hydrophilic environment (Morrell et al., 2008). It has further been suggested that, in Drosophila, Wg might form multimers to hide its hydrophobic areas away from the outside, or bind to lipoprotein particles in order to travel extracellularly (Panáková et al., 2005; Katanaev et al., 2008). Secreted wg interacting molecule (Swim) has recently been identified as a WNT chaperone. In Drosophila, Swim allows WNT to travel throughout the extracellular milieu, while retaining its activity (Mulligan et al., 2012). A mammalian orthologue of Swim has not yet been identified. WNT retains its functionality only in the presence of serum when produced in vitro. However, it was unclear for a long time which component of the serum keeps WNT soluble and still functional. When kept in solution in the absence of serum, WNT loses its biological activity. Recently, Mihare and colleagues demonstrated that afamin (α‐albumin) is the component of serum that is responsible for keeping WNT in solution (Mihara et al., 2016). The authors show that WNT remains active in serum‐free medium in the presence of afamin, which seems to interact in a 1:1 molar ratio with multiple WNTs of both human and murine origin. How afamin stabilizes WNT and if afamin itself is involved in physiological WNT signalling in vivo is currently not known.

A recent study on intestinal epithelial organoid cultures, as well as crypts in vivo, demonstrated that WNT does not diffuse or is packaged into liposomes, but rather directly binds to the neighbouring stem cell and stays bound by FZD on the membrane. Subsequently, a WNT gradient is created by cell divisions diluting the WNT molecules by the increased cell membrane surface (Farin et al., 2016). Furthermore, it was shown that intestinal crypt stem cells maintain WNT on the membrane to prolong active signalling and do not internalize WNT. This work also showed that the receiving cell can still modulate the signal strength, a regulatory role previously only associated with the amount of WNT released from the secreting cell, or with the mechanism of its transport.

Regulation of WNT signalling at the cell membrane

At the cell membrane of the signal‐receiving cell, WNT interacts with co‐receptors FZD and LRP5/6 to initiate β‐catenin‐dependent WNT signalling. Multiple levels of regulation of this interaction were revealed over the past years and very likely more will be found in the future. We are beginning to understand that WNT signalling is tightly regulated by activators and inhibitors acting at different stages of the signalling cascade. This tight regulation underscores the importance of the pathway, which ‐if not properly controlled‐ can cause developmental defects and human diseases such as cancers, degenerative illness, metabolic diseases, disorders of endocrine function and autoimmune disease (Ho and Keller, 2015).

A growing number of publications suggests that WNT activity is still modifiable in WNT pathway‐mutant lines, and that even Apc mutant cells still depend on upstream signal components such as WNT or FZD. In 2002, this led Ron Smits and Riccardo Fodde to propose the Goldilocks (or ‘just right’) model of WNT signalling in cancer (Albuquerque et al., 2002). This model claims that not the high levels of activation but rather a specific increase in WNT signalling supports tumour formation. Indeed, the promoter regions of WNT inhibitors are often hypermethylated in colon cancers, most of which carry an inactive form of APC, suggesting constitutive WNT signalling. However, reactivation of these WNT inhibitors upstream of APC still attenuates tumour cell growth (Caldwell et al., 2004; Suzuki et al., 2004). Furthermore, in a colon cancer model, extracellular WNT‐trapping by the FZD ectodomain inhibits tumour growth in vivo (Vincan et al., 2005). Similar results were obtained in hepatocellular carcinoma cells, where soluble FZDs did inhibit tumour cell growth, but left normal hepatocytes unaffected (Wei et al., 2011). These data suggest that targeting the extracellular or membrane‐associated components of the WNT pathway might prove effective in treating cancer, even if the WNT pathway is mutated downstream.

The next section of this review focuses on the regulatory mechanisms that govern WNT signalling at the membrane of the receiving cell. Additionally, we will provide a detailed description of the ongoing work that explores possibilities to alter WNT signalling in cancer. Most approaches that are currently explored aim to modify these cell‐borne ways to regulate WNT signalling and, thus, to exploit the intrinsic WNT regulatory mechanisms of the cell to therapeutically target aberrant signalling in disease. For an overview of the molecules (both endogenous and exogenous) that have been tested to therapeutically interfere with WNT signalling, see Table 1.

Table 1.

An overview of the molecules tested to therapeutically interfere with WNT signalling at the membrane. Both endogenous and exogenous molecules are listed

Regulator	Target	Tested in tumour types	Clinical trial	References
miRNA‐126	FZD7	Acute myeloid leukaemia	No	(Li et al., 2015)
miRNA‐204	FZD1	Head and neck, endometrial cancer	No	(Chung et al., 2012)
miRNA‐23b	FZD7	Colon cancer	No	(Zhang et al., 2011)
miRNA‐27b	FZD7	Gastric cancer	No	(Geng et al., 2016)
miRNA‐142‐3p	FZD7	Cervical cancer	No	(Deng et al., 2015)
miRNA‐493	FZD4	Bladder cancer	No	(Ueno et al., 2012)
OMP‐18R5: anti‐FZD7 (Vantictumab, Oncomed)	FZD	Breast, colorectal, lung, pancreatic and neuroendocrine tumours, sarcoma	Yes	(Chung et al., 2012; Smith et al., 2013).
OMP‐54F28: FZD8‐Fc(Ipafricept, Oncomed)	WNT	Pancreatic, ovarian and hepatocellular cancer	No	(DeAlmeida et al., 2007; Kahn, 2014)
Anti‐WNT1 and anti‐WNT10b antibodies	WNT	Head and neck, colon cancer	No	(Rhee et al., 2002; He et al., 2005)
Fc‐sFRP1	WNT	Hepatocellular carcinoma	No	(Hu et al., 2009)
Fc‐WIF1	WNT	Hepatocellular carcinoma	No	(Hu et al., 2009)
Anti‐LRP antibodies	LRP6	Breast cancer, teratoma	No	(Gong et al., 2010)
Anti‐DKK1	LRP5/6	Multiple Myeloma	No	(Yaccoby et al., 2007)
OMP‐131R10	RSPO3	Colon, lung and ovarian cancer	Yes	www.oncomed.com/pipeline.html

WNT/FZD/LRP subtypes and levels at the cell membrane

WNT/FZD subtypes

There are 19 different WNT proteins and ten different FZD family members identified in humans (Bhanot et al., 1996; Kikuchi et al., 2011). Although certain WNTs preferentially activate β‐catenin‐dependent WNT signalling (WNT1, WNT3A and WNT8), it is virtually impossible to classify WNTs according to the pathway they induce. The combination of a certain WNT subtype together with the receptor they encounter on the cell membrane of the recipient cell ultimately determines the outcome (Niehrs, 2012). This combinatorial code can be seen as the first level of WNT regulation: It is this specific combination of WNT subtypes and receptors present at the membrane that is going to determine which signalling pathway is initiated, independent of the extensive levels of regulation that we will discuss later.

WNT/FZD/LRP levels

Overexpression of certain WNT and FZD subtypes have been reported in cancer. Also, FZD expression is increased during carcinogenesis (Tanaka et al., 1998; Vincan et al., 2007; Ueno et al., 2009; Yang et al., 2011; Phesse et al., 2016; Xu et al., 2016). Down‐regulation of FZD led to decreased tumour cell growth both in vivo and in vivo. A recent study revealed an additional path to FZD up‐regulation. This work showed that miRNA miR‐23b, which is down‐regulated in many types of cancer, targets amongst others FZD₇ (Zhang et al., 2011). In acute myeloid leukaemia, knockdown of miRNA‐126 increases FZD expression. This results in acceleration of the disease, but at the same time sensitizes the tumour to standard chemotherapy (Li et al., 2015). miRNA‐204 down‐regulates FZD₁ on the membrane and acts as a tumour suppressor in endometrial cancers (Chung et al., 2012). Other miRNAs targeting FZDs that are down‐regulated in tumours are miRNA, 23b, miRNA‐27b, miRNA‐142‐p3, miRNA‐493 and miRNA‐194 (Zhang et al., 2011; Krutzfeldt et al., 2012; Ueno et al., 2012; Deng et al., 2015; Geng et al., 2016). Their down‐regulation leads to FZD up‐regulation in these cancers. As the previously mentioned work shows, reintroducing these miRNAs indirectly affects FZD levels in tumours and, thus, alters WNT signalling. Further research is required to show if such an approach is applicable in the clinic.

FZD are located on the cell surface and are therefore easy targets for antibodies. The fully humanized antibody OMP‐18R5 (or Vanticumab) binds to FZD_1,2,5,7,8 and prevents its interaction with WNT (Gurney et al., 2012). Its efficacy as an anti‐cancer agent is currently being explored in clinical trials. Initial results with three patients treated for neuroendocrine tumours that presented with prolonged stable disease appear promising (Chung et al., 2012; Smith et al., 2013). Another agent, OMP‐54F28, is an Fc‐fusion protein with FZD₈, whose efficacy was originally tested in teratoma cell lines (DeAlmeida et al., 2007). The molecule, which binds to all WNT subtypes, has entered clinical trials and is currently being tested for drug safety (Kahn, 2014).

WNT subtypes are also overexpressed in cancer (Leethanakul et al., 2000; Holcombe et al., 2002; Benhaj et al., 2006). However, the difference of WNT expression between healthy and tumour cells can vary depending on the WNT subtype. As the effect on cell behaviour for each WNT subtype is not yet completely understood, this makes WNT expression data difficult to interpret. However, Rhee et al. showed that treatment with anti‐WNT antibodies decreases β‐catenin expression levels in head and neck tumour cell lines and induces tumour cell apoptosis in vitro (Rhee et al., 2002). Furthermore, anti‐WNT1 antibodies were effective in patient‐derived primary colon cancer cultures, and that the use of such antibodies showed synergism with standard chemotherapy (He et al., 2005).

A study testing different antibodies raised against LRP6 revealed that different antibodies can inhibit or potentiate WNT signalling depending on where they bind to LRP (Gong et al., 2010). Antagonizing antibodies prevent interaction of WNT with LRP5/6 and stimulating antibodies seem to induce dimerization of LRP5/6. However, the effect depends on the cell line is tested and also on the WNT isoforms expressed. Nevertheless, if classified based on the WNT subtype expressed, the responses were consistent. This needs to be taken into account when applying LRP5/6‐antibodies in the clinic.

FZDs as GPCRs

As FZDs are serpentine receptors, they are typically coupled to trimeric G protein complexes and classed as GPCRs (Gilman, 1987). Indeed, FZDs function as GPCRs in β‐catenin‐independent WNT signalling. However, β‐catenin‐dependent signalling also requires FZDs to function (at least partly) as GPCRs (Katanaev et al., 2005; Kopein and Katanaev, 2009; Egger‐Adam and Katanaev, 2010; Koval and Katanaev, 2011; Koval et al., 2011; Halleskog and Schulte, 2013). Treatment with Pertussis toxin, a widely used G protein inhibitor, reduced WNT/β‐catenin signalling in a reporter teratocarcinoma cell line (Liu, 2001). More targeted inhibition of only certain G protein subunits showed a similar effect. In addition, it was shown that Pertussis toxin prevented β‐catenin stabilization in mouse primary microglia (Halleskog and Schulte, 2013). Tomlinsen and colleagues showed a G protein dependence in WNT signalling in Drosophila development (Katanaev et al., 2005). In addition, it was shown that G proteins directly bind to Axin2 and DVL, thereby stimulating WNT signalling (Egger‐Adam and Katanaev, 2010).

Although the exact molecular mechanisms are still not understood, the fact that FZD receptors might function as GPCRs is highly relevant when thinking about pharmacological interference with the WNT pathway. A recent review describes the role of FZDs as GPCRs in more detail (Schulte, 2015). It is likely that we will learn more about the molecular details of this route in the coming years, which will lead to the identification of new molecular targets of the WNT pathway.

WNT signalling enhancers or inhibitors

At the level of ligand perception, there are multiple WNT signalling enhancers and inhbiitors that control WNT signalling (Lau et al., 2012). Secreted FZD‐related proteins and the WNT inhibitory factor (WIF) directly bind to WNT and prevent it from binding to its receptor. Dickkopf proteins (DKK), WNT modulator in surface ectoderm (Wise) and Sclerostin (SOST) bind to LRP and prevent WNT from binding to it (Leyns et al., 1997; Hsieh et al., 1999; Jones and Jomary, 2002; Itasaki et al., 2003; Kawano and Kypta, 2003; Niehrs, 2006). Norrin can bind to FZD and actually induce WNT signalling. Another WNT signalling enhancer is GPR124 that activates specifically WNT7‐induced signalling. Although the biological effect of these molecules on WNT signalling has been known for a while, only recently their structural interactions with WNT pathway components are beginning to be revealed. These interactions, described and summarized by Malinauskas and Jones (2014), are pivotal in the development of future therapies to modulate WNT signalling. Through studying structural interactions used by these endogenous modulators, we can learn which synthetic compounds or molecules are most likely to be functional.

Norrin

Norrin can bind to FZD and induce formation of the FZD/LRP complex, thereby inducing WNT signalling (Malinauskas and Jones, 2014). So far, Norrin has only been shown to bind FZD₄ but not any other of the FZD family members (Xu et al., 2004). The protein is mostly known from Norrie disease, an X‐linked congenital retinal dysplasia where, in most cases, affected individuals present with blindness at birth. Patients lacking functional Norrin present with additional symptoms such as mental retardation and progressive hearing loss. In 2013, the crystal structure of Norrin revealed that the protein functions as a dimer containing separate binding sites for both FZD and LRP (Ke et al., 2013). Surprisingly, Norrin shows no sequence similarity with WNT. However, a structural similarity between the two was recently found: a cytokine‐like fold in the C‐terminal domain of the protein (Ke et al., 2013).

Tetraspanin‐12 (TSPAN12) is a member of the tetraspanin family (Serru et al., 2000) that is essential for the positive effect of Norrin on WNT signalling (Junge et al., 2009). Tetraspanins are four‐transmembrane proteins that cluster together and serve as signalling platforms on the membrane (Hemler, 2005). Ablation of TSPAN12 expression leads to a decrease of Norrin‐dependent but not WNT‐dependent signalling suggesting TSPAN12 specificity for the Norrin/FZD interaction. Later data confirmed that TSPAN12 stabilizes the interaction between FZD and LRP5 to induce β‐catenin signalling (Knoblich et al., 2014). Also TSPAN12 functions in tumour development and metastasis (Knoblich et al., 2014), suggesting that targeting of this protein might prove effective in the treatment of cancer. Further studies may therefore provide new insights into the molecular mechanism by which TSPAN12 affects Norrin function. A schematic representation of the Norrin/TSPAN12‐mediated activation of the WNT signalling pathway is given in Figure 2.

Figure 2.

The effect of Norrin and TSPAN12 on Wnt signalling. Norrin forms dimers to interact with FZD₄ receptor, thereby activating the Wnt signalling pathway in a Wnt ligand independent fashion. TSPAN12 reinforces this effect by stabilizing the interaction between LRP and FZD₄.

GPR124

Another enhancer of WNT signalling was found in 2015.The adhesion GPR124 (now called ADGRA2), proved to be a WNT7‐specific enhancer of WNT signalling (Zhou and Nathans, 2014). At that time, this protein was already known to be essential, during embryogenesis, for CNS angiogenesis and the integrity of the blood brain barrier. Focusing on these systems, GPR124 loss in embryonic mice resulted in a phenotype that was highly similar to that resulting from WNT signalling defects (Kuhnert et al., 2010; Anderson et al., 2011; Cullen et al., 2011). Therefore, researchers suggested that this protein played a role in WNT signalling. Indeed, in 2015, GPR124 was shown to enhance WNT signalling, and to do so in a WNT7‐specific manner. Furthermore, the authors showed that inactivation of this gene resulted in a relatively mild phenotype and that this was due to partial redundancy with Norrin‐induced signalling. Indeed, inactivation of both pathways resulted in a severely disturbed development of the CNS.

As GPR124 is up‐regulated in tumour vasculature (Croix, 2000; Carson‐Walter et al., 2001), it is likely that this newly discovered modifier of the WNT pathway might also function outside the CNS. Following this rationale, inhibition of GPR124 could synergize with other anti‐angiogenesis therapies that are already applied in the clinic to prevent tumour growth. Although only hypothetical at this time, we expect that such therapies will be explored in the future. The regulatory effects of GPR124 on WNT7 induced WNT signalling is summarized in Figure 3.

Figure 3.

The effect of GPR124 on Wnt7‐induced canonical Wnt signalling. GPR124 enhances the effect of Wnt7 binding to FZD1/4. The molecular mechanism explaining this positive regulatory role of GPR124 remains to be elucidated.

Secreted Frizzled related proteins

The five known secreted FZD related proteins (sFRPs) comprise the largest family of WNT inhibitor proteins. At their N‐terminal, SFRPs contain a CRD that is very similar to the CRD of FZD suggesting that they compete with FZD for interaction with WNT (Dann et al., 2001). However, Rubin et al. actually showed that sFRP1 lacking its CRD is still capable of inhibiting WNT signalling. This suggests that the regulatory effect of sFRP1 is independent of the CRD (Üren et al., 2000). Furthermore, the authors were the first to reveal a biphasic effect of sFRPs on WNT signalling. At higher levels, sFRP1 blocked WNT activity. However, at lower concentrations, the presence of sFRP resulted in the opposite effect, an increase in WNT signalling. The molecular mechanisms underlying this effect remain to be elucidated. The first family member to be identified was Frzb, which was shown to bind to WNT8 and block WNT signalling in Xenopus (Leyns et al., 1997; Wang et al., 1997). Thereafter, it was quickly recognized that sFRPs not only bind to WNT but can also bind to one another to regulate their activity (Yoshino et al., 2001) and bind to FZD (Bafico et al., 1999; Rodriguez et al., 2005). Interestingly, aberrant expression of these proteins was observed in several types of cancer (Rubin et al., 2006) and other diseases (Jones et al., 2000; Bodine et al., 2004; Berndt and Kumar, 2007) suggesting interference with these molecules might be useful as a therapeutic approach. Indeed, treatment with antibodies consisting of human Fc‐domains fused to sFRPs resulted in reduced tumour growth in hepatocellular xenograft models (Hu et al., 2009). There are also reports describing low MW inhibitors of sFRPs that were effective in treating bone disease (Bodine et al., 2009), but to date these have not been tested in cancer models.

WNT inhibitory protein 1

WNT inhibitory protein 1 also directly binds WNT to inhibit its activity. The protein contains five EGF repeats and a WIF domain, the latter being responsible for its WNT inhibitory effect⁴. Although the WIF domain has been shown to be sufficient for both WNT binding and inhibition of WNT signalling, recent data suggest that the EGF‐like domains enhance binding of WNT to WIF1 (Malinauskas et al., 2011). Further work demonstrated that WIF1 interacts with WNT protein at multiple sites (Kerekes et al., 2015). In a wide range of cancers, the Wif1 promoter region is methylated (Ai et al., 2006; Kawamoto et al., 2008; Veeck et al., 2009), down‐regulating WIF1 expression. Indeed, re‐establishing WIF1 expression in renal cell, bladder and cervical cancer models induced apoptosis and suppressed tumour growth and invasion in vivo (Kawakami et al., 2009; Tang et al., 2009; Ramachandran et al., 2012). Similar to the above‐described Fc‐sFRP antibodies, Fc‐WIF1 antibodies also reduced growth in hepatocellular xenograft models (Hu et al., 2009).

Dickkopf proteins

The Dickkopf family of proteins consists of four DKK members (DKK1‐4) and a DKK3 related protein named Soggy (Niehrs, 2006). The C‐terminal domain of DKK is the most conserved between family members and is sufficient to inhibit WNT signalling (Brott and Sokol, 2002; Li et al., 2002). In contrast to the above‐described regulators, DKKs specifically inhibit the β‐catenin dependent WNT pathway, as they interact with LRPs, which are restricted to this specific WNT signalling pathway (Bafico et al., 2001; Mao et al., 2001; Semenov et al., 2001). As an exception to this rule, accumulating data shows that DKK subtype DKK3 does not affect β‐catenin‐dependent WNT signalling, but exerts its role in other (β‐catenin‐independent) WNT pathways (Krupnik et al., 1999; Mao et al., 2001; Mao and Niehrs, 2003). DKK1 and DKK2 have been well documented as modulators of the WNT pathway. Intriguingly, and in contrast to the strong inhibitory effect of DKK1, the effect of DKK2 on WNT signalling is context‐dependent. For example, in vitro studies with human fibroblasts have suggested it functions as an inhibitor (Wu et al., 2000), while in mouse fibroblasts it serves as a weak activator of the pathway (Li et al., 2002). A later study suggested that this difference in effect might be explained by the absence or presence of Kremen (Nakamura et al., 2001), which, when present, forms a trimeric complex with LRP and DKK. This trimer is subsequently internalized, leaving less LRP on the cell membrane to serve as WNT receptor (Mao et al., 2002). Both DKK1 and DKK2 interact with Kremen but DKK3 functions in a Kremen‐independent way (Mao and Niehrs, 2003). In contrast, other studies report that Kremen is dispensible for, at least, DKK1 to function as an inhibitor (Wang et al., 2008) and that interaction of Kremen with DKK1 does not lead to internalization of LRP (Semenov et al., 2008).

A fully human anti‐DKK1 antibody proved to be effective in mouse and primate models of postmenopausal osteoporosis (Glantschnig et al., 2011). To our knowledge, there are currently no trials ongoing to test the efficacy of such antibodies specifically in cancer‐associated bone loss. However, the potential of these antibodies to treat cancer was recently explored. In a rat model for multiple myeloma, where DKK‐producing tumour plasma cells home in bone and destroy bone structure, treatment with anti‐DKK1 antibodies not only decreased bone density loss but also led to reduced tumour growth in vivo (Yaccoby et al., 2007). In contrast, a multitude of studies show that in solid tumours, DKK proteins are actually down‐regulated by epigenetic silencing (Aguilera et al., 2006; Hayashi et al., 2012) or other means (Tsuji et al., 2000, 2001; Nozaki et al., 2001; Kurose et al., 2004; Tanimoto et al., 2007; Du et al., 2011). Overexpression of DKKs inhibits proliferation in different tumour types in vitro (Tsuji et al., 2001; Hsieh et al., 2004; Abarzua, 2005; Tanimoto et al., 2007; Kawasaki et al., 2009), suggesting that re‐activation of DKKs (or administration of these molecules in vivo) might prove effective in the treatments of these type of tumours.

Wise and sclerostin

Wise was discovered in Xenopus embryos as a negative regulator of the WNT pathway. Alternative names for this gene are USAG‐1, Ectodin or Sostdc1. SOST is a Wise‐related protein that contains 38% sequence similarity to Wise. Both Wise and SOST were shown to interact with LRP5/6 and compete with WNT for binding (Itasaki et al., 2003; Li et al., 2005; Semenov et al., 2005; Lintern et al., 2009). In different stages of development, Wise actually functions as an activator of the WNT pathway suggesting a context‐dependent effect. The mechanism behind this regulation still requires clarification, and for this reason, the role of extracellular Wise remains somewhat controversial. However, a study describing Wise in the ER showed a more dedicated role for the protein in WNT signalling. ER retained Wise constitutively inhibits the WNT pathway by reducing LRP6 expression on the cell membrane (Guidato and Itasaki, 2007). As SOST is involved in bone mass maintenance, anti‐SOST antibodies have proven to be effective to treat osteoporosis (Li et al., 2009; Ominsky et al., 2010) and such antibodies are currently tested in clinical trials to treat this disease (Padhi et al., 2011; Cosman et al., 2016; MacNabb et al., 2016). Anti‐SOST antibodies have not yet been tested in cancer models. However, as SOST is overexpressed in tumours (Colucci et al., 2011; Mendoza‐Villanueva et al., 2011; Gkotzamanidou et al., 2012; Garcia‐Fontana et al., 2014), it is worth exploring whether such molecules can provide us with new opportunities for cancer treatment.

WNT modifying enzymes

TIKI, a WNT protease

TIKI1, named after a large‐headed humanoid in Polynesian mythology due to its effect on Xenopus embryos, has been shown to be a negative regulator of WNT signalling that functions upstream of WNT receptors (Zhang et al., 2012). This membrane‐tethered enzyme functions by cleaving WNT on its N‐terminal. This cleavage of WNT leads to loss of its activity most likely due to WNT oxidation and oligomerization. Both subtypes of TIKI that are found in humans (TIKI1 or TIKI2) can eliminate WNT function in WNT‐producing and WNT‐responding cells. Based on sequence similarity, TIKI has been suggested to be related to metalloproteases that are found in bacteria (Bazan et al., 2013) although others claim TIKI might actually be the first member of a new enzyme family (Sanchez‐Pulido and Ponting, 2013). In vitro, TIKI cleaves 10 of the 19 known human WNTs, but there seems to be no consensus motif between the cleavage sites (Zhang et al., 2016).

Notum

Discovered in Drosophila as a WNT pathway‐specific inhibitor, Notum was originally suggested to function as a phospholipase; cleaving the membrane‐anchor of glypicans (large polysaccharide molecules that form complexes with, inter alia, WNTs). It was proposed that in this way Notum decreases the tethering of WNT to the membrane reducing its signalling potency (Gerlitz and Basler, 2002; Giráldez et al., 2002). Notum, which is a WNT target itself, would thus function in a negative feedback loop to regulate WNT signalling. However, glypicans serve as a membrane scaffold for all sorts of signalling molecules such as Fgf, Hedgehog ligand and TGF‐β. This finding did not agree with the WNT pathway‐specific effect of Notum and suggested that it must serve as a WNT inhibitor in a different way. In 2015, it was revealed that Notum indeed does not cleave the membrane anchor of glypicans but rather acts as a de‐acylase. It interacts with glypicans to obtain close proximity to WNT and, subsequently, cleaves WNT from its palmitoylation moiety (Kakugawa et al., 2015). As this modification on WNTs serves as a docking site for FZD, its enzymatic cleavage prevents interaction of WNT with FZD and, thereby, inhibits WNT signalling. To date, Notum is the first extracellular protein de‐acylase ever described and its specificity makes it an interesting target for therapy. Indeed, it has been shown that Notum is overexpressed in colorectal cancers and hepatocellular carcinoma (De Robertis et al., 2015; Torisu et al., 2008). Additionally, Notum levels might serve as a biomarker for the effect of WNT inhibitory therapies, as it is a secreted protein and its levels in blood correlate with tumour growth (Madan et al., 2016). Figure 4 gives an overview of the previously discussed inhibitors of WNT signalling that engage either on WNT itself, or LRP5/6 or FZD.

Figure 4.

Inhibitors of the FZD/LRP/Wnt signalling cascade. This figure gives an overview of the inhibitors of LRP, FZD and Wnt that are discussed in this review. SOST, Wise and DKks interact with LRP and prevent its interaction with Wnt, thus preventing signal transduction. Inhibition of Wnt ligand can occur via WIF‐1, an inhibitor molecule that binds to Wnt and prevents its interaction with LRP/FZD. Furthermore, Wnt can be inactivated by the membrane‐tethered protease TIKI and extracellular de‐acylase Notum. sFRPs can function as Wnt signalling inhibitors by interacting with both Wnt and FZD. In addition, intracellular mi‐RNAs are known to interact with FZD‐coding mRNA to prevent translation, thereby decreasing FZD levels on the membrane. In addition to endogenous Wnt pathway inhibitors, two antibodies (OMP‐54F28 targeting Wnt and OMP‐18R5 targeting FZD) are being currently explored as therapeutic agents.

The RSPO/RNF43/Lgr5 module

Only 5 years ago, an unexpected, new level of WNT regulation was discovered that involved RSPO, the leucine‐rich repeat‐containing GPCR (Lgr) and E3 ligases. RSPOs were found to be a family of proteins that increase WNT signalling (Kazanskaya et al., 2004) and up‐regulate FZD on the cell membrane. Shortly after, it was shown that the stem cell marker Lgr5 and its relatives Lgr4 and ‐6 serve as a receptor for RSPO and that they are essential for the effect of RSPO on WNT signalling (Carmon et al., 2011; Glinka et al., 2011). Later research revealed that the E3 ligases RNF43 and ZNRF3 represented a missing link that explains the effect of RSPO on membrane FZD levels (Hao et al., 2012; Koo et al., 2012).

RSPOs are members of a protein superfamily characterized by the presence of thrombospondin type 1 repeats domains. A clear description of their origin, protein structure and involvement in human disease can be found in the study of Lau et al. (2012). RSPOs were first linked to the WNT pathway in 2004 when it was shown that RSPO enhanced WNT signalling in early frog embryos (Kazanskaya et al., 2004). The fact that RSPO genes were found, similar to WNT genes, as preferential integration sites for mammary tumour virus, further supported these data (Lowther et al., 2005).

Shortly after, it was shown that the stem cell markers Lgr4, Lgr5 and Lgr6 serve as receptors for RSPO (Carmon et al., 2011; de Lau et al., 2011; Glinka et al., 2011) and it was not much later that the tumour suppressors RNF43 and ZNRF3 were identified as essential components of the WNT‐amplifying effect of RSPO (Hao et al., 2012; Koo et al., 2012). Two independent studies demonstrated that, in the absence of RSPO, the E3 ligases RNF43 and ZNRF3 cause FZD to be ubiquinated and internalized for lyzosomal degradation (Hao et al., 2012; Koo et al., 2012). However, when RSPO binds to its receptor Lgr, this prevents the E3 ligases from ubiquitinating FZD, thus, making a cell more sensitive to WNT.

In 2013, Chen et al. (2013) used crystallography to confirm that Lgr5, RNF43 and RSPO form a trimeric complex. This work which studied the interactions of protein domains, together with other structural studies (Peng et al., 2013a, b; Wang et al., 2013; Xie et al., 2013; Xu et al., 2013; Zebisch et al., 2013) validated the physical linkage of these three proteins. Additionally, they revealed that RNF43/ZNRF3 serves as the effector receptor in this complex, as Lgr5 supports the affinity of RSPO to RNF43/ZNRF3, but does not interact with these molecules itself. Only recently, the molecular mechanisms were elucidated by which RNF43 and ZNRF3 recognize the WNT receptor complex. Jiang and colleagues proposed that the protein Dishevelled (DVL) serves as a linker between FZD and the E3 ligases (Jiang et al., 2015). This complicates the role of DVL in WNT signalling, as it was originally defined as a positive regulator of the pathway.

Loss‐of‐function mutations in RNF43 and ZNRF3 have been found in a variety of human cancers such as colorectal (Giannakis et al., 2014; Shinmura et al., 2014), gastric (Wang et al., 2014), ovarian (Ryland et al., 2013; Giannakis et al., 2014) and pancreatic cancers (Wu et al., 2011; Jiang et al., 2013; Witkiewicz et al., 2015), cholangiocarcinoma (Ong et al., 2012) and adrenocortical carcinoma (Assie et al., 2014). RSPO overexpression via RSPO gene translocations have been observed in colon cancer (Seshagiri et al., 2012), prostate cancer (Robinson et al., 2015) and Schwann cell tumours (Watson et al., 2013). Indeed, treatment of these tumours with WNT‐inhibitory drugs such as porcupine inhibitors decreased tumour growth (Jiang et al., 2013; Madan and Virshup, 2015). Translocation‐independent overexpression of RSPOs in cancers has also been reported and patient‐derived xenografts responded to anti‐RSPO antibodies in vivo (Chartier et al., 2016). Thus, the cancers in this paragraph are particularly interesting targets for treatment by well‐defined extracellular WNT signalling inhibitors. Figure 5 summarizes the effects of the RSPO/RNF43/Lgr5 module on WNT signalling.

Figure 5.

The R‐spondin/Rnf43/Lgr5 module and its effects on canonical Wnt signalling. When extracellular RSPO is present, it can interact with membrane‐spanning E3 ligases RNF43/ZNRF3 and LGR5. This interaction results in internalization of the complex and consequent removal from the membrane. In the absence of RSPO, the E3 ligases ubiquinate FZD, thus marking it for degradation by the proteasome and inhibiting Wnt signalling. As RNF43 and ZNRF3 are Wnt targets themselves, this module functions in a negative feedback loop to control Wnt signalling.

Concluding remarks

We are beginning to understand the interactions between extracellular WNT, WNT modifiers and their membrane‐bound receptors. An increase in the number of clinical trials that begin to explore the therapeutic effect of antibodies against these molecules supports a rapid translation to the clinic. So far, most of these clinical trials have focused on diseases other than cancer. However, many of the targets of these antibodies are also deregulated in cancer suggesting that these compounds might prove effective as cancer therapy. This research, together with the development of new and better low MW compounds or antibodies will, we hope, help to translate current knowledge on WNT regulation at the membrane to the clinic and aid treatment of patients. The multitude of regulatory mechanisms that have evolved to modify the WNT pathway makes clear that negative feedback loops where the inhibitors of the WNT pathways are actually WNT targets themselves (e.g. Notum, RNF43, Lgr5) constitute an important part of the mechanisms. Such regulatory loops are widespread throughout developmental biology and serve to fine‐tune highly complex biological processes. This ‐together with the diverse roles of the WNT signalling in development and homeostasis and its involvement in human disease‐ underscores the importance of a fast‐responding, subtle and extensive regulatory network for WNT signalling. Given the discoveries on WNT regulation of the past years, it is likely that new studies will reveal even more regulatory molecules or pathways. This will not only allow for a better understanding of this intriguing and exciting signalling system but will also increase the window of opportunities for the therapeutic modification of the WNT pathway.

Conflict of interest

The authors declare no conflicts of interest.

Driehuis, E. , and Clevers, H. (2017) WNT signalling events near the cell membrane and their pharmacological targeting for the treatment of cancer. British Journal of Pharmacology, 174: 4547–4563. doi: 10.1111/bph.13758.