Regulation of Wnt Signaling by Protocadherins

Abstract

The ~70 protocadherins comprise the largest group within the cadherin superfamily. Their diversity, the complexity of the mechanisms through which their genes are regulated, and their many critical functions in nervous system development have engendered a growing interest in elucidating the intracellular signaling pathways through which they act. Recently, multiple protocadherins across several subfamilies have been implicated as modulators of Wnt signaling pathways, and through this as potential tumor suppressors. Here, we review the extant data on the regulation by protocadherins of Wnt signaling pathways and components, and highlight some key unanswered questions that could shape future research.

1. Introduction

The ~70 protocadherins (Pcdhs) have, since the turn of the 21st century, emerged as some of the most interesting regulators of neural development. Pcdhs make up the largest group within the broader cadherin superfamily of cell adhesion molecules, which also includes the canonical classical cadherins, the seven-transmembrane domain cadherins, and atypical cadherins such as Fat and Dachsous (for more information, see the other reviews in this special issue). Functional studies have implicated numerous Pcdhs in the regulation of neuronal survival, axon outgrowth and targeting, dendrite arbor complexity, the self-avoidance of sister axon and dendrite branches, and synaptogenesis. Several Pcdh genes also have been implicated, either by mutation or epigenetic dysregulation, in a wide variety of neurological and neurodevelopmental disorders, including epilepsy, mood disorders, autism, and schizophrenia (see reviews by Peek, et al. [1] and Keeler et al. [2], as well as the other reviews in this special issue).

Despite this progress in identifying functional roles for many Pcdhs, less is understood about the intracellular signaling pathways with which they engage in order to play these roles. Recently, we and others have uncovered interactions between Pcdh molecules and Wnt signaling pathways, which are known to be critical for embryonic development in general, and neural development in particular, as well as in the etiology and progression of multiple types of cancer. In this brief review, we will summarize the studies that have begun to elucidate links between Pcdh adhesion molecules and Wnt signaling components.

2. Wnt Signaling Pathways

We will begin by providing a brief overview of Wnt signaling pathways; multiple aspects of these complicated pathways are discussed more fully in a number of recent reviews [3–10].

2.1 Wnt proteins and their receptors

Wnts are cysteine-rich proteins that are evolutionarily conserved; humans and mice have 19 Wnt ligands [11]. Originally named Int-1, the mouse Wnt1 gene was identified in the early 1980s by Nusse and Varmus as a preferential integration site of mouse mammary tumor virus (MMTV), an oncogenic retrovirus [12]. Subsequently, it was determined that Drosophila wingless (wg), which plays a role in segment polarity during larval development, is a homolog of Wnt1 [13]. Wnts possess an N-terminal signal peptide sequence for secretion and are subject to a number of protein modifications, most prominently glycosylation and acylation, both of which are crucial for their function [14–18]. While Wnts have been studied extensively, to date crystal structures of only 2 Wnts, XWnt8 (Xenopus) and WntD (Drosophila), have been elucidated [19, 20].

The main Wnt receptors are Frizzleds (Fzd), a family of seven-transmembrane G-protein-coupled receptors that possess a large extracellular cysteine-rich domain that mediates Wnt binding [20, 21]. Mammals have 10 Fzd receptors [22]. The intracellular domain (ICD) of Fzd binds Dishevelled proteins (Dvls) through a conserved KTXXXW motif; Dvl interacts with a large number of Wnt co-receptors to activate the different Wnt pathways [23]. Fzds have been found to be phosphorylated at the ICD, an event that downregulates their functions [24]. Furthermore, ubiquitination and de-ubiquitination of Fzds are important for the regulation of events downstream of Wnt-Fzd binding [25].

Two highly homologous proteins, Lrp5 and Lrp6 in vertebrates, as well as their Drosophila homolog Arrow, act as Wnt co-receptors with Fzds to initiate canonical Wnt signaling (see below, 2.2). Lrp6 remains the best studied Lrp (Low-density lipoprotein-related rececptor) and is a large, single transmembrane domain protein with greater affinity to Wnts complexed with Fzd than to Wnts on its own. The extracellular domain of Lrp6 contains many independent Wnt-binding sites, allowing simultaneous interaction with many Wnt-Fzd complexes [26]. A key event in the regulation of Lrp6 function is the phosphorylation of several sites within its ICD, with the first event being phosphorylation of the 5 PPPSP repeats by several proline targeted kinases; this then primes Lrp6 for a second phosphorylation event by the casein kinase I (CKI) family at neighboring Ser residues (PPPSPXS) [27–29].

Additionally, ROR1 and ROR2 (Receptor tyrosine kinase-like Orphan Receptor) have been shown to be co-receptors for Wnt5a to facilitate planar-cell polarity (PCP) signaling in verterbrates [30, 31] (see below, 2.2). The binding of Wnt5a leads to the homodimerization of ROR2, which together forms a ternary complex with Fzd [32, 33]. This leads to the recruitment of the actin-binding protein filamin A and activation of c-Jun N-terminal kinase (JNK) [34]. ROR receptors are also phosphorylated, in a Wnt5a-dependent manner, to activate PCP signaling [32, 35]. A final Wnt-binding co-receptor is the receptor tyrosine kinase Ryk, which has been implicated in multiple Wnt signaling pathways [36–38].

2.2 Canonical, PCP, and Wnt/Ca²⁺ pathways

The binding of secreted Wnts to Fzd, Lrp5/6, Ryk, and ROR co-receptors has been shown to activate at least three distinct pathways: the “canonical” (β-catenin-dependent) pathway, the Wnt/PCP pathway, and the Wnt/Ca²⁺ pathway. The canonical Wnt/β-catenin pathway, the one that is best characterized molecularly, is dependent on the proteolytic state of cytoplasmic β-catenin (Figure 1). In the Wnt-OFF state, a large proportion of β-catenin is found at the plasma membrane, complexed with classical cadherin cytoplasmic domains, where it regulates cell adhesion at adherens junctions and synapses via its partner α-catenin’s ability to bind actin filaments [39]. In the Wnt-OFF state, cytoplasmic β-catenin levels are kept low by the constitutive activity of a “β-catenin destruction complex” consisting of Axin1, adenomatous polyposis coli (APC), CK1α, and glycogen synthase kinase 3β (GSK3β; reviewed by Clevers and Nusse [5]; Nusse and Clevers [8]). Axin1 serves as a scaffold for the destruction complex, and associated CK1α and GSK3β sequentially phosphorylate β-catenin, targeting it for ubiquitin-dependent degradation by the proteasome [40, 41].

Figure 1. Wnt signaling pathways.

A and B: Canonical (β-catenin-dependent) pathway. In the “Wnt-OFF” state (A), β-catenin levels are kept low in the cytoplasm by the action of the “destruction complex”. GSK3β phosphorylates β-catenin, which targets it for desctruction by the proteasome. Wnt binding to Frizzled and Lrp5/6 co-receptors results in disruption of the destruction complex, allowing β-catenin to accumulate in the cytoplasm. β-catenin can then translocate to the nucleus and promote the activation of Wnt target genes by displacing co-repressors, and recruiting co-activators, of TCF/Lef. C: Wnt/PCP pathway. Binding of Wnt to Frizzled and ROR or Ryk co-receptors leads to recruitment of Dvl, which can act through Rac1 or Daam1 to initate changes in cytoskeletal dynamics important for cell orientation and movement. D: Wnt/Ca²⁺ pathway. Wnt binding and recruitment of Dvl leads to activation of PLC, which cleaves PIP₂ to generate IP₃ and DAG. This leads to release of Ca²⁺ from intracellular stores, and downstream signaling through a number of Ca²⁺-dependent kinases and phosphatases.

When Wnt ligands bind to Fzd and Lrp5/6, the Wnt-ON state is triggered. This results in Dvl binding to the C-terminus of Fzd [42], which recruits Axin1 to the cytoplasmic tail of Lrp5/6, facilitating the phosphorylation of Lrp5/6 by GSK3β and CK1α [27–29, 43, 44]. The fully phosphorylated Lrp6 can then bind Axin1, which recruits the β-catenin destruction complex. The activity of GSK3β in the destruction complex is inhibited by phosphorylated Lrp5/6, resulting in decreased phosphorylation of both β-catenin and Axin1 [45–47]. Dephosphorylated Axin1 is dissociated from the receptor complex and also from β-catenin, which inactivates the destruction complex until Axin1 is again phosphorylated [48, 49]. This cascade of events causes β-catenin to accumulate in the cytoplasm and to translocate to the nucleus, where it interacts with members of the T-cell factor (TCF)/lymphoid enhancer factor (Lef) family of transcription factors to activate a wide variety of Wnt target genes [50] (Figure 1).

A separate Wnt-receptor interaction outcome that is β-catenin independent is the Wnt/PCP pathway. In this pathway, Wnt-Fzd binding, along with its coreceptors ROR or Ryk, recruits Dvl, which can: 1) form a complex with Dishevelled-associated activator of morphogenesis 1 (Daam1) to activate RhoA; or 2) activate JNK through Rac1 to affect cytoskeletal dynamics and cell polarity [7]. The PCP pathway is critical in establishing cell polarity in morphogenetic processes such as the regulation of cell movements during gastrulation, neural tube closure and the orientation of stereocilia in the inner ear in vertebrates [23, 51]. Finally, a distinct Wnt/Ca²⁺ pathway has been implicated in cancer, inflammation and neurodegeneration, as well as a variety of critical events in embryonic development (reviewed by Slusarski and Peligri [9]; De [10]). In this pathway, Wnt-Fzd binding activates phospholipase C (PLC) in a Dvl-dependent manner. This leads to a release of Ca²⁺ from intracellular stores as PLC hydrolyzes PIP₂ to form IP₃/DAG, which then activates calmodulin-dependent kinase II (CaMKII), protein kinase C (PKC) and calcineurin [10]. CaMKII activates TAK, which promotes Nemo-like kinase (NLK) activity to inhibit Wnt/β-catenin transcriptional activity [52]; thus, the Wnt/Ca²⁺ pathway can antagonize the canonical Wnt pathway. PKC phosphorylates the small GTPase Cdc42, a major mediator of actin cytoskeleton reorganization [53], while calcineurin dephosphorylates nuclear factor of activated T cell (NFAT), promoting its translocation to the nucleus to upregulate genes controlling cell fate and cell migration [54–57].

Wnt signaling pathways have long been associated with cancer and carcinogenesis. The link between Wnt and cancer was established with their initial discovery, as enhanced expression of the int1 (Wnt1) gene due to MMTV insertion caused mammary hyperplasia and tumors in mice [12, 58, 59]. Subsequently, two studies identified mutations in the APC gene (whose gene product, as mentioned above, interacts with β-catenin as part of the destruction complex) as the underlying cause of hereditary colon cancer syndrome [60, 61]. Others have identified mutations of other components of Wnt signaling pathways in various cancers: for example, β-catenin mutation in gastric cancer [62], Axin1 mutations in hepatocellular carcinomas and medulloblastomas [63, 64], and β-catenin, Axin1, Axin2, and TCF4 mutations in colon cancer [65]. While the canonical Wnt pathway remains best defined in its role in cancer, the relatively more diverse and less understood non-canonical Wnt pathways have also been implicated in metastasis formation and cell migration of cancer cells [66, 67]. Other reviews can be consulted for insights on the misregulation of canonical and non-canonical Wnt signaling in cancer cells [68–70].

Since the initial discovery of Wnt signaling’s role in tumorigenesis, we have learned that Wnt signaling plays a vast array of roles in embryonic development, including cell fate determination, early patterning events, and organ morphogenesis, as well as in the adult, where Wnt signaling regulates stem cell renewal and tissue homeostasis [71–77]. The variety of intracellular pathways, some of which are mutually antagonistic, downstream of Wnt-Fzd binding allows for a wide range of cellular outcomes mediated by common signals. Understanding the signaling partners, such as Pcdhs, that can influence signaling downstream of Wnts will be important for understanding the many roles they play.

3. Protocadherins

3.1 Pcdh gene and protein structure

In this review, we will focus only on the “true” protocadherins (Pcdhs): the ~60 members of the clustered α-, β-, and γ-Pcdh families and the 10 non-clustered δ-Pcdhs. Other “atypical” cadherin molecules (sometimes referred to colloquially as Pcdhs), including the seven-transmembrane Flamingo/Celsr family and the large Fat and Dachsous cadherins, have been implicated extensively in the control of PCP pathways. Flamingo, in particular, is known to interact physically and functionally with Frizzled receptors in order to regulate PCP [78]. Thorough discussions of the seven-transmembrane cadherins and the Fat cadherins can be found, respectively, in the articles by Goffinet and Tissir (reference to be added later in revision) and by Avilés and Goodrich [79] in this special issue. Here, we will confine our discussion to the emerging roles for clustered and non-clustered Pcdhs in the regulation of the canonical (primarily) and non-canonical Wnt pathways.

Pcdhs represent the largest group within the cadherin superfamily of cell adhesion molecules known to play critical roles in several biological processes, including embryonic morphogenesis, neural circuit formation, angiogenesis, and cancer [80–82]. Members of the cadherin superfamily are characterized by extracellular cadherin (EC) motifs that are approximately 100 amino acids long and that mediate trans-interactions between cells. Shintaro Suzuki and colleagues used degenerate PCR to search for additional cadherin-related molecules, and discovered and named the first Pcdhs (including one of the clustered Pcdhs and one of the δ1-Pcdhs) in the early 1990s [83]. Like “classical” cadherins, Pcdhs are type I transmembrane proteins; however, unlike cadherins which have five EC domains, Pcdhs have six (clustered Pcdhs, δ2-Pcdhs) or seven (δ1-Pcdhs) EC domains, and distinct cytoplasmic domains that lack catenin-binding sites, allowing for integration into distinct signaling pathways [83–94].

The clustered Pcdhs consist of ~60 proteins, termed α-, β-, and γ-Pcdhs, that are encoded by three tandem gene clusters (Pcdha, Pcdhb, and Pcdhg) encompassing about 1 MB at human chromosome 5q31 and mouse chromosome 18 [93, 95, 96]. Eight of these clustered Pcdh genes had been identified previously as “Cadherin-related Neuronal Receptor” (CNR) proteins by Kohmura et al. [97]. Each large “variable” exon (expressed from its own promoter) encodes six EC domains, a transmembrane domain and a variable cytoplasmic domain of approximately 90 amino acids, and is spliced to three constant exons that encode a shared ~125 amino acid C-terminal domain for the Pcdha and Pcdhg clusters. The Pcdhb cluster does not contain such constant exons and thus are expressed as a single-exon transmembrane molecule [98, 99] (Figure 2). The Pcdhg cluster encodes 22 γ-Pcdh proteins that can be grouped into three subfamilies on the basis of sequence similarity, termed γ-Pcdh-A, -B, and –C; γ-Pcdh-C3, -C4 and –C5 are more similar to α-Pcdh-C1 and –C2, found within the Pcdha cluster, than they are to any other γ-Pcdhs [93].

Figure 2. The protocadherin gene clusters.

A: Schematic of the human PCDHA, PCDHB, and PCDHG gene clusters on chromosome 5q31. A very similar structure is observed for the mouse clusters at chromosome 18. B: The exon structure of the PCDHG cluster is expanded below, with an example of the transcription initiation and splicing pattern (for A6, in this instance). C: Schematic of the PCDHG spliced transcripts generated by the cluster; each mature transcript consists of one large variable exon and the three small constant exons. D: Protein structure of the γ-Pcdhs (α-Pcdhs are identical in structure; β-Pcdhs lack any constant domain). Six extracellular cadherin (EC) repeats, a transmembrane domain, and a variable cytoplasmic domain are encoded by each variable exon; the constant exons encode a 125 amino acid C-terminal domain. Stars indicate the sites of “cluster control regions”, enhancers required for normal expression patterns of the Pcdh clusters.

Many nonclustered Pcdh genes have an exon structure similar to the clustered Pcdhs: In most δ-Pcdh genes, a single large first exon encodes the N-terminal signal peptide, multiple EC domains, a single transmembrane domain, and a small part of the intracellular domain, while the remainder of the intracellular domain is encoded by several remaining exons. Exceptions are Pcdh1 and Pcdh11 (a.k.a. PcdhX/Y; a Pcdh11 gene is found on the homologous region of both the X and Y chromosomes in humans, but only on the X chromosome in other mammals; [100]), the EC domains of which are encoded by 2 exons [91]. While alternative splicing of small exons is observed in all δ1-Pcdh transcripts, this is particularly prominent within Pcdh11 and Pcdh19, resulting in numerous splice variants for these genes [91, 101–103].

3.2 Regulation of clustered Pcdh gene expression

Pcdh genes are predominantly expressed in the developing and adult nervous system [83, 91, 97, 99, 104–111], though they are also expressed at low levels in other organs such as lung and kidney [107, 108, 112, 113]. A number of studies have uncovered a complex mechanism of transcription for the Pcdha, Pcdhb, and Pcdhg clusters that results in each neuron expressing a semi-stochastic repertoire of isoforms. Each variable exon has its own promoter region that includes a ~20 base pair conserved sequence element (CSE) required for expression. A variable exon promoter is “chosen” for activation by a DNA looping mechanism involving enhancers outside of the clusters, in a manner dependent on the transcription factor CTCF, cohesin, and the methyltransferase SETDB1 ([114–119], reviewed by Hirayama and Yagi [120] in this special issue). A long transcript through the remainder of the cluster, including all downstream variable exons as well as (for the Pcdha and Pcdhg clusters) the three constant exons, is generated. The 5’ variable exon is cis-spliced to the three downstream constant exons to generate a mature Pcdha or Pcdhg transcript (note that Pcdhb transcripts include only a single variable exon; [98, 99]. Based on single-cell RT-PCR studies, it is believed that the majority of Pcdha, Pcdhb, and Pcdhg variable exons are monoallelically expressed, while PcdhaC1, PcdhC2, PcdhgC3, PcdhgC4, and PcdhgC5, the C isoforms of α-Pcdh and γ-Pcdh, are biallelically expressed. Both clusters are transcriptionally active in any given cell, but a particular variable exon promoter is “chosen” from only one of the two alleles. This produces a stochastic expression pattern for most of the α-, β-, and γ-Pcdh family members, while the C isoforms are ubiquitously expressed [121–123]. Single-cell RT-PCR analysis of cerebellar Purkinje neurons found that each cell expressed approximately 4 α-Pcdh isoforms, 2 β-Pcdh isoforms, and 7 γ-Pcdh isoforms [124, 125].

It has been found that this differential expression of clustered Pcdh genes is controlled by epigenetic modifications, specifically silencing by methylation ([98, 126–128]; reviewed by Hirayama and Yagi [120], in this special issue). Work from Takeshi Yagi and colleagues, in particular, has demonstrated that the degree of methylation of Pcdha promoters, as well as the 5’ regions of each variable exon, was negatively correlated with that exon’s expression level; PcdhaC1 and PcdhaC2, both of which are ubiquitously expressed, possess hypomethylated promoters [127, 129]. Demethylation through the application of 5-azacytidine was sufficient to increase the transcription of Pcdha genes, and experimental hypermethylation of a promoter repressed its transcriptional activity [127]. In lines of transgenic mice harboring deletions or duplications of exons within the Pcdha cluster, a decrease in methylation of any variable exon situated at the 3’ end of the cluster was observed. These hypomethylated exons subsequently were found to be ubiquitously expressed, while duplicated copies of the normally ubiquitous PcdhaC1 and PcdhaC2 became hypermethylated and stochastically expressed when situated farther 5’ in the cluster [129, 130]. Work from the same laboratory identified Dnmt3b as the DNA methyltransferase responsible for regulating methylation patterns of stochastically expressed Pcdh isoforms in neural cells at early embryonic stages [131]. As will be seen below (section 4), such mechanisms are relevant to the epigenetic dysregulation of Pcdh genes in cancer and other disorders (reviewed by El Hajj et al., [132], in this special issue).

3.3 Pcdh functions and signaling

Most studies on Pcdhs have focused on their roles in neurodevelopment. The clustered Pcdh proteins have been detected on axons, in dendrites, and in some, but far from all, synapses [97, 109, 113, 133–139]. These observations are consistent with functions revealed by analyses of Pcdha or Pcdhg mutant mice, which indicated that α- and γ-Pcdhs play critical roles in neuronal survival, axon and dendrite arborization, self-avoidance, and tiling, and synaptogenesis, depending on the neuronal subtype examined [109, 137, 140–152]. Comparison of mice lacking all three Pcdh gene clusters with those lacking only the Pcdha, Pcdhb, or Pcdhg clusters indicates overlapping and synergistic functions for the clustered Pcdh families, though the most severe phenotypes are generally attributable to loss of the γ-Pcdhs [153, 154]. Individual non-clustered δ-Pcdhs are expressed by discrete neuronal subsets throughout the brain, and functional studies collectively have revealed roles in axon outgrowth and pathfinding, synaptic plasticity, and synapse elimination [84, 86, 110, 155–163]. Though not the focus of this review, the many roles of Pcdhs in the nervous system have been discussed extensively in several recent reviews ([1, 2, 164–166]; from this special issue, see Light and Jontes, 2017; El Hajj et al., [132]; Aviles and Lefebvre, [79], Rubinstein et al., 2017; Phillips et al., [167]; Hirayama and Yagi, [120]).

Pcdhs have been found to engage with a number of intracellular signaling partners of potential relevance to multiple Wnt pathways. The α- and γ-Pcdhs interact with focal adhesion kinase (FAK) and Pyk2 (also known as FAK2) via their respective constant regions [94]. Interaction with these Pcdhs inhibits activity of these kinases by suppressing their autophosphorylation, which is the first step for kinase activation [94]. Consistent with this, analyses of Pcdhg mutant cerebral cortex revealed hyperactivation of a pathway including PLC, FAK and protein kinase C (PKC) [85]. This report is bolstered by concurrent work showing hyperactivation of Pyk2 and FAK, and reduced activity of both Rac1 and RhoA in animals with a deletion of the Pcdha gene cluster [168]. Recent work identified a serine residue within the γ-Pcdh constant domain that is phosphorylated by PKC in vitro and in vivo; this phosphorylation event reduces inhibition of FAK by γ-Pcdhs [87]. Together, these studies provide the strongest evidence thus far for a signaling pathway downstream of the clustered Pcdhs; notably, PLC, PKC, Rac1 and RhoA are all of relevance to multiple Wnt signaling pathways (Figure 1).

Yeast two-hybrid experiments revealed an interaction of the intracellular domain of Pcdh7 with protein phosphatase 1 alpha (PP1α), a protein implicated in synaptic plasticity [169, 170]. It was later determined that this interaction is conserved for all δ1-Pcdhs, facilitated through the CM3 motif, which is absent in δ2-Pcdhs [91]. Several δ2-Pcdhs (Pcdh10, Pcdh17, Pcdh18b, Pcdh19), on the other hand, have been found to interact with Nap1, a component of the WAVE complex, through the WRC interacting receptor sequence (WIRS) [84, 86, 88, 90, 171]. The WAVE protein complex comprises WAVE1, Cyfip1, Abi2, Nap1 and HSPC300, and is activated by Rac1 and Arf GTPases. The WAVE complex interacts with the Arp2/3 complex to promote actin assembly [172]. The WIRS, a conserved motif, is not exclusive to δ2-Pcdhs, but is also present in α-Pcdhs and Pcdh9, as well as many other adhesion molecules [171]. Note that these pathways leading to actin rearrangements may have several points of contact with components of Wnt/PCP signaling as well.

4. Regulation of Wnt signaling by Pcdhs

4.1 Epigenetic dysregulation of Pcdhs in cancer

Considering that Pcdhs are thought of primarily as neuronal cell adhesion molecules, and neurons are intrinsically postmitotic and terminally differentiated, it is perhaps surprising that several studies have also implicated Pcdhs in many types of cancer. Several groups have reported silencing of δ-Pcdh expression, due to promoter hypermethylation, in many primary tumors or cell lines: Pcdh1 in breast cancer [173]; Pcdh7 in bladder cancer [174]; Pcdh8 in renal cell carcinoma [175], nonmuscle invasive bladder [176], hematologic [177] and breast [178] cancers; Pcdh9 in glioblastoma [179] and hepatocellular carcinoma [180]; Pcdh10 in breast [181], lung, nasopharyngeal, esophageal [182], hepatocellular carcinoma [182, 183], hematologic [184], colorectal, pancreatic [185], gastric [185, 186], cervical [187, 188], prostate [189], and testicular [190] cancers; Pcdh17 in laryngeal and esophageal squamous cell carcinoma [191, 192], urological [193, 194], gastric, and colorectal cancers [195]; and Pcdh20 in non-small-cell lung [196] and hepatocellular carcinoma [197]. A smaller number of studies has implicated dysregulation of the clustered Pcdh genes in cancer as well. A microarray-based methylation study of astrocytomas revealed that Pcdhga11 is hypermethylated in these cells, which resulted in decreased transcription. Transcript levels of Pcdhga11 were restored when the authors treated these astrocytomas with a demethylating agent [198]. The Pcdhb gene cluster has been associated with “CpG island methylator phenotype”, a term used to describe concordant methylation of multiple loci in various cancers [199], as the pattern of methylation within the cluster was able to distinguish two groups of neuroblastoma patients at opposite ends of the International Neuroblastoma Risk Group classification system [200]. Other studies have found clustered Pcdhs to be differentially methylated in prostate cancers, and hypermethylated in breast cancers [201, 202]. There is now increasing evidence that Pcdhs are potential tumor suppressor genes, as reexpression of Pcdh8, Pcdh10, Pcdh17 and Pcdh20 suppresses tumor cell proliferation, inhibits cell migration, and induces apoptosis and autophagy in cancer cell lines [178, 192, 195, 196].

4.2 Clustered Pcdhs

The well-established role of Wnt signaling in tumorigenecity thus suggests that Pcdhs could act as tumor suppressor genes via Wnt pathway regulation. A growing body of literature has, in fact, begun to identify mechanisms by which Pcdhs can regulate Wnt signaling (Figure 3). A genome-wide analysis of promoter methylation in Wilm’s tumor (WiT), a pediatric kidney cancer, identified a region spanning 800 kilobases at chromosome 5q31 that was hypermethylated; this is the region containing the ~60 genes of the Pcdh gene clusters [112]. Consistent with this hypermethylation, the authors demonstrated silencing of Pcdhg gene expression in WiT. While there is extensive hypermethylation across the Pcdhg gene cluster in WiT, some individual Pcdh genes (such as Pcdhga6 and Pcdhgc3) are not hypermethylated and remain expressed; therefore, the authors performed a knockdown of all Pcdhg genes using an siRNA targeting the constant exons in a WiT cell line. This led to an increase in β-catenin/TCF/Lef reporter gene activity and a corresponding increase in expression of target genes in the Wnt signaling pathway [112]. In addition, overexpression of individual γ-Pcdh isoforms in HEK293 and WiT cell lines led to a decrease in Wnt signaling activity and inhibition of colony formation and tumor cell growth in vitro [112]. This group subsequently expanded on this study to show that long range epigenetic silencing of the Pcdhg cluster is also observed in colorectal cancer, and is associated with the early stages of colorectal tumorigenesis [203]. Focusing on the most abundantly expressed γ-Pcdh isoform in the colon, Dallosso et al. [203] also observed that Pcdhgc3 expression is silenced in colorectal cancer cells. Overexpression of γ-Pcdh-C3 in human colon cancer cell lines increased apoptosis, inhibited growth, and led to a reduction of β-catenin/TCF/Lef reporter activity as well as a decrease in the levels of endogenous “active” β-catenin. The authors go on to implicate the mTOR pathway, as siRNA knockdown of Pcdhgc3 led to an increase in phosphophorylation of mTOR at serine residue 2448 (S2448) as well as an increase in total mTOR levels as assayed by Western blot; the opposite effects on mTOR were observed when Pcdhgc3 was overexpressed [203]. Together, this work provided the first firm evidence that clustered Pcdhs could regulate tumor cell behavior by inhibiting canonical Wnt signaling.

Figure 3. Regulation of Wnt pathways by Pcdhs.

Summary of results implicating γ-Pcdhs (left) and δ-Pcdhs (right) in the regulation of Wnt signaling, as discussed in the main text. The γ-Pcdh-C3 isoform, and the δ-Pcdhs Pcdh8, 9, 10, and 17 have been reported to suppress Wnt-induced expression of target genes, while some other γ-Pcdh isoforms and the δ1 protein Pcdh11 have been reported to have the opposite effect (top). Some Pcdhs have known cytoplasmic interactors that have been shown to, or that potentially could, impinge upon Wnt signaling pathways (bottom). Pcdh8 has been shown to interact with Frizzeld7 to promote Wnt/PCP signaling (far right). Long lines with a short perpendicular line indicate inhibition, while arrows indicate activation. Dashed lines indicate possible signaling connections, based on the literature, that remain to be demonstrated directly.

We subsequently sought to elucidate the specificity of Wnt pathway suppression by γ-Pcdh isoforms, and to identify the molecular mechanisms through which it is achieved [204]. Using the TOPFLASH assay, in which HEK293 cells transfected with reporter constructs that yield a quantifiable luciferase signal upon exposure to Wnt3a, as well as quantitative PCR (qPCR) for Wnt target genes, we confirmed that the γ-Pcdh-C3 isoform, specifically, inhibits the canonical transcriptional pathway. Surprisingly, however, we found that 13 other γ-Pcdh isoforms can actually potentiate β-catenin/TCF/Lef luciferase reporter activity in response to Wnt3a. We determined that the variable cytoplasmic domain (VCD), unique to each γ-Pcdh isoform, is important in this regulation of Wnt signaling: expression of constructs encoding only the VCD of C3 or A1 isoforms was sufficient to, respectively, suppress or potentiate Wnt signaling [204]. We identified Axin1, a key component of the destruction complex, as an evolutionarily-conserved physical interactor of the γ-Pcdh-C3 VCD, and showed that the C3 VCD competes with Dvl for binding to the DIX domain of Axin1. This interaction stabilized Axin1 at the membrane, and reduced phosphorylation of Lrp6 [204].

We were also able to confirm that β-catenin/TCF/Lef reporter activity can be modulated up (by overexpression of γ-Pcdh-A1) or down (by overexpression of γ-Pcdh-C3) in the mouse cerebral cortex in vivo, using conditional transgenic alleles [204]. Our data suggest a novel mechanism in which the interaction of γ-Pcdh-C3 with Axin1 can potentially sequester it away from other Wnt signaling components, which leads to Lrp6 hypophosphorylation and reduced Wnt target gene expression through an as-yet undetermined pathway. It is likely that γ-Pcdhs act as modulators or buffers for Wnt signaling activity, rather than as primary co-factors, as a grossly normal cerebral cortex still forms in mice overexpressing γ-Pcdh-C3 or –A1 [204]; major disruption of Wnt signaling in embryonic telencephalon (e.g., through ablation of the Wnt-producing cortical hem; [205]) leads to a severe disruption of cortical development. As it stands, it will be important to determine if any of the many roles that γ-Pcdhs play in the brain are dependent on its modulation of Wnt pathways. In this respect, the interaction of the C3 VCD with Axin1 is particularly interesting: both Axin1 [206] and the γ-Pcdhs [85, 147, 168] are required for complex dendrite arborization in cortical or hippocampal neurons.

4.3 Non-clustered Pcdhs

Like some γ-Pcdhs, a number of non-clustered δ-Pcdhs have been implicated as potential tumor suppressors through regulation of Wnt signaling pathways (Figure 3). Reduction of Pcdh10 expression due to promoter hypermethylation has been reported in several tumors [207], while its experimental reexpression inhibits cell growth, decreases colony formation, prevents cell invasion and promotes cell apoptosis, all of which indicate a possible role for Pcdh10 as a tumor suppressor. It appears that Pcdh10 can act through the canonical Wnt/β-catenin pathway, as Zhao et al. [208] reported that the expression of Pcdh10 negatively regulated Wnt transcriptional response, as shown using TOPFLASH and qPCR. One Wnt target gene that was shown to be suppressed is a long noncoding RNA called MALAT1; overexpression of Pcdh10 decreased β-catenin binding at MALAT1’s promoter through an unknown mechanism [208]. Another study implicated Pcdh10 in the negative regulation of Wnt/β-catenin signaling [209]. Pcdh10 overexpression was found to inhibit myeloma cell proliferation, even when cells were treated with LiCl to activate Wnt signaling [209]. Pcdh10 negatively regulated a canonical Wnt pathway, as assayed by TOPFLASH and qPCR for several Wnt target genes (Xu et al., 2015); additionally, the expression of several Wnt pathway proteins, including β-catenin, was reduced while GSK3β was upregulated [209]. Lastly, the authors observed a decrease in the expression of B-cell CLL/lymphoma 9 (BCL-9), a coactivator of β-catenin, in myeloma cell lines that overexpress Pcdh10, confirming the negative regulation of the Wnt/β-catenin signaling by Pcdh10 [209].

Pcdh8 (for simplicity, we will use this gene name here regardless of organism; original papers reported this gene as paraxial protocadherin (PAPC) in amphibians [210] and arcadlin in rats [161]) has been intensively studied for its roles in development mediated through the Wnt/PCP pathway. Pcdh8 was first discovered as a gene expressed in the Spemann organizer of Xenopus, and later in the paraxial mesoderm during embryo gastrulation [210]. Pcdh8 expression was found to be upregulated as a result of signaling through a non-canonical Wnt5a/ROR2-JNK (PCP) pathway in early Xenopus development [211]. The overexpression of Pcdh8 RNA was sufficient to trigger gastrulation movements in Xenopus animal cap explants [210], and Pcdh8 has also been shown to coordinate tissue separation and convergent extension in Xenopus development by acting through the Wnt/PCP pathway to simultaneously activate JNK via Rho A while inactivating Rac1 [212, 213]. It has recently been determined that glycosylation of Wnt5a is required in vivo for Pcdh8 synthesis, further lending support to Pcdh8’s links to the Wnt/PCP pathway [214].

Much progress has been made in identifying key interactors of Pcdh8 in its regulation of the PCP pathway. Pcdh8 physically interacts with Fzd7 through its extracellular region, as determined by co-immunoprecipitation and bimolecular fluorescence complementation. This interaction regulates tissue separation in the mesoderm of Xenopus embryos [212, 215]. It has also been found that the intracellular domain of Pcdh8 recruits Sprouty, an inhibitor of convergence-extension movements, to the membrane, and antagonizes Sprouty’s ability to inhibit the PCP pathway [216]. Concurrent work revealed that Xenopus ankyrin repeats domain protein 5 (xANR5) physically interacts with Pcdh8 to promote the PCP pathway indirectly, by regulating JNK and Rho activity [217]. Jung et al. [218] discovered that the activation of RhoA signaling through the binding of Pcdh8 and Fzd7, in conjunction with the Wnt/PCP pathway, regulates invagination of the ear placode in Xenopus. Finally, Pcdh8 stability appears to be a regulation point for Wnt signaling pathways. The regulation of Pcdh8 localization and stability is determined by its phosphorylation by GSK3, and subsequent polyubiquitination by the E3 ubiquitin ligase, β-TrCP [219]. A recent study also demonstrated a physical interaction between the Pcdh8 intracellular domain and NLK, which feeds back onto the canonical β-catenin transcription pathway; this interaction stabilizes both proteins by inhibiting their ubiquitination and is required for Pcdh8’s promotion of Wnt/PCP signaling [220]. Finally, in a study by Kietzmann et al. [221], it was found that the intracellular domain of Pcdh8 interacts with casein kinase 2β (CKIIβ) and recruits it to the membrane. This prevents CKIIβ from forming a tetrameric complex consisting of 2 CKIIα and 2 CKIIβ subunits that is required for CK2 phosphorylation of β-catenin, which potentiates Wnt signaling by stabilizing β-catenin [221, 222]. This shift in CKIIβ localization mediated by Pcdh8 decreases Wnt/β-catenin signaling activity and target gene expression [221].

This general down-regulation of canonical Wnt/β-catenin-dependent signaling pathways by non-clustered Pcdhs is supported by several other studies. As noted above, Pcdh9 has been found to be downregulated in hepatocellular carcinoma (HCC) [180]. Over-expression of Pcdh9 in HCC-derived cell lines decreases migration and results in reduced phosphorylation of GSK3β at serine residue 9 (and thus presumably higher GSK3β activity); however, this did not seem to suppress the transcription of Wnt target genes [180], so a distinct pathway may be involved. Overexpression of Pcdh17 in tumor cells led to a reduction of active (that is, non-GSK3β phosphorylated) β-catenin, decreased levels of β-catenin mRNA and of proteins encoded by Wnt target genes, and a suppression of tumor growth [223]. Pcdh20 (a δ-Pcdh-related protein sometimes classified as its own δ0 subgroup) was also found to negatively regulate Wnt signaling activity in some HCC-derived cells as well as in HEK293T [197]. Pcdh20 overexpression suppresses HCC cell migration in vitro and cell growth both in vitro and in vivo; the authors posit that Pcdh20 activates GSK3β by modulating the Erk and Akt pathway [197]. This is consistent with a concurrent study demonstrating lower levels of active β-catenin, translocation of β-catenin from the nucleus to the cytoplasm and membrane, and reduced expression of Wnt target genes in cells transfected with Pcdh20 [224]. Interestingly, there is one Pcdh, Pcdh11Y, that may upregulate canonical Wnt signaling. Unusually for Pcdhs, Pcdh11Y has a β-catenin binding site localized within its COOH terminus and was found to interact with β-catenin by immunoprecipitation assay [225]. Overexpression of Pcdh11Y in human prostate or colon cancer cell lines was found to activate Wnt signaling, as measured by a TOPFlash assay and semi-quantitative PCR for Wnt target genes [226].

5. Conclusions

From the forgoing discussion, we hope it is clear that the varied groups of both clustered and non-clustered Pcdh adhesion molecules are increasingly becoming implicated in regulation of Wnt signaling pathways, particularly as this relates to cancer and morphological development of embryos. Nevertheless, this is still an emerging interaction and much remains unclear about how Pcdhs might impinge upon the multiple pathways activated downstream of Wnt binding to its receptors. In closing, we highlight a few of the questions that should be explored in future studies.

First, why do some Pcdhs suppress canonical (Wnt/β-catenin-dependent gene transcription) signaling, while others appear to potentiate it? Most reports thus far demonstrate the former rather than the latter, but this may simply reflect the small number of Pcdhs and cellular contexts that have been examined. Presumably, as we have shown for the γ-Pcdhs, a major component of this is differential interaction of Pcdh cytoplasmic domains with particular signaling partners. For example, we showed that the variable cytoplasmic domain of γ-Pcdh-C3, but not several other γ-Pcdh isoforms, competes with Dvl for binding to Axin1, leading to suppression of canonical Wnt signaling [203]. The challenge now is to identify other signaling proteins that might mediate the opposite effect on Wnt target genes that we found when overexpressing other γ-Pcdh isoforms. In addition to the variable cytoplasmic domains, which are unique to each γ-Pcdh isoform, all γ-Pcdhs share a constant cytoplasmic domain; any role for this domain in Wnt signaling remains to be determined, though its binding to, and inhibition of, FAK and Pyk2 affects signaling pathways of potential relevance (Figure 3). In the case of δ-Pcdhs, all of those thus far examined inhibit the canonical pathway except Pcdh11Y. In this respect, it is interesting to note that among the δ-Pcdhs, only Pcdh11 harbors a cytoplasmic β-catenin binding site resembling that found in classical cadherins [225]. It remains to be shown, however, whether this somehow underlies Pcdh11’s ability to upregulate the canonical pathway. Clearly, a more complete catalog of Pcdh cytoplasmic interactors is needed in order to generate hypotheses about mechanisms of Wnt pathway modulation. Additionally, it is totally unknown whether homophilic engagement of Pcdh extracellular domains causes conformational changes that may activate intracellular signaling. One interesting thing to note is that γ-Pcdh-C3, the only isoform that we found to inhibit canonical Wnt signaling [204], appears to be ubiquitously expressed, at least among neurons [123]. Thus, it may be that a basal level of homophilic interaction, mediated by C3, obtains between any two cells. If homophilic interaction potentiates signaling via Axin1, then C3 may mediate a basal suppression of the canonical pathway. Introduction of matching between other γ-Pcdh isoforms, many of which seem to potentiate Wnt signaling (e.g., A1 or B1), could strengthen cell-cell adhesion while counterbalancing C3’s inhibition of the canonical pathway. Though it is too early to speculate further on such a mechanism, it is clear that a major question is whether Pcdhs regulate Wnt signaling pathways constitutively, or whether this is dependent on cell-cell interaction; if so, the isoform composition of adhesion complexes may determine ultimate signaling outcomes.

Second, it remains unclear the extent to which changes in TOPFlash output or altered expression of Wnt target genes induced by Pcdh experimental over-expression or down-regulation due to altered methylation in vivo represents true canonical Wnt signaling. In our own unpublished experiments, for example, we have found that γ-Pcdh-C3 inhibits the expression of Wnt target genes or TCF/Lef reporter constructs without significantly altering the levels or localization of active β-catenin. Similarly, overexpression of Pcdh9 could reduce phosphorylation of GSK3β at Ser9, which should increase its activity, without a concomitant increase in assayed Wnt target genes [179]. Like many intracellular signaling proteins, Wnt components can interact with many partners and link into many distinct pathways. It is quite possible that several Pcdhs can modulate the downstream activity of β-catenin/TCF/Lef-dependent gene transcription through novel “non-canonical” pathways, including but not limited to regulation of Wnt/PCP and/or Wnt/Ca²⁺ branches of signaling. An example may be Pcdh8’s interaction with NLK [219], which can negatively feedback on β-catenin/TCF/Lef-dependent gene transcription (Figure 3).

Finally, most of the evidence linking Pcdhs in to Wnt signaling pathways comes from the study of cancer cells, and yet most major functional roles for Pcdhs identified thus far in vivo concern aspects of neuronal development, including axon outgrowth, dendrite arborization, and synaptogenesis (reviewed by [1]). Wnt signaling is known to play many important roles in the developing nervous system (reviewed by [227]); thus it will be important to determine whether altered Wnt signaling plays any role in the many neural functions identified for Pcdh proteins. Certainly, major disruption of Wnt signaling is incompatible with the formation of a structurally normal nervous system, so the discrete phenotypes observed when clustered or non-clustered Pcdhs are deleted in mice could only reflect fairly subtle modulation of Wnt pathways. Additionally, some Wnt pathway proteins can affect neural circuit formation on their own (e.g., Axin1 is important for the formation of dendritic arbors in hippocampal neurons [206]), reinforcing the point above that the Pcdhs may affect Wnt pathway components without acting through the accepted “canonical” signaling pathways.

Abbreviations

Pcdhs

protocadherins

PP1a

protein phosphatase 1a

WIRS

WRC interacting receptor sequence

FAK

focal adhesion kinase

PKC

protein kinase C

MMTV

mouse mammary tumor virus

Fzd

Frizzleds

ICD

Intracellular domain

Dvls

Dishevelled proteins

Lrp

Low-density lipoprotein-related rececptor

CKI

Casein kinase I

ROR

Receptor tyrosine kinase-like Orphan Receptor

PCP

planar-cell polarity

JNK

c-Jun N-terminal kinase

APC

adenomatous polyposis coli

GSK3β

glycogen synthase kinase 3β

TCF

T-cell factor

Lef

lymphoid enhancer factor

Daam1

Dishevelled-associated activator of morphogenesis 1

PLC

phospholipase C

CaMKII

calmodulin-dependent kinase II

PKC

protein kinase C

NLK

Nemo-like kinase

NFAT

nuclear factor of activated T cell

CSE

conserved sequence element

qPCR

quantitative PCR

VCD

variable cytoplasmic domain

BCL-9

B-cell CLL/lymphoma 9

PAPC

Paraxial protocadherin

xANR5

Xenopus ankyrin repeats domain protein 5

HCC

hepatocellular carcinoma