**Not All Wnt Activation Is Equal: Ligand-Dependent versus Ligand-Independent Wnt Activation in Colorectal Cancer**

### **Sam O. Kleeman 1,2 and Simon J. Leedham 2,\***


Received: 16 October 2020; Accepted: 12 November 2020; Published: 13 November 2020

**Simple Summary:** Colorectal cancer is the third most common cause of cancer-related deaths. The Wnt signaling pathway is activated by genetic mutations in most patients with colorectal cancer. A number of different types of Wnt pathway mutation have been described: some increase the sensitivity of tumor cells to Wnt ligands produced by stromal cells (ligand-dependent), while others drive downstream activation of the pathway (ligand-independent). Ligand-dependent tumors are of particular interest as there are a number of emerging treatment options, such as porcupine inhibitors, that can specifically target these tumors. In this review, we discuss what is known about these different types of Wnt activating mutations. We propose that ligand-dependent tumors should be viewed as a separate subset of colorectal cancer with its own biomarkers, prognosis and targeted therapies.

**Abstract:** Wnt signaling is ubiquitously activated in colorectal tumors and driver mutations are identified in genes such as APC, CTNNB1, RNF43 and R-spondin (RSPO2/3). Adenomatous polyposis coli (APC) and CTNNB1 mutations lead to downstream constitutive activation (ligand-independent), while RNF43 and RSPO mutations require exogenous Wnt ligand to activate signaling (ligand-dependent). Here, we present evidence that these mutations are not equivalent and that ligand-dependent and ligand-independent tumors differ in terms of underlyingWnt biology, molecular pathogenesis, morphology and prognosis. These non-overlapping characteristics can be harnessed to develop biomarkers and targeted treatments for ligand-dependent tumors, including porcupine inhibitors, anti-RSPO3 antibodies and asparaginase. There is emerging evidence that these therapies may synergize with immunotherapy in ligand-dependent tumors. In summary, we propose that ligand-dependent tumors are an underappreciated separate disease entity in colorectal cancer.

**Keywords:** Wnt; signaling; colorectal; cancer; porcupine; R-spondin; serrated; immunotherapy

#### **1. Introduction**

Metastatic colorectal cancer (CRC) is a lethal malignancy with a five-year survival of less than 15% [1]. Patients with metastatic CRC are treated with combination cytotoxic chemotherapy alongside monoclonal antibodies targeting angiogenesis or epidermal growth factor receptor (EGFR) [2]. There is a need to develop new therapeutic strategies for metastatic cancer, especially in light of evidence showing rapid increases in CRC incidence affecting younger patients [3]. Molecular profiling of CRC has shown considerable disease heterogeneity, suggesting that patients might benefit from precision medicine, in which treatments are personalized for their tumor profile [4]. For example, immunotherapy targeting PD-1/PD-L1 signaling is only active in hypermutated tumors, while anti-EGFR antibodies are only effective in tumors without downstream mutations [5–7].

Colorectal cancer is characterized by near-ubiquitous activation of the Wnt signaling pathway [8]. The Wnt pathway is an evolutionarily conserved mechanism for intercellular communication, with essential roles in embryogenesis and adult tissue development [9]. In the colonic crypt, Wnt signaling is necessary to maintain the adult intestinal stem cell niche and epithelial homeostasis [10]. Colorectal tumors are dependent upon aberrant Wnt signaling to maintain stemness and a de-differentiated phenotype and genetic Wnt inhibition leads to rapid tumor regression [11]. Additionally, Wnt signaling can protect cells from immune surveillance, thus restricting anti-tumoral immunity [12,13]. Altogether, this suggests that the Wnt pathway could be a viable therapeutic target for patients with CRC.

Colorectal tumors are thought to evolve through the sequential acquisition of mutations driving progression from a normal founder cell to adenoma and then carcinoma [14]. Adenomas can be histologically classified as either conventional, such as tubular or tubulovillous (TVA), or serrated, such as sessile serrated lesions (SSL) or traditional serrated adenomas (TSA) [15]. Serrated adenomas are characterized histologically by a saw-tooth morphology. The cell-of-origin for TVA is likely the crypt-based columnar stem cell [16] while the cell-of-origin for serrated lesions is unknown, but may derive from ectopic crypt foci in the rare traditional serrated adenoma subtype [17].

Here, we present evidence for a new model of CRC in which Wnt pathway activation can take one of two distinct trajectories, ligand-dependent (LD) and ligand-independent (LI), with implications spanning tumor biology, screening, diagnosis and treatment. We will first outline the Wnt signaling pathway in the normal colon and types of recurrent Wnt mutations in CRC. We will then discuss morphological and molecular biomarkers that can be used to identify LD tumors in the clinic. Finally, we will argue that this model has the potential to transform the landscape of precision medicine in CRC.

#### **2. Wnt Signaling Pathway**

The Wnt signaling pathway in normal colon crypts is summarized in Figure 1. Briefly, canonical Wnt ligands are secreted by the cells in the stem cell niche following O-acylation by porcupine [18,19]. Wnt ligands bind to Frizzled (FZD) and lipoprotein receptor-related protein (LRP) receptor complexes on the plasma membrane of neighboring cells [20]. Both Wnt ligand secretion and binding to FZD depend upon acylation of Wnt ligands [21]. The downstream effector of Wnt signaling is the transcriptional co-activator β-catenin (CTNNB1). In the absence of Wnt ligands, CTNNB1 is degraded by the action of a destruction complex containing adenomatous polyposis coli (APC), axin-like protein (AXIN1/2), glycogen synthase kinase (GSK3) and casein kinase (CSNK1A) [22]. Wnt ligand binding inhibits the destruction complex, thus stabilizing CTNNB1 and activating expression of Wnt target genes. An additional level of regulation comes from E3 ubiquitin ligases ring finger protein 43 (RNF43) and zinc and ring finger 3 (ZNRF3), which constitutively degrade FZD to repress Wnt signaling [23]. R-spondin (RSPO) ligands bind to leucine-rich repeat-containing G protein-coupled (LGR) receptors, inhibiting RNF43/ZNRF3 and substantially amplifying Wnt signaling [24]. There are four homologous human RSPO ligands (RSPO1-4) and, while all four can bind to LGR-family receptors, the EC50 for activation of Wnt signaling varies 100-fold, with RSPO2 and RSPO3 demonstrating the highest potency (0.02–0.05 nM) [25]. R-spondins are produced by stromal cells adjacent to the stem cell niche [26]. Consistent with this, R-spondin signaling is necessary to maintain the stem cell niche, both in vivo and as part of organoid culture systems [24,27,28]. Wnt target genes, such as notum palmitoleoyl-protein carboxylesterase (NOTUM) and AXIN2 are negative regulators of Wnt signaling, functioning as negative feedback loops to fine-tune and limit downstream signaling [29]. AXIN2 is an inducible component of the destruction complex, while NOTUM works in the extracellular space to deacetylate and inactivate Wnt ligands [30].

**Figure 1.** Overview of Wnt signaling pathway. Wnt ligands secreted from stromal cells are activated by porcupine-mediated post-translational modification and bind to Frizzled (FZD) receptors on Wnt receiving cells. This functions to inhibit a destruction complex containing axin-like protein (AXIN)1/2 and adenomatous polyposis coli (APC) thus disinhibiting β-catenin (CTNNB1), the master transcriptional regulator ofWnt singaling. Frizzled receptors are degraded due to the action of ring finger protein 43 (RNF43), which is in turn inhibited by binding of R-spondin (RSPO) ligands to leucine-rich repeat-containing G protein-coupled (LGR) family receptors, thus augmenting Wnt signaling tone. Wnt pathway activation is regulated at multiple levels by negative feedback loops, including those mediated by AXIN2 and notum palmitoleoyl-protein carboxylesterase (NOTUM). Recurrent mutations in CTNNB1 and APC result in ligand-independent pathway activation while mutations in RSPO and RNF43 depend upon binding of Wnt ligands to Frizzled receptors. GSK: glycogen synthase kinase; LRP: lipoprotein receptor-related protein.

#### **3. Ligand-Dependent and Ligand-Independent Alterations in Colorectal Cancer**

Large-scale sequencing studies in CRC have established the presence of pervasive Wnt pathway mutations. Recurrent mutations include loss-of-function mutations in APC and RNF43, and gain-of-function mutations in CTNNB1 and RSPO2/3 (Figure 1, Table 1) [8,31,32]. Consistent with driver mutation status, in vivo modeling indicates that these mutations can be sufficient for colorectal tumorigenesis [33–36]. For tumor suppressors APC and RNF43, we only consider protein-truncating mutations or deletions as potential driver alterations [37]. ZNRF3 is a homolog of RNF43 but truncating mutations are rare in colorectal tumors, potentially reflecting its comparably low mRNA expression in normal colon and colorectal tumors [8,38]. This is in contrast to the situation in murine intestine, in which Znrf3 and Rnf43 gene expression is comparable, and loss-of-function alterations in both Znrf3 and Rnf43 are necessary to activate Wnt signaling [36,39].

While *APC* and *CTNNB1* alterations drive downstream, constitutive activation of the Wnt pathway that is independent of Wnt ligand binding (ligand-independent, LI), *RSPO* and *RNF43* alterations disrupt the synergistic RSPO axis and by doing so, amplify endogenous and otherwise intact, Wnt ligand signaling (ligand-dependent, LD). APC mutations are characteristically nonsense or frameshift alterations affecting the "mutation cluster region", often with a second "hit" from loss of heterozygosity [40]. CTNNB1 mutations are gain-of-function missense mutations affecting specific amino acid residues that are phosphorylation sites for components of the destruction complex [41].


**Table 1.** Driver Wnt alterations in colorectal cancer. Prevalence refers to the frequency of each mutation in the subset of colorectal tumors with a detectable driver Wnt alteration, as derived from [42]. Loss-of-function alterations in APC and RNF43 are frequently accompanied by loss of heterozygosity (LOH) affecting the second allele [13].

RSPO mutations induce R-spondin ligand overexpression either from epithelial cells (autocrine), as RSPO fusion genes [42]. R-spondin gain-of-function is only observed for RSPO2 and RSPO3, consistent with their enhanced potency to induce Wnt signaling in vitro [25]. RSPO3 fusion genes commonly result in the replacement of RSPO3 exon one and promoter with that of a gene with higher basal expression, resulting in a functional epithelial-expressed protein [32]. A wide range of fusion partners have been identified including PTPRK, EIF3E, NRIP1 and PIEZO1 [43,44], all of which are associated with relatively high constitutive gene expression [38]. RSPO fusions cannot be reliably identified even from whole-genome sequencing due to large inconsistency in genomic alterations, while the transcript breakpoints are more stereotypical [32]. Alternatively, in a rare subset of colorectal tumors, we identified R-spondin overexpression in the absence of RSPO fusions or any other detectable Wnt driver alteration [42]. In situ hybridization demonstrated high stromal RSPO3 expression in these tumors, implicating a role for paracrine R-spondin signaling driven by stromal overexpression [42]. RSPO3 overexpression in the absence of gene fusions has been previously detected in lung cancer, where it was associated with RSPO3 hypomethylation [45]. The concept that RSPO overexpression can derive from either epithelial or stromal sources is consistent with previous evidence that RSPO3 expression is significantly and positively correlated with stromal expression signatures [46]. RNF43 mutations are mostly recurrent frameshift mutations at amino acid positions 117 and 659 that result in a truncated gene product [31]. These recurrent mutations occur at tandem repeats called microsatellites whose stability is dependent upon proficient mismatch repair (MMR) [47]. As a result, these mutations tend to occur in tumors with MMR deficiency, detected as microsatellite instability (MSI), which is often caused by promoter hypermethylation of MLH1 [8,48].

Recently, there has been some controversy about whether the RNF43 G659Vfs\*41 mutation demonstrably leads to impaired protein function. In vitro transfection experiments have indicated that this mutant RNF43 protein retains the ability to bind R-spondin and repress Frizzled [49]. However, this alteration is associated with significantly reduced RNF43 expression, potentially consistent with nonsense-mediated decay [49], and CRISPR-Cas9 editing of the endogenous RNF43 locus to mimic the G659Vfs\*41 mutation, was sufficient to increase cell surface Frizzled expression [50]. Furthermore, the G659Vfs\*41 mutation occurs substantially more often than would be expected by chance in microsatellite-unstable tumors, indicating strong positive selection [31,51].

It is important to note that driver Wnt alterations affecting APC, CTNNB1, RNF43 and RSPO in pre-cancerous polyps and tumors show marked mutual exclusivity [42]. There are two logical implications from this: firstly, these alterations are redundantly able to activate Wnt signaling. Secondly, there may be selection against the accumulation of driver alterations in more than one gene. This is consistent with the "just right" theory of Wnt signaling: that there is an optimal level of Wnt activation to drive tumorigenesis. It has been observed that there is a non-random distribution of second "hit" mutations in APC that is consistent with selection for APC genotypes that retain some CTNNB1 repression [52]. Additionally, ectopic expression of R-spondin in APC-mutant mice results in reduced proliferation and increased apoptosis [53], consistent with evidence that Wnt can directly promote apoptosis [54].

#### **4. Mutation Selection in Lesion Subtypes**

Molecular profiling in pre-cancerous polyps has shown that ligand-dependent alterations are predominantly seen in the serrated pathway (Figure 2) [44,55]. A total of 55% of TSAs have ligand-dependent alterations, namely truncating RNF43 mutations or RSPO fusions (mostly PTPRK-RSPO3) [44]. In sessile serrated lesions (SSL), mutations in the Wnt signaling pathway are not thought to be initiating lesions as Wnt disruption is observed predominantly in dysplastic rather than non-dysplastic lesions. A total of 50% of SSLs had ligand-dependent RNF43 mutations [55], whereas APC mutations are much rarer in serrated lesions, being detected in 13% and 9% of TSAs and dysplastic SSLs, respectively [44,55]. In contrast, conventional adenomas have a high frequency (>85%) of ligand-independent alterations [56]. APC mutation is sufficient to initiate adenoma pathogenesis [57] and no ligand-dependent alterations have been reported in conventional adenomas [44].

**Figure 2.** Molecular pathogenesis of different colorectal precursor subtypes. Colorectal cancer develops from three types of pre-cancerous polyps: conventional and serrated adenomas—divided into traditional serrated adenomas (TSAs) and sessile serrated lesions (SSLs). Conventional adenomas are driven by ligand-independent mutations that likely arise in the crypt base columnar (CBC) stem cells. TSAs arise from APC, RSPO or RNF43 mutations, possibly in ectopic crypts. SSL pathogenesis is characterized by the late acquisition of APC or RNF43 mutations, concurrent with the onset of the detectable dysplasia. Ligand-dependent CRC arises from TSAs and SSLs while ligand-independent CRC arises from all three types of polyp (bottom panel).

Altogether, this raises the possibility that different intestinal lesions follow distinct molecular carcinogenesis pathways. These different evolutionary trajectories appear to result in the selection of either ligand-dependent or independent mutations. This may also partly explain the mutual exclusivity of Wnt driver mutations discussed above. Why polyp subtypes acquire apparent obligatory Wnt disruption through these different mechanisms is unknown, but may be influenced by the variable cell-of-origin in different lesion subtypes (Figure 2). Indeed, APC mutations induce tumorigenesis in vivo if introduced into the LGR5+ intestinal stem cell but not transit-amplifying cells [16], while RSPO fusions significantly co-occur with loss-of-function mutations in the Bone morphogenic protein (BMP) signaling pathway that are known to induce ectopic crypt formation [58,59]. These data would also suggest that RSPO-mutant colorectal tumors are wholly derived from TSAs.

#### **5. Negative Regulation of Wnt Signaling**

In some ways, it is surprising that despite multiple levels of negative feedback, ligand-dependent mutations, which act upstream in an otherwise normal pathway, can induce activation of Wnt signaling at all. Ligand-dependent pathway activation would be expected to induce physiological expression of Wnt negative regulators such as AXIN2 or NOTUM, which would function to proportionately constrain activation of the pathway. In contrast, ligand-independent alterations result in downstream, constitutive activation that is uncoupled from the action of negative regulators. We have recently shown that tumors with ligand-dependent alterations are associated with significant repression of Wnt negative regulators, especially AXIN2 [42]. This repression may be at least partly explained by AXIN2 methylation [60]. This raises the possibility that ligand-dependent Wnt activation requires two "hits"—firstly a driver mutation affecting RNF43 or RSPO, and secondly, epigenetic downregulation of Wnt negative regulators. Indeed, serrated adenomas which are enriched for ligand dependent mutations, have lower AXIN2 expression and increased AXIN2 methylation compared to conventional tubulovillous adenomas [61,62], as do MSI-high cancers that progress via this pathway [45,53]. AXIN2 expression is also decreased in an in vivo model of ligand-dependent tumors, generated by orthotopic engraftment of CRISPR-edited organoids [63]. Furthermore, ectopic expression of AXIN2, leading to re-activation of Wnt negative feedback in an RNF43-mutant cell line (HCT116) resulted in rapid cell death [60,61]. In fact, AXIN2 is not the only Wnt negative regulator known to be silenced by promoter hypermethylation in colorectal cancer: hypermethylation has been detected in negative regulators including WIF1, SFRP1/2/4, DKK1–3 and NOTUM [42,62,64]. These genes are predominantly hypermethylated in ligand-dependent or microsatellite-unstable tumors. This suggests that repression of negative regulators is a more global phenomenon in ligand-dependent CRC, with loss of negative feedback mechanisms at multiple levels of the Wnt signaling pathway.

#### **6. Application of AXIN2 as a Biomarker for Ligand-Dependent Wnt Biology**

Our finding that ligand-dependent tumors exhibit suppressed expression of negative regulators of Wnt can be harnessed to utilize AXIN2 as a single-gene biomarker to distinguish between ligand-dependent and ligand-independent tumors at the point of diagnosis. This is particularly important as otherwise ligand-dependent tumors would need to be identified from expensive and time-consuming analysis of paired DNA (for APC, CTNNB1 and RNF43) and RNA sequencing (RSPO fusions). Paired DNA and RNA sequencing is simply not practical for routine diagnostic assessment in the clinic, both in terms of cost and the relatively high failure rate of sequencing (>10%) from diagnostic clinical samples [65].

We recently demonstrated that AXIN2 mRNA expression could be used as a discriminatory biomarker with an area under the curve (AUC) greater than 0.93 in three independent cohorts, indicating excellent diagnostic performance. This analysis incorporated both RNA sequencing and microarray profiling to assay gene expression in resection and biopsy specimens [42]. The diagnostic performance corresponded to sensitivity and specificity >90%. We also demonstrated similar results with high-throughput AXIN2 profiling by quantitative real-time polymerase chain reaction

(qRT-PCR). These findings were recently supported by the use of an organoid biobank derived from patients with colorectal cancer, in which organoids with RSPO fusions or RNF43 mutations exhibited lower AXIN2 expression than APC-mutant organoids [58]. Our analysis of paired qRT-PCR and immunohistochemistry for AXIN2 showed that there was only weak correlation between AXIN2 mRNA and scored AXIN2 protein expression, suggesting that AXIN2 may undergo significant translational regulation, as has been described previously [66]. This would suggest that profiling of AXIN2 mRNA expression would be the preferred approach to translate this biomarker into the clinic.

It is worth noting that AXIN2 gene expression is widely used as a read-out of global Wnt pathway activation [67] and our findings suggest that this should be interpreted with caution, as AXIN2 expression can be confounded by the type of acquired Wnt disrupting pathway mutation. This confounding has important implications for the interpretation of analyses that have demonstrated inverse correlations between AXIN2 (used as a read-out of Wnt activation) and immune infiltration [13]. Tumors with low AXIN2 expression are enriched with RNF43-mutant MSI-high tumors that have enhanced anti-tumoral immune responses, thought to result from an increased neoantigen load [68]. As a result, the inverse relationship between AXIN2 and immune infiltration may be partly explained by increased mutational load in RNF43-mutant ligand-dependent tumors, rather than reduced Wnt activation.

In summary, the distinction between ligand-dependent and ligand-independent tumors is clinically-actionable because tumors can be robustly discriminated using a low-cost single-gene molecular biomarker.

#### **7. Non-Overlapping Clinicopathological Features of Ligand-Dependent Tumors**

Consistent with altered Wnt pathway biology and an altered trajectory through the serrated pathway, ligand-dependent tumors have non-overlapping morphological and clinical characteristics with ligand-independent tumors, reflecting an underappreciated separate disease entity in colorectal cancer. Using manual and automated digital pathological approaches, we have demonstrated that ligand-dependent tumors are enriched with mucin [13,42]. Mucin is a high molecular-weight glycoprotein that is secreted by goblet cells and forms a key component of the mucous layer that provides physical protection in the gastrointestinal tract [69]. Mucinous differentiation has long been recognized in a subset of colorectal tumors (around 10%) and is diagnosed in tumors where mucin comprises >50% of the tumor volume [70]. Indeed, mucinous differentiation is associated with microsatellite instability, implicating a link with RNF43-mutant tumors. We have demonstrated that computational-scored mucin area alone could discriminate between ligand-dependent and ligand-independent tumors with an AUC > 0.75. Based on our findings, we propose that mucinous differentiation may well either be induced by ligand-dependent Wnt signaling or reflect the association with the serrated pathway. Consistent with the former hypothesis, the induction of ligand-dependent alterations in organoids is sufficient to generate orthotopic colon tumors with mucinous differentiation [63]. Furthermore, RNF43 mutations in biliary malignancies are associated with mucin hypersecretion [71]. Altogether, this suggests that mucin content, which is routinely scored by histopathologists, can be used as a phenotypic biomarker for ligand-dependent tumors with good diagnostic performance.

In our comparison of ligand-dependent and ligand-independent tumors in a pooled cohort of over 600 tumors with available outcome data, we did not identify any significant differences in prognosis [42]. However, this is likely to mask, considerably, the prognostic heterogeneity between the subsets of ligand-dependent tumors. One way to examine this is to compare specific subsets with their consensus molecular subtype (CMS) classifications, as this study was well-powered to identify prognostic associations incorporating over 2000 patients [72]. Ligand-dependent tumors appear to lie on a continuum between RNF43-mutant tumors which mostly classify as CMS1 (associated with good prognosis) and tumors with stromal RSPO overexpression which mostly classify as CMS4 (associated with poor prognosis). Consistent with this, we observed a high frequency of tumor budding and enriched desmoplastic stroma in tumors with stromal RSPO overexpression, both of which are

associated with poor prognosis [73,74]. Of note, mucinous differentiation is associated with marginally reduced overall survival [75]. These data suggest that the prognostic implications of ligand-dependent Wnt biology are likely to be highly heterogenous.

#### **8. Selective Vulnerabilities in Ligand-Dependent Tumors**

Downstream ligand-independent Wnt signaling has proved difficult to target in solid tumors, reflecting challenges in designing small-molecule inhibitors to inhibit constitutive pathway activation through transcription factors such as beta-catenin [76]. In contrast, from a conceptual and experimental standpoint, ligand-dependent Wnt activation is inherently "druggable" through deprivation of extracellular ligand (Wnt or R-spondin) or attenuation of negative regulator suppression with demethylating agents (Figure 3) [27,60,77]. Furthermore, emerging evidence would indicate that these selective vulnerabilities in ligand-dependent tumors could synergize with immunotherapy targeting PD-1/PD-L1 signaling in tumors [78]. This makes ligand-dependent tumors a fascinating subset of colorectal cancer, with the real possibility of new transformative treatments.

**Figure 3.** Target therapies for ligand-dependent tumors. All ligand-dependent tumors require Wnt ligands for pathway activation and so are sensitive to porcupine inhibitors that impair Wnt ligand activation. RSPO3 overexpression can be antagonized by anti-RSPO3 antibodies. Ligand-dependent GSK3 inhibition results in reduced proteasomal degradation to generate amino acids such as asparagine, making tumor cells sensitive to asparagine depletion with asparaginase treatment. Ligand-dependent tumors are characterized by AXIN2 repression which can be antagonized with licensed demethylating agents such as azacitidine.

By definition, ligand-dependent Wnt alterations can only induce downstream Wnt pathway activation in the presence of Wnt ligand. As a result, depletion and inactivation of Wnt ligand by inhibition of porcupine is a viable therapeutic approach for ligand-dependent tumors. In vitro models of ligand-dependent tumors, including organoids with RNF43 mutations [79] and cell lines with RSPO fusions [46], are exquisitely sensitive to porcupine inhibitors. This has also been demonstrated in various in vivo settings, including xenografts with RSPO fusions [77] and autochthonous Rnf43/Znrf3-null intestinal tumors [80]. Porcupine inhibition is associated with marked repression of Wnt pathway activity, reduced tumor size and substantial remodeling the transcriptomic landscape that includes increased intestinal differentiation [77,80]. Porcupine inhibitors have entered early-phase clinical trials (NCT01351103, NCT03447470, NCT03507998). Preliminary evidence from a phase 1 trial identified a

partial response in one patient with a detectable RNF43 mutation [81] while porcupine inhibition was associated with reduced AXIN2 expression, suggesting on-target effects [82].

However, in vitro modeling of porcupine inhibition in ligand-dependent CRC cell lines has identified selection for resistance mutations, such as loss-of-function alterations to AXIN1, leading to loss of function of the destruction complex and downstream constitutive pathway activation [46]. It is worth noting that AXIN2 repression seen in ligand-dependent tumors does not result in downstream pathway activation because of redundancy with AXIN1. AXIN1 is a constitutive component of the destruction complex and not a Wnt pathway target. This would suggest that AXIN1 inactivation alone would not be sufficient to drive Wnt pathway activation unless AXIN2 was concurrently repressed—we would hypothesize that this situation could only arise in ligand-dependent tumors. This might explain the relatively low frequency (<0.05%) of truncating AXIN1 mutations seen in CRC [8].

In tumors with epithelial RSPO fusions, the autocrine signaling loop can be blocked by an anti-RSPO3 antibody. For example, treatment with anti-RSPO3 antibody has been shown to result in inhibition of xenograft tumor growth with tumor regression in some cases [27,83–85]. As with porcupine inhibitors, this was associated with evidence of increased intestinal differentiation on morphological and transcriptomic analysis [27,83]. This differentiated phenotype was associated with reduced expression of stem cell markers and key Wnt targets such as LGR5 and ASCL2. A phase 1 trial of an anti-RSPO3 antibody in patients with metastatic colorectal cancer was associated with partial responses in some patients, although this was not clearly associated with baseline RSPO3 expression [86]. In addition, while it has not been formally tested, it is entirely plausible that anti-RSPO3 therapy would also be effective for tumors with stromal RSPO overexpression.

The Wnt pathway plays a critical role in bone homeostasis [87] and unsurprisingly inhibition of ligand-dependent Wnt signaling via porcupine inhibitors or anti-RSPO3 antibodies results in on-target bone toxicity, including reduced bone strength and pathological fractures [86,88,89]. Consistent with this, porcupine-null mice have widespread bone defects, while germline loss-of-function Wnt ligand mutations in humans are associated with high fracture risk [90–92]. Preliminary evidence has shown that bone toxicity could be reduced with co-administration of denosumab, which inhibits bone degradation [89]. Altogether, concerns about resistance and on-target toxicity would likely limit the use of direct Wnt inhibitions (porcupine, anti-RSPO3) to short durations of time, likely in conjunction with other treatments.

In light of evidence that ligand-dependent tumors may depend upon repression of negative regulators, possibly via promoter hypermethylation [42], demethylating agents could be a viable therapeutic strategy in ligand-dependent tumors. Demethylation treatment with azacitidine in HCT116, a colorectal cancer cell line with an RNF43 mutation and comparatively low AXIN2 expression, resulted in increased AXIN2 expression and increased cell death [60,93]. Azacitidine is an approved treatment for myelodysplastic syndrome with a well-established toxicity profile suggesting that this would be a feasible treatment for ligand-dependent CRC [94].

Unexpectedly, recent work in acute myeloid leukemia found that asparaginase treatment was synthetically lethal with inhibition of GSK3 [95]. Asparaginase functions to deaminate and so degrade the nonessential amino acid asparagine, which is required for leukemic cell growth [96]. GSK3 mediates ubiquitination of a wide range of proteins, such as APC, and resulting proteasomal degradation provides a source of asparagine in the cell. Asparaginase treatment has a relatively favorable toxicity profile and is licensed for acute myeloid leukemia [97]. In contrast to ligand-independent alterations, which act downstream and by-pass GSK3, ligand-dependent mutations directly lead to inhibition of GSK3 (Figure 1) through activation of the canonical Wnt pathway, thus explaining a unique selective vulnerability for asparaginase treatment in ligand-dependent tumors. Specifically, asparaginase treatments were highly toxic for organoids with RSPO fusions but had no activity against organoids with APC or CTNNB1 mutations [98]. Treatment of mice with subcutaneous implantation of RSPO-mutant organoids was associated with marked tumor regression and prolonged progression-free survival, with no evidence of early therapy resistance [98]. No benefit

was seen for implanted APC-mutant organoids. Altogether, these data would suggest that asparaginase could be a viable and well-tolerated treatment for patients with ligand-dependent CRC.

In summary, ligand-dependent Wnt biology is associated with a range of therapeutic vulnerabilities that could be exploited as effective anti-cancer therapy.

#### **9. Combination Therapy for Ligand-Dependent Tumors**

Considering that direct inhibition of the Wnt pathway is unlikely to be feasible for extended periods of time, it is important to consider how treatments for ligand-dependent tumors might synergize with existing anti-cancer therapy. Wnt pathway activation is often detected as a marker of resistance to cytotoxic chemotherapy [99]. Resistance to paclitaxel, which is a type of cytotoxic chemotherapy that inhibits microtubule detachment from centrosomes, is associated with Wnt pathway activation, detected as increased CTNNB1 protein expression. Considering that Wnt functions as a regulator of centrosome separation [100], it is feasible that Wnt activation could directly promote survival of tumor cells. Consistent with this, anti-Wnt treatments such as anti-RSPO3 antibodies synergize with paclitaxel in patient-derived xenografts with RSPO3 fusions [84].

More generally, inhibition of Wnt signaling in ligand-dependent tumors is consistently shown to skew cells from a stem-like phenotype to a more differentiated phenotype [27,101,102]. Resistance to cancer radiotherapy and chemotherapy is often driven by acquisition of stem-like phenotypes, with enrichment of tumor cells that are able to repopulate a tumor on transplantation, often termed cancer stem cells [103,104]. This suggests that short courses of Wnt pathway inhibitors could be synergistic with a wide range of existing and innovative drug regimens, especially if Wnt inhibitors are early in the treatment schedule.

The Wnt signaling pathway appears to play a role in protecting cells from immune surveillance. As a result, there is considerable interest in the combination of immunotherapy that targets PD-L1/PD-1 signaling and direct inhibition of the Wnt signaling pathway. Signaling through the PD-1 receptor is thought to promote an exhausted phenotype in cytotoxic T cells that impairs effective anti-tumoral immunity [105]. While immunotherapy has demonstrated activity in diverse tumor types, it has proved ineffective in unselected patients with CRC [5]. There are multiple lines of evidence that the Wnt signaling pathway can directly promote an immune suppressive environment [106]. Early data from trials of porcupine inhibitors have shown evidence for increased expression of activated immune signatures [82]. Furthermore, porcupine inhibition was synergistic with anti-CTLA4 immunotherapy in a murine melanoma model [107]. Altogether, this raises the question of whether anti-Wnt therapies would act synergistically with immunotherapy in colorectal tumors and this hypothesis is under active investigation in several early-phase clinical trials (NCT01351103, NCT02521844, NCT02675946).

Furthermore, as discussed above, the microsatellite-unstable subset of colorectal tumors ligand-dependent tumors is enriched with tumors, which have enhanced responses to immunotherapy [6]. Unexpectedly, a recent analysis incorporating a large cohort of patients with colorectal cancer who were treated with immunotherapy, demonstrated that RNF43-mutant tumors responded significantly better to immunotherapy than would have been expected from their mutational burden [108]. This is an exciting finding that raises the possibility the ligand-dependent Wnt biology might be independently associated with responses to immunotherapy and warrants further investigation in additional cohorts.

#### **10. Outlook—Landscape of Precision Medicine in CRC**

Approximately 15% of colorectal tumors have ligand-dependent alterations in the Wnt signaling pathway, affecting RNF43 or RSPO2/3. This unique Wnt biology is associated with a range of specific therapeutic vulnerabilities, especially to depletion of Wnt ligand by porcupine inhibitors. There is a strong theoretical basis for the combination of immunotherapy with a time-limited course of porcupine inhibition. Inhibitors of ligand-dependent Wnt signaling are known to be ineffectual in tumors with ligand-independent alterations such as APC mutations [79]. As a result, due to the low frequency of ligand-dependent alterations, clinical trials of these selective treatments will fail in unselected patients. Precision medicine depends upon the ability to stratify patients into clinically meaningful subsets, followed by targeting with biologically appropriate therapies. It is contingent on the existence of biomarkers specific for each subset that can be feasibly adopted into routine clinic practice. We propose that AXIN2 is one such biomarker and could be measured at low cost from routine clinical specimens. It can be measured by high-throughput qRT-PCR and does not require costly and time-consuming DNA and RNA sequencing. On the basis of AXIN2 expression, it would be possible to identify patients with ligand-dependent Wnt biology who could then be targeted with effective personalized therapies. In summary, we propose that the concept of ligand-dependent tumors as an individual disease entity has the potential to revolutionize precision medicine and improve the outcomes for patients with colorectal cancer.

**Author Contributions:** S.O.K. and S.J.L. co-wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Wellcome Trust Senior Clinical Research Fellowship (206314/Z/17/Z).

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Wnt and Vitamin D at the Crossroads in Solid Cancer**

### **José Manuel González-Sancho 1,2,3, María Jesús Larriba 1,3,4 and Alberto Muñoz 1,3,4,\***


Received: 29 October 2020; Accepted: 17 November 2020; Published: 19 November 2020

**Simple Summary:** The Wnt/β-catenin signaling pathway is aberrantly activated in most colorectal cancers and less frequently in a variety of other solid neoplasias. Many epidemiological and experimental studies and some clinical trials suggest an anticancer action of vitamin D, mainly against colorectal cancer. The aim of this review was to analyze the literature supporting the interference of Wnt/β-catenin signaling by the active vitamin D metabolite 1α,25-dihydroxyvitamin D3. We discuss the molecular mechanisms of this antagonism in colorectal cancer and other cancer types. Additionally, we summarize the available data indicating a reciprocal inhibition of vitamin D action by the activated Wnt/β-catenin pathway. Thus, a complex mutual antagonism between Wnt/β-catenin signaling and the vitamin D system seems to be at the root of many solid cancers.

**Abstract:** Abnormal activation of the Wnt/β-catenin pathway is common in many types of solid cancers. Likewise, a large proportion of cancer patients have vitamin D deficiency. In line with these observations, Wnt/β-catenin signaling and 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3), the active vitamin D metabolite, usually have opposite effects on cancer cell proliferation and phenotype. In recent years, an increasing number of studies performed in a variety of cancer types have revealed a complex crosstalk between Wnt/β-catenin signaling and 1,25(OH)2D3. Here we review the mechanisms by which 1,25(OH)2D3 inhibits Wnt/β-catenin signaling and, conversely, how the activated Wnt/β-catenin pathway may abrogate vitamin D action. The available data suggest that interaction between Wnt/β-catenin signaling and the vitamin D system is at the crossroads in solid cancers and may have therapeutic applications.

**Keywords:** wnt; β-catenin; vitamin D; cancer; colon cancer

#### **1. Introduction**

#### *1.1. Wnt*

Wnt proteins are extracellular signaling molecules that control many key processes during embryonic development and regulate the homeostasis of adult tissues, mainly by modulating the survival, self-renewal, and proliferation of stem cells. They are secreted by a variety of cell types and typically have a short range of action, mediating communication between neighboring cells. Wnt proteins bind to cell surface receptors, of which several classes have been described. Specific Wnt-receptor combinations and cellular contexts determine which of the existing Wnt signaling pathways is engaged [1].

The Wnt/β-catenin pathway is triggered by Wnt binding to cell membrane receptors of the Frizzled and low-density lipoprotein receptor-related protein (LRP) families. In the absence of a Wnt ligand, β-catenin protein is mainly located at cell-cell contacts and free cytoplasmic β-catenin is kept low because of a proteolytic destruction machinery. A complex containing the tumor suppressor proteins APC (*adenomatous polyposis coli*) and axin, and the kinases casein kinase 1 (CK1) and glycogen synthase kinase-3β (GSK-3β) targets β-catenin for *N*-terminal phosphorylation and subsequent ubiquitination and proteasome-mediated degradation. Wnt binding to Frizzled and LRP5/6 leads to inhibition of the β-catenin destruction complex and, therefore, to the accumulation of β-catenin in the cytoplasm. A proportion of β-catenin enters the nucleus and binds to transcription factors of the LEF/TCF family acting as a co-activator and regulating the expression of a large variety of genes. Wnt target genes are cell and tissue type-dependent and affect many cellular functions and processes, such as cell proliferation, stemness, migration, and invasion. Some of these targets include the *c*-*MYC* and *CCND1*/cyclin D1 oncogenes, the Wnt inhibitors *DKK1* and *NKD1*/*2*, and the Wnt effector *LEF1*. In addition, the Wnt inhibitor *AXIN2* is the most ubiquitously regulated β-catenin target gene [1–3].

Wnt/β-catenin signaling is highly dependent on the number of Frizzled receptor molecules present on the cell surface. Vertebrates have evolved a complex regulatory mechanism to control the amount of Frizzled on the plasma membrane that involves three types of proteins: leucine-rich repeat-containing G-protein coupled receptors (LGR4-6), their extracellular ligands R-spondins (RSPO1-4), and the E3 transmembrane ubiquitin ligases ZNRF3 and RNF43 [4,5]. In the absence of RSPO, Frizzled receptors are targeted for degradation by ZNRF3/RNF43-mediated ubiquitination, which results in low Frizzled membrane concentration and, therefore, in attenuated Wnt signaling. In contrast, RSPO binding to LGR4-6 sequesters ZNRF3/RNF43 in a ternary complex and prevents ubiquitin tagging of Frizzled. Thus, RSPOs are responsible for the accumulation of Frizzled receptors on the cell surface and the potentiation of Wnt/β-catenin signaling in target cells [6–8].

Dysregulation of Wnt/β-catenin signaling is involved in human diseases including cancer. In many types of cancer, e.g., colorectal, breast, and liver carcinoma, melanoma and leukemia, β-catenin constitutively accumulates within the nucleus of tumor cells [9–13]. In fact, aberrant activation of the Wnt/β-catenin pathway is the most common event in human colorectal cancer (CRC) [14,15] in which massive sequencing has estimated that over 94% of primary colon tumors carry mutations in one or more genes involved in this pathway [16]. Truncation mutations and allelic losses in the tumor suppressor gene *APC* are present in around 80% of sporadic CRC cases, whereas a small proportion carries mutations in *AXIN2* or *CTNNB1*/β-catenin genes. Moreover, chromosomal rearrangements in R-spondin family members *RSPO2* and *RSPO3* have been found in 10% of human CRC leading to enhanced Wnt signaling [17,18]. In addition, *RNF43* is mutated in a proportion of mismatch repair-deficient colon tumors. Alterations in genes encoding components of the Wnt/β-catenin pathway are frequently mutually exclusive, which confirms that aberrant activation of this pathway is a hallmark of CRC. Mutations in *CTNNB1*/β-catenin or *AXIN2* have been reported in other human tumors, e.g., hepatocellular carcinomas [19–21], whereas overexpression of Wnt factors/receptors or silencing of extracellular Wnt inhibitors are the preferred mechanisms of Wnt/β-catenin sustained activation in other cancers, e.g., breast and lung cancers [22–27].

#### *1.2. Vitamin D*

Vitamin D3 (cholecalciferol) is a natural seco-steroid whose main source is non-enzymatic production in human skin from UV-B exposed 7-dehydrocholesterol, an abundant cholesterol precursor [28,29]. Vitamin D3 from skin production and from dietary uptake is hydroxylated in the liver to 25-hydroxyvitamin D3 (25(OH)D3, calcidiol), a stable compound that is used as a biomarker for the vitamin D status of a person [29–31]. Subsequent hydroxylation of 25(OH)D3 at carbon 1, which occurs mainly in the kidneys but also in several types of epithelial and immune cells, renders 1,25-dihydroxyvitamin D3 (1,25(OH)2D3, calcitriol). This is the most active vitamin D3 metabolite and a high affinity ligand for the vitamin D receptor (VDR) [29,31,32].

VDR is a member of the nuclear receptor superfamily of transcription factors, which includes receptors for other hormones such as estrogen, progesterone or glucocorticoids, as well as a number of orphan receptors. Nuclear receptors present a highly conserved ligand-binding domain, which in the case of VDR fixes 1,25(OH)2D3 or its synthetic analogues with high specificity [33]. Binding of 1,25(OH)2D3 to VDR promotes the formation of complexes with RXR, the receptor for 9-*cis*-retinoic acid, and the binding of these VDR/RXR heterodimers to DNA. This leads to epigenetic changes that affect the transcription rate of hundreds of target genes involved in many cellular processes, including proliferation, differentiation and survival [29,31]. Moreover, a proportion of VDR molecules locate in the cytoplasm of some cell types where, on ligand binding, they trigger rapid, non-genomic, modulatory effects on signaling pathways by acting on kinases, phosphatases, and ion channels [34,35].

Current evidence indicates that 1,25(OH)2D3 and its derivatives modulate signaling pathways that affect cell survival, growth, and differentiation [29,36,37], key processes that are dysregulated in human cancers. One of these signaling routes is the Wnt/β-catenin pathway. This review will focus on the crosstalk between Wnt and vitamin D in solid tumors.

#### **2. Antagonism of Wnt**/β**-Catenin Signaling by 1,25(OH)2D3 in Solid Cancers**

#### *2.1. Colorectal Cancer*

Four decades ago, an epidemiological study hinted at the protective effects of vitamin D3 against CRC by indicating that high UVB exposure or life at lower latitudes, both of which result in higher vitamin D3 synthesis, lead to lower incidence of CRC [38]. Since then, a large number of epidemiological studies, experimental work performed in cultured cells and animal models, and also some, but not all, vitamin D3 supplementation human clinical trials have strongly suggested that 1,25(OH)2D3 has anticancer effects, particularly in CRC [31,37,39–44].

Our group was a pioneer in demonstrating that 1,25(OH)2D3 antagonizes the Wnt/β-catenin signaling pathway in colon carcinoma cells [45], a mechanism that could at least partly account for the protective effects of vitamin D3 observed in epidemiological and animal studies. Previously, other groups had reported a crosstalk between Wnt signaling and other nuclear receptors, such as those for retinoid acid and androgen [46,47].

Results from our laboratory showed that 1,25(OH)2D3 interferes with Wnt/β-catenin signaling in human colon carcinoma cells by at least three mechanisms (Figure 1). Firstly, ligand-activated nuclear VDR binds and sequesters β-catenin, which prevents its binding to LEF/TCF transcription factors and thus blocks β-catenin/TCF-mediated transcription of Wnt target genes [45]. VDR/β-catenin physical interaction was later confirmed in this and other cell systems and involves the *C*-terminal region of β-catenin and the activator function-2 domain of VDR [48]. Interestingly, wild-type APC potentiates VDR/β-catenin binding [49]. Lithocholic acid, a low affinity VDR ligand, also prompts this interaction, although less efficiently than 1,25(OH)2D3 [49].

Secondly, 1,25(OH)2D3 induces the expression of E-cadherin protein, which sequesters newly synthesized β-catenin at subcortical cell-cell adherens junctions, thus avoiding its translocation to the nucleus and β-catenin/TCF-mediated transcription [45]. Our data suggest that the small GTPase RhoA, the protease inhibitor cystatin D, the regulator of tyrosine kinase receptor signaling Sprouty-2, and the histone demethylase JMJD3 are involved in this mechanism [34,50–52]. Induction of E-cadherin by 1,25(OH)2D3 and concomitant inhibition of the Wnt/β-catenin pathway have also been reported in other cell types [53]. However, 1,25(OH)2D3 can antagonize Wnt/β-catenin signaling in colon carcinoma cells that do not express E-cadherin, which implies that this mechanism is not strictly required [45].

**Figure 1.** Schematic representation of the mechanisms by which 1,25(OH)2D3 interferes the Wnt/β-catenin signaling pathway in human CRC cells. 1,25(OH)2D3 binds to its high affinity receptor VDR inducing the formation of β-catenin/VDR complexes and thus preventing that of transcriptionally active β-catenin/TCF4 complexes. In addition, 1,25(OH)2D3 increases the transcription of the *CDH1* gene encoding E-cadherin, which sequesters newly synthesized β-catenin protein at the subcortical adherens junctions. Furthermore, 1,25(OH)2D3 upregulates the expression of the negative regulators of the Wnt/β-catenin pathway *TCF7L2* (encoding TCF4), *DKK1* and *AXIN1*. 1,25(OH)2D3 also antagonizes the pathway by reducing the secretion by nearby macrophages of IL-1β, which inhibits GSK-3β activity in CRC cells leading to an increase in β-catenin levels.

Thirdly, 1,25(OH)2D3 promotes the expression of Dickkopf 1 (DKK1), a member of a family of extracellular inhibitors of the Wnt/β-catenin pathway [54]. DKK1 can inhibit Wnt/β-catenin signaling by two mechanisms. On the one hand, DKK1 direct binding to LRP5/6 blocks Wnt-LRP interaction [55]. On the other, DKK1 can engage a ternary complex with LRP5/6 and Kremen receptors, which prompts rapid endocytosis and removal of LRP5/6 from the plasma membrane [56]. In addition to a Wnt inhibitor, *DKK1* is a β-catenin/TCF target gene [57–59]. Although most CRC carry mutations that activate the Wnt/β-catenin pathway downstream of DKK1, evidence suggests that this extracellular inhibitor has antitumor effects that are independent of β-catenin/TCF transcriptional activity [60–62]. Supporting the relevance of DKK1 in CRC, we and others have demonstrated that DKK1 expression is frequently downregulated in this neoplasia [59], in part due to gene promoter hypermethylation [62–64]. Moreover, the expression levels of VDR and DKK1 in human CRC biopsies directly correlate [54], and dietary vitamin D intake is inversely associated with *DKK1* promoter methylation in a large cohort of CRC patients [65].

DKK4 is another member of the Dickkopf family of Wnt extracellular inhibitors, although it is a weaker Wnt antagonist than DKK1. We reported that 1,25(OH)2D3 downregulates the expression of DKK4 in both human colon and breast cancer cells. Accordingly, a significant inverse correlation between DKK4 and VDR expression exists in human CRC biopsies [66]. Unexpectedly, overexpression of DKK4 in human CRC cells enhances their migratory, invasive, and angiogenic potential [66]. These effects are probably unrelated to Wnt/β-catenin inhibition and imply additional mechanisms of action of DKK4. In this regard, we and others found that DKK4 transcripts are overexpressed in human CRC samples and in biopsies from patients with inflammatory bowel disease [66–68]. These data suggest that downregulation of DKK4 by 1,25(OH)2D3 may be another mechanism for the antitumor action of 1,25(OH)2D3 in CRC.

The *c*-*MYC* oncogene is a well-known β-catenin/TCF target gene that is frequently deregulated in human cancers and activates genetic programs that orchestrate biological processes to promote cell growth and proliferation [69]. Therefore, targeting the function of MYC oncoproteins holds the promise of achieving new, effective anticancer therapies that can be applied to a broad range of tumors [70]. In particular, mutational and integrative analyses have stressed the essential role of *c*-*MYC* in CRC [16]. A study reported by Meyer and colleagues using chromatin immunoprecipitation assays followed by high-throughput DNA sequencing (ChIP-Seq) in the CRC cell line LS180 concluded that β-catenin/TCF4 and VDR/RXR heterodimers colocalize at 74 sites near a limited set of genes that included *c*-*MYC* and *c*-*FOS* oncogenes [71]. These data strongly suggest a direct action of both complexes at these gene *loci*. In fact, 1,25(OH)2D3 effects on *c*-*MYC* gene expression may count as another mechanism of crosstalk between 1,25(OH)2D3 and Wnt/β-catenin signaling pathways. Firstly, ligand-activated VDR represses *c*-*MYC* expression by direct interaction with two vitamin D response elements (VDRE) in the promoter region [71,72]. Secondly, the antagonism exerted by 1,25(OH)2D3 on Wnt/β-catenin signaling impairs the transcription of *c*-*MYC* mediated by β-catenin/TCF complexes through their binding to several Wnt responsive elements (WRE) at the *c*-*MYC* promoter [45,73].

Some authors have proposed additional mechanisms of 1,25(OH)2D3 crosstalk with Wnt/β-catenin signaling in CRC cells (Figure 1). Beildeck and colleagues showed that 1,25(OH)2D3 increases TCF4 expression in several human CRC cell lines. The effect is indirect but completely dependent on VDR [74]. Tang and colleagues have reported that TCF4 functions as a transcriptional repressor that restricts CRC cell growth [75]. Therefore, 1,25(OH)2D3/VDR-mediated upregulation of TCF4 possibly has a protective effect on CRC. Furthermore, 1,25(OH)2D3/VDR induces expression of the negative regulator of the Wnt/β-catenin pathway AXIN1 in CRC cells through a VDRE localized in the regulatory region of the gene [76]. In addition, Gröschel and colleagues found that 1,25(OH)2D3 reduces nuclear β-catenin levels in LT97 colon microadenoma cells and thus downregulates the expression of Wnt target genes such as *BCL2*, *CCND1*/cyclin D1, *SNAI1*, *CD44*, and *LGR5* [77]. Moreover, in healthy colon of mice fed a high vitamin D diet, β-catenin protein expression is decreased and the same effect is observed for TCF4 [77], which contrasts with the results of Beildeck and colleagues [74].

Kaler and colleagues described a paracrine mechanism that involves not only a crosstalk between 1,25(OH)2D3 and Wnt/β-catenin signaling pathways but also between carcinoma cells and the tumor microenvironment (Figure 1). They demonstrated that colon carcinoma cells induce the release of interleukin-1β (IL-1β) from macrophagic THP-1 cells in a process that requires constitutive activation of STAT1 [78]. Secreted IL-1β then acts on colon carcinoma cells where it triggers the inactivation of GSK-3β and thus the stabilization of β-catenin and subsequent expression of Wnt target genes. 1,25(OH)2D3 interrupts this crosstalk by blocking the constitutive activation of STAT1 and thus the production of IL-1β in macrophages in a VDR-dependent manner, which hampers the ability of

macrophages to activate Wnt/β-catenin signaling in CRC cells [78]. The possibility that this mechanism works in vivo with tumor-associated macrophages is highly interesting.

Our group has also studied the interplay between 1,25(OH)2D3 and Wnt3A (an activator of the Wnt/β-catenin pathway) in human colon fibroblasts. Both agents strongly modulate the gene expression profile and phenotype of these cells. However, in contrast to the antagonism exerted by 1,25(OH)2D3 on the Wnt/β-catenin pathway in colon carcinoma cells, they have a partially overlapping effect. Both compounds inhibit fibroblast proliferation and migration, but while 1,25(OH)2D3 reduces, Wnt3A increases fibroblast capacity to remodel the extracellular matrix [79]. In addition, in contrast to the effects observed in established colon carcinoma cell lines, 1,25(OH)2D3 does not affect the expression of key genes of the Wnt/β-catenin pathway (*AXIN2*, *CCND1*, *DKK1* and *c*-*MYC*) in human colon tumor or normal organoids derived from CRC patients, where only the *DKK4* Wnt/β-catenin target gene is repressed by 1,25(OH)2D3 [80]. This shows that antagonism of the Wnt/β-catenin pathway is not a universal action of 1,25(OH)2D3 in tumor contexts. Moreover, 1,25(OH)2D3 cooperates with Wnt factors in the differentiation of bone (osteoblasts), skin (keratinocytes) and brain (neuronal precursors) cells under physiologic conditions [81,82]. Together, available data indicate a mostly repressive action of 1,25(OH)2D3 on overactivation of the Wnt/β-catenin pathway with different effects in particular scenarios.

The interplay between 1,25(OH)2D3 and Wnt/β-catenin signaling in CRC has also been studied in vivo in animal models and patients. Our group showed that the 1,25(OH)2D3 analogue EB1089 reduces the growth of xenografts generated by human CRC cells in immunosuppressed mice. In line with data obtained in cell cultures, this inhibition is associated with an increase of E-cadherin and DKK1 levels, and a decrease of β-catenin nuclear content and of the expression of the β-catenin/TCF target gene *ENC1* in the xenografts [54,83,84]. Likewise, the antitumor action of 1,25(OH)2D3 on chemically induced mouse intestinal tumors is concomitant with increased expression of E-cadherin and the inhibition of β-catenin/TCF target genes such as *c*-*Myc* and *Ccnd1*/cyclin D1 in the intestinal crypts of these animals [85,86]. Concordantly, Xu and colleagues reported that 1,25(OH)2D3 and two of its analogues reduce the tumor load in the *Apc*min/+ mouse model of intestinal tumorigenesis associated with an increase of E-cadherin protein and a decrease of nuclear β-catenin levels and of the expression of the Wnt target genes *c*-*Myc* and *Tcf1* [87].

A Western-style diet that is high in fat and low in calcium and vitamin D is a risk factor for gastrointestinal carcinogenesis. This diet increases the frequency of intestinal tumors in normal mice and speeds up tumor formation in mouse models for intestinal cancer [88]. Several groups have shown that a Western-style diet alters components of the Wnt/β-catenin pathway in intestinal epithelial cells of normal mice [88,89]. These effects can be reversed by calcium and vitamin D supplementation, which prevents the increase of β-catenin/TCF transcriptional activity and reduces the expression of β-catenin, Ephb2 and Frizzled-2, -5, and -10 [89,90].

*Vdr* knockout mice have also been used to study the role of the vitamin D pathway on CRC. Larriba and colleagues [91] and Zheng and colleagues [92] generated *Apc*min/+ *Vdr*−/<sup>−</sup> mice and discovered that the absence of Vdr results in a higher tumor load and an increased number of premalignant lesions. Interestingly, nuclear staining of β-catenin and expression of Wnt target genes *Ccnd1*/cyclin D1 and *Lef1* are higher in *Apc*min/+ *Vdr*−/<sup>−</sup> than in *Apc*min/+ *Vdr*+/+ mice. This suggests that *Vdr* inactivation facilitates intestinal tumorigenesis fostered by Wnt/β-catenin activation [91,92].

Remarkably, in a randomized, double-blinded, placebo-controlled clinical trial, Bostick's group reported that vitamin D supplements increase the expression of APC, E-cadherin and other differentiation markers, and decrease that of β-catenin in the upper part of the crypt of normal rectal mucosa from sporadic colorectal adenoma patients [93–96]. In addition, a recent study with 67 CRC patients has revealed that a high circulating 25(OH)D3 level associates with low promoter methylation of secreted frizzled-related protein 2 (*SFRP2*) gene that encodes a soluble inhibitor of the Wnt/β-catenin pathway [97]. These data support an inhibitory effect of vitamin D on Wnt signaling in the human colon in vivo.

#### *2.2. Other Solid Tumors*

Although Wnt signaling was first described as inducing breast tumors in mice [98] and Wnt/β-catenin signaling is activated in a proportion of multiple subtypes of human breast cancers [10,99], the typical mutations in components of the pathway found in CRC (*APC*, *CTNNB1*, *AXIN2*) are rare in breast carcinomas [27]. The elevated level of nuclear β-catenin and Wnt signaling in these tumors may be due to high expression of Wnt factors in the tumor environment, loss of APC, Wnt inhibitors (DKK1, SFRPs), and/or E-cadherin expression by epigenetic modification/gene silencing, or alterations in the expression of other genes that encode constituents of the pathway (*RSPO2*, *FZD6*) [27].

Notably, nuclear β-catenin accumulation in a subset of triple-negative and basal-like breast cancer subtypes has been associated with a poor patient outcome [10,99]. Our group has shown that 1,25(OH)2D3 downregulates the expression of myoepithelial/basal markers, such as P-cadherin, smooth muscle α-actin, and α6 and β4 integrins in a panel of breast carcinoma cells, and that *Vdr*−/<sup>−</sup> mice express higher levels of P-cadherin and smooth muscle α-actin in the mammary gland than *wt* littermates [100]. These results suggest that 1,25(OH)2D3/VDR antagonizes the Wnt/β-catenin pathway in breast cancer cells, which might protect against the triple-negative and basal-like phenotype. In line with this, 1,25(OH)2D3 induces DKK1 expression and reduces β-catenin transcriptional activity in R7 murine breast cancer cells, and *Vdr* deletion and 1,25(OH)2D3 treatment increases and inhibits, respectively, the tumor expression of several Wnt/β-catenin target genes in breast cancer mouse models [101,102]. The capacity of 1,25(OH)2D3 to inhibit spheroid formation by breast cancer stem cells is overcome by β-catenin overexpression, which suggests that inhibition of the Wnt/β-catenin pathway is essential for this action of 1,25(OH)2D3 [101]. Furthermore, Zheng and colleagues have reported that VDR overexpression in a stem cell-enriched subpopulation of MCF-7 breast cancer cells inhibits Wnt/β-catenin signaling and increases cell sensitivity to tamoxifen [103]. Surprisingly, however, in another study the stable knockdown of *VDR* expression leads to attenuation of the Wnt/β-catenin pathway in MDA-MB-231 breast cancer cells: cytoplasmic and nuclear levels of β-catenin are reduced with the subsequent downregulation of its target genes *AXIN2*, *CCND1*/cyclin D1, *IL6*, and *IL8* [104].

Long non-coding RNA colon cancer-associated transcript 2 (*CCAT2*) is upregulated in ovarian cancer cells and promotes epithelial-mesenchymal transition (EMT) at least partially through the Wnt/β-catenin pathway. *CCAT2* knockdown represses the expression of β-catenin and the activity of TCF/LEF factors and inhibits EMT by upregulating E-cadherin and downregulating N-cadherin, Snail1, and Twist1 [105]. Of note, 1,25(OH)2D3 inhibits CCAT2 expression in ovarian cancer cells concomitantly with a reduction in cell proliferation, migration, and invasion. This is linked to decreased binding of TCF4 to the *c*-*MYC* promoter and, thus, to repression of *c*-MYC protein expression [106]. Thus, inhibition of CCAT2 represents a novel mechanism of Wnt/β-catenin antagonism by 1,25(OH)2D3. In addition, Srivastava and colleagues have shown that 1,25(OH)2D3/VDR can deplete ovarian cancer stem cells via inhibition of the Wnt/β-catenin pathway [107].

A recent study has investigated whether 1,25(OH)2D3 can affect Wnt/β-catenin signaling in human uterine leiomyoma primary cells using a Wnt pathway PCR array. Up to 75% of the β-catenin/TCF target genes analyzed are repressed by 1,25(OH)2D3. Similarly, 1,25(OH)2D3 inhibits the expression of 73.3% and 77.2% of the Wnt-related genes involved in tissue polarity and cell migration, and in cell cycle, cell growth and proliferation, respectively [108]. These results suggest that not only Wnt/β-catenin but probably also Wnt non-canonical pathways are inhibited by 1,25(OH)2D3 in this cellular context.

1,25(OH)2D3 and its analogue TX527 increase β-catenin protein levels in the nucleus and at the plasma membrane in a Kaposi's sarcoma cellular model and potentiate β-catenin/VDR interaction. The net outcome is downregulation of the β-catenin/TCF target genes *c*-*MYC*, *MMP9* and *CCND1*/cyclin D1. Moreover, VE-cadherin protein and *DKK1* RNA levels are increased [109]. As in Kaposi's sarcoma cells, 1,25(OH)2D3 augments the level of total β-catenin (both cytoplasmic and nuclear pools), while it reduces that of phosphorylated β-catenin in renal cell carcinoma cells [110]. More importantly, 1,25(OH)2D3 enhances VDR/β-catenin interaction while attenuating β-catenin/TCF

binding. Accordingly, 1,25(OH)2D3 downregulates the expression of *CCND1*/cyclin D1 and *AXIN2* genes. In addition, 1,25(OH)2D3 upregulates E-cadherin expression and blocks TGFβ1-induced nuclear translocation of ZEB1, Snail1 and Twist1, which contributes to the suppression of EMT and the inhibition of cell migration and invasion [110]. Thus, the effects of 1,25(OH)2D3 on Kaposi's sarcoma and renal cell carcinoma cells are largely in agreement with those observed on CRC cells. Distinctly, in pancreatic carcinoma cells, the 1,25(OH)2D3 analogue calcipotriol inhibits Wnt/β-catenin signaling by a different mechanism: the promotion of lysosomal degradation of the Wnt membrane receptor LRP6 [111].

Concomitant with an increase in Wnt/β-catenin signaling, global and epidermal-specific *Vdr* deletion predispose mice to either chemical [112] or long-term UVB-induced [113,114] skin tumor formation. 1,25(OH)2D3 enhances β-catenin binding to E-cadherin at the plasma membrane, which promotes epidermal cell differentiation. Moreover, VDR competes with LEF/TCF to recruit β-catenin to gene promoters [115,116] and both 1,25(OH)2D3 and VDR suppress β-catenin-stimulated LEF1/TCF-driven reporter activity [116,117]. The 1,25(OH)2D3 analogue EB1089 prevents the development of β-catenin-induced trichofolliculomas, while β-catenin activation in the absence of Vdr results in basal cell carcinomas [115]. Recently, Muralidhar and colleagues analyzed 703 primary melanoma transcriptomes and found that high tumor *VDR* expression is associated with upregulation of pathways mediating antitumor immunity and downregulation of proliferative pathways, notably Wnt/β-catenin [118]. Functional validation in vitro showed that 1,25(OH)2D3 inhibits the expression of Wnt/β-catenin pathway genes. These results suggest that 1,25(OH)2D3/VDR inhibits the pro-proliferative and immunosuppressive Wnt/β-catenin pathway in melanoma and that this is associated with less metastatic disease and stronger host immune responses [118].

Salehi-Tabar and colleagues have reported that *VDR* knockdown induces, while 1,25(OH)2D3 inhibits, β-catenin binding to and activation of *c*-*MYC* promoter in head and neck squamous cell carcinoma [119]. In this neoplasia, two vitamin D hydroxyderivatives, 20(OH)D3 and 1,20(OH)2D3, interfere with β-catenin nuclear translocation [120]. In a recent study, Rubin and colleagues analyzed the antitumor effects of 1,25(OH)2D3 and mitotane, the only chemotherapeutic agent available for adrenocortical carcinoma treatment. These authors reported a reduction in adrenocortical carcinoma cell growth and migration in response to either of the two agents, which is stronger when they are combined [121]. 1,25(OH)2D3 triggers a decrease in β-catenin RNA and nuclear protein levels, and both 1,25(OH)2D3 and mitotane induce RNA expression of the Wnt inhibitor *DKK1*, with a more marked effect with the combined treatment, although neither of them can reduce expression of the Wnt target gene *c*-*MYC* [121].

Vitamin D deficiency has been shown to promote hepatocellular carcinoma growth in *Smad3*+/- mice via upregulation of toll-like receptor 7 expression and β-catenin activation and, accordingly, vitamin D supplementation reduced β-catenin levels [122]. In contrast, Matsuda and colleagues reported that neither dietary supplements of vitamin D nor treatment with vitamin D analogues affect tumor formation or growth in a mouse model of hepatocarcinogenesis induced by mutant β-catenin and c-*MET* overexpression. Hence, they questioned the utility of vitamin D for hepatocellular carcinoma therapy in that setting [123].

In summary, available data show that 1,25(OH)2D3 and its analogues interfere with Wnt/β-catenin signaling in a variety of human solid tumors using mechanisms that mostly resemble those observed in CRC cells (Table 1).


**Table 1.** Mechanisms of Wnt/β-catenin pathway interference by 1,25(OH)2D3 in solid cancers.

#### **3. Antagonism of 1,25(OH)2D3**/**VDR Signaling by the Wnt**/β**-Catenin Pathway**

The abovementioned data indicate that 1,25(OH)2D3 antagonizes Wnt/β-catenin signaling in several neoplasias. However, the interplay between both pathways is a two-way road, that is, activation of the Wnt/β-catenin pathway may also result in 1,25(OH)2D3/VDR inhibition.

VDR is the only high affinity receptor for 1,25(OH)2D3 and mediates most if not all 1,25(OH)2D3 effects. Thus, cellular VDR expression is the main determinant of 1,25(OH)2D3 action and its downregulation leads to 1,25(OH)2D3 unresponsiveness. VDR is expressed in most normal human cell types and tissues, but also in cancer cell lines and tumors of diverse origins. In line with the antitumor effects of 1,25(OH)2D3 observed in several neoplasias, high VDR expression in human cancer is usually a hallmark of good prognosis [29,31,36,37,124]. VDR expression and activity is regulated transcriptionally, posttranscriptionally by several microRNAs (miRs), and posttranslationally (via phosphorylation, ubiquitination, acetylation, and sumoylation) [125].

#### *3.1. Repression of VDR by Snail Transcription Factors*

Wnt/β-catenin signaling is known to promote EMT through upregulation of the expression and activity of key EMT transcription factors such as Snail1, Snail2, Zeb1 and Twist1 by several mechanisms [126,127]. Snail1 and Snail2 are phosphorylated by GSK-3βand tagged forβ-TrCP-mediated ubiquitination and subsequent proteasomal degradation [128–130]. Thus, GSK-3β inhibition in response to Wnt/β-catenin signaling results in Snail1 and Snail2 protein stabilization. Inhibition of GSK-3β also increases *SNAI1* transcription via NFκB activation [131]. Furthermore, the Wnt/β-catenin target gene *AXIN2* contributes to Snail1 protein stabilization in breast cancer cells by regulating GSK-3β localization. When levels of AXIN2 increase in response to β-catenin/TCF signaling, GSK-3β is exported from the nuclear compartment leaving Snail1 in its non-phosphorylated transcriptionally active form [132]. In addition, induction of *SNAI2* RNA levels by Wnt3 has been described in breast cancer cells [133].

Interestingly, Snail1 and Snail2 are the best-characterized transcriptional repressors of the human *VDR*gene. Our group demonstrated that Snail1 represses the expression of *VDR*by two mechanisms [83] (Figure 2). On the one hand, Snail1 inhibits *VDR* gene transcription by binding to three E-box sequences in its promoter. On the other, Snail1 reduces *VDR* RNA half-life. As a consequence, Snail1 strongly decreases the level of VDR RNA and protein and the cellular response to 1,25(OH)2D3 [83,84]. Moreover, forced expression of Snail1 in human CRC cells prevents the upregulation of E-cadherin and the subsequent cell differentiation triggered by 1,25(OH)2D3. Therefore, by repressing *VDR* and *CDH1*/E-cadherin genes, Snail1 abolishes 1,25(OH)2D3 action and favors the accumulation of β-catenin in the nucleus and the transcription of β-catenin/TCF target genes [83,84]. Later, Snail2 was found to also inhibit *VDR* gene expression in CRC cells through the same E-boxes in the promoter used by Snail1 (Figure 2). Actually, both transcription factors present an additive repressive effect on the *VDR* gene [134]. *SNAI1* and/or *SNAI2* upregulation is observed in 76% of human CRC and is associated with *VDR* downregulation [83,134–136]. Not surprisingly, the lowest *VDR* RNA levels are found in tumors with upregulation of both *SNAI1* and *SNAI2* genes [134]. *VDR* expression is also reduced in normal colonic tissue surrounding the tumor, which suggests that Snail1 expression in tumor cells promotes the secretion of factors that reduce *VDR* expression in neighboring normal cells [137]. In addition to CRC cells, Snail1 and Snail2 also repress *VDR* gene expression and antagonize the antitumor action of 1,25(OH)2D3 in human osteosarcoma and breast cancer cells [138,139]. Knackstedt and colleagues showed that downregulation of Vdr observed in the colon of a colitis mouse model is associated with an increase in the expression of Snail1 and Snail2 [140]. Altogether, these results support that activation of the Wnt/β-catenin pathway upregulates Snail1 and Snail2, which antagonizes 1,25(OH)2D3/VDR signaling by inhibiting *VDR* gene expression.

**Figure 2.** The Wnt/β-catenin signaling pathway represses VDR expression. A major mechanism of this effect is the upregulation of Snail1 and Snail2, which repress *VDR* gene transcription and decrease *VDR* RNA half-life. The Wnt/β-catenin pathway also antagonizes 1,25(OH)2D3/VDR signaling by the upregulation of *miR*-*372* and *miR*-*373*, which reduce the level of VDR RNA and protein.

#### *3.2. Posttranscriptional Repression of VDR by miRNAs*

A novel, recently described mechanism of Wnt/β-catenin-mediated antagonism of 1,25(OH)2D3/ VDR signaling involves the *miR*-*372*/*373* cluster (Figure 2). *miR*-*372*/*373* expression is induced by β-catenin/TCF in several human cancer cell lines through three TCF/LEF binding sites located in its promoter region [141]. Accordingly, this cluster of stem cell-specific miRs is dysregulated in various cancers, particularly in CRC due to the constitutive activation of the Wnt/β-catenin pathway [142–144]. Wang and colleagues have shown that overexpression of *miR*-*372*/*373* enhances the stemness of CRC cells and promotes their self-renewal, chemotherapy resistance and invasive potential [145]. These authors found that overexpression of *miR*-*372*/*373* results in upregulation of stemness-related pathways, e.g., Nanog and Hedgehog and, conversely, downregulation of differentiation-related pathways, e.g., NFκB, MAPK/ERK, and VDR. Interestingly, they demonstrated that *miR*-*372*/*373* overexpression leads to reduced expression of VDR RNA and protein in CRC cells, which contributes to the maintenance of the cancer stem cell phenotype [145]. These data suggest that the Wnt/β-catenin pathway also inhibits VDR expression through the induction of *miR*-*372*/*373*.

#### **4. Conclusions**

The Wnt/β-catenin pathway is frequently overactivated in cancer and promotes tumorigenesis, which makes it an attractive candidate for therapeutic intervention. The active vitamin D metabolite 1,25(OH)2D3, a major regulator of the human genome, cooperates with the Wnt/β-catenin pathway in the physiological control of tissues and organs such as bone, skin, and brain. Conversely, 1,25(OH)2D3 attenuates aberrant activation of the Wnt/β-catenin pathway that takes place in most CRC and in a variable proportion of other solid tumors. To do this, 1,25(OH)2D3 modulates a series of genes and mechanisms acting at different levels of the Wnt/β-catenin pathway that vary among cancer types. 1,25(OH)2D3 does not completely block the pathway but rather reduces its overactivation. This probably helps to maintain the physiological effects of Wnt/β-catenin in healthy organs, with few toxic side-effects. As expected from two crucial regulators of the organism and its necessary homeostasis, 1,25(OH)2D3 action is counterbalanced by Wnt factors.

The multilevel antagonistic action of 1,25(OH)2D3 on aberrantly activated Wnt/β-catenin signaling strongly supports the therapeutic utility of vitamin D compounds in cancer prevention and treatment.

**Author Contributions:** Conceptualization, A.M.; writing-original draft preparation, J.M.G.-S.; writing—review and editing, M.J.L. and A.M.; artwork, M.J.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** The work in the authors' laboratory is funded by the Agencia Estatal de Investigación (PID2019-104867RB-I00/AEI/10.13039/501100011033), the Agencia Estatal de Investigación—Fondo Europeo de Desarrollo Regional (SAF2016-76377-R, MINECO/AEI/FEDER, EU), the Ministerio de Economía y Competitividad (SAF2017-90604-REDT/NuRCaMeIn), and the Instituto de Salud Carlos III—Fondo Europeo de Desarrollo Regional (CIBERONC; CB16/12/00273).

**Acknowledgments:** We thank Lucille Banham and Javier Pérez for their valuable assistance in the preparation of the English manuscript and the artwork, respectively.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the writing of the manuscript or in the decision to publish it.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Wnt**/β**-Catenin Target Genes in Colon Cancer Metastasis: The Special Case of L1CAM**

#### **Sanith Cheriyamundath and Avri Ben-Ze'ev \***

Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot 76100, Israel; sanith.cheriyamundath@weizmann.ac.il

**\*** Correspondence: avri.ben-zeev@weizmann.ac.il; Tel.: +972-8-934-2422

Received: 5 October 2020; Accepted: 17 November 2020; Published: 19 November 2020

**Simple Summary:** The Wnt/β-catenin cell–cell signaling pathway is one of the most basic and highly conserved pathways for intercellular communications regulating key steps during development, differentiation, and cancer. In colorectal cancer (CRC), in particular, aberrant activation of the Wnt/β-catenin pathway is believed to be responsible for perpetuating the disease from the very early stages of cancer development. A large number of downstream target genes of β-catenin-T-cell factor (TCF), including oncogenes, were detected as regulators of CRC development. In this review, we will summarize studies mainly on one such target gene, the L1CAM (L1) cell adhesion receptor, that is selectively induced in invasive and metastatic CRC cells and in regenerating cells of the intestine following injury. We will describe studies on the genes activated when the levels of L1 are increased in CRC cells and their effectiveness in propagating CRC development. These downstream targets of L1-signaling can serve in diagnosis and may provide additional targets for CRC therapy.

**Abstract:** Cell adhesion to neighboring cells is a fundamental biological process in multicellular organisms that is required for tissue morphogenesis. A tight coordination between cell–cell adhesion, signaling, and gene expression is a characteristic feature of normal tissues. Changes, and often disruption of this coordination, are common during invasive and metastatic cancer development. The Wnt/β-catenin signaling pathway is an excellent model for studying the role of adhesion-mediated signaling in colorectal cancer (CRC) invasion and metastasis, because β-catenin has a dual role in the cell; it is a major adhesion linker of cadherin transmembrane receptors to the cytoskeleton and, in addition, it is also a key transducer of Wnt signaling to the nucleus, where it acts as a co-transcriptional activator of Wnt target genes. Hyperactivation of Wnt/β-catenin signaling is a common feature in the majority of CRC patients. We found that the neural cell adhesion receptor L1CAM (L1) is a target gene of β-catenin signaling and is induced in carcinoma cells of CRC patients, where it plays an important role in CRC metastasis. In this review, we will discuss studies on β-catenin target genes activated during CRC development (in particular, L1), the signaling pathways affected by L1, and the role of downstream target genes activated by L1 overexpression, especially those that are also part of the intestinal stem cell gene signature. As intestinal stem cells are highly regulated by Wnt signaling and are believed to also play major roles in CRC progression, unravelling the mechanisms underlying the regulation of these genes will shed light on both normal intestinal homeostasis and the development of invasive and metastatic CRC.

**Keywords:** L1; Wnt target genes; β-catenin; cell adhesion; colon cancer; NF-κB; invasion and metastasis; cancer stem cells; EMT; Lgr5

#### **1. Introduction**

Cell–cell adhesion is a basic biological process in multicellular organisms that determines cellular and tissue morphogenesis, and its disruption is a hallmark of cancer development. Aberrant signaling mediated by changes in cell–cell adhesion is a characteristic feature of invasive and metastatic cancer cells. Wnt/β-catenin signaling is a key signaling pathway that is hyperactivated in the majority of inherited colorectal cancer (CRC) patients and serves as a very useful model for studying adhesion-mediated mechanisms underlying CRC development [1,2]. This notion is supported by findings demonstrating that β-catenin plays a dual role in the cell. It is a major linker of cell–cell adhesion receptors (of the cadherin type) to the actin-cytoskeleton and, in addition, β-catenin plays a critical role in transmitting the Wnt signal to the nucleus by being a co-transcriptional activator [together with T-cell factor (TCF)] of Wnt target genes in the nucleus [3,4]. These two seemingly unrelated roles of β-catenin and the characteristic hyperactivation of Wnt/β-catenin signaling in CRC can serve as a useful system for investigating the roles of adhesion-mediated and Wnt signaling in CRC invasion and metastasis.

Wnt signaling was discovered over 40 years ago [5,6] and was first shown to play a role in determining the segmentation pattern in *Drosophila* [7]. Following these original studies, in the coming years, a role for Wnt signaling in embryonic axis determination in vertebrates was reported [8], and the potential involvement of the Wnt pathway in cancer development in humans was suggested [9]. In parallel, numerous studies addressed the identification of downstream components in the Wnt signaling pathway and discovered that inactivating mutations in the adenomatous polyposis coli (APC) gene, which is involved in β-catenin degradation, is a key step in the activation of Wnt signaling during CRC development [10]. In addition, stabilizing mutations in β-catenin against degradation by the ubiquitin-proteasomal system were also identified in a minority of CRC cases [11,12]. At this stage, an important avenue of research consisted of unraveling the target genes of Wnt/β-catenin signaling that are responsible for human CRC development. As the early steps in tumorigenesis are driven by changes that lead to uncontrolled proliferation of cells, initial studies focused on asking whether key regulators of the cell cycle (especially those leading to increased cell proliferation) are target genes of Wnt signaling and contain β-catenin/TCF binding sites in their promoter region. These studies led to the discovery of c-myc [13] and cyclin D1 [14,15] as target genes of β-catenin/TCF transactivation. Since then, hundreds of additional β-catenin-TCF target genes were discovered; for most of these genes, their role in CRC development remains to be determined [16]. Initial immunohistochemical studies of human CRC tissue did not detect a significant accumulation of β-catenin in the nuclei of early-stage CRC tissue and β-catenin localization remained mostly at cell–cell junctions in both normal colonic epithelial cells and in differentiated areas of CRC tissue [17,18]. However, at the later stages of CRC development, especially during the invasive and metastatic stages of tumor progression, a vast accumulation of β-catenin could be demonstrated, mostly in the nuclei of cancer cells [17,18], in addition to a specific expression of β-catenin target genes at the invasive areas of the tumor [19].

In this review, we will describe studies mainly on one such β-catenin-TCF target gene, the neuronal cell adhesion receptor L1CAM (L1) and its downstream targets, and its role in CRC invasion and metastasis. We will also discuss studies suggesting that some genes induced by L1 overexpression are known genes of the colonic stem cell signature that control the homeostasis of the intestinal stem cell compartment. Because CRC is believed to originate from tumorigenic intestinal stem cells [20], we hope that studies on L1 and downstream Wnt/β-catenin target genes will provide novel insights into the control of normal intestinal homeostasis and will also provide new targets for CRC therapy.

#### **2. Members of the L1 Family of Cell Adhesion Receptors Are** β**-Catenin-TCF Target Genes**

Initial DNA microarray analyses of genes induced by activated β-catenin-TCF signaling in cancer cells identified two members of the L1 family of immunoglobulin-like cell adhesion receptors, NrCAM [21,22] and L1 [19]. These findings were unexpected because both L1 and NrCAM were known to be present mostly in nerve cells, playing key roles during brain development by regulating a number of dynamic cellular processes including axonal growth, fasciculation, and pathfinding [23,24]. In previous studies, numerous point mutations were discovered in the L1 molecule that have severe

consequences on brain development, leading to mental retardation by a group of syndromes known as L1 syndrome, MASA syndrome, and X-linked hydrocephalus [25–29].

L1 is a cell adhesion transmembrane receptor, believed to act mostly by homophilic interactions with L1 on the surface of neighboring cells. L1 belongs to the superfamily of immunoglobulin-like cell adhesion receptors, containing six Ig-like domains and five fibronectin type III repeats; a transmembrane sequence; and a highly conserved (from *C. elegans* to man) cytoplasmic tail that has binding sites for ezrin, ankyrin, and other PDZ-containing proteins (Figure 1). In addition, L1 can be cleaved in the juxtamembrane region, outside the cell, by the metalloprotease ADAM10, and inside the cell, it has binding sites for the γ-secretase cleavage complex (Figure 1). Unlike cadherins that are characterized by strong homophilic interactions, L1 can interact via both homophilic and heterophilic binding to other neuronal cell adhesion molecules including neurocan, neuropilin1, axonin, and N-CAM [30]. In addition, L1 can associate with ECM components (fibronectin, laminin, tenascin) and ECM receptors (integrins) and can also bind to growth factor receptors, such as EGFR and basic FGFR [31]. Because of these numerous weak interactions of L1 with a variety of molecules, an increase in the expression of L1 in cancer cells could be advantageous for promoting the motile, invasive, and metastatic stages of tumorigenesis.

**Figure 1.** Domain structure and binding partners of L1. Note the numerous types of L1 ligands in the ectodomain as well as in the cytoplasmic tail domain of the molecule.

#### **3. The Roles of L1 in Promoting CRC Cell Proliferation, Motility, Tumorigenesis, and Metastasis**

Overexpression of L1 in 3T3 mouse fibroblasts and in human CRC cell lines results in elevated cell proliferation under stress (in the absence of serum); increased motility; invasion; tumorigenesis upon s.c injection into mice [19]; and, in the case of CRC cells, metastasis to the liver, a hallmark of human CRC progression [32]. The metalloprotease ADAM10, also a target gene of β-catenin-TCF transactivation, cleaves the ectodomain of L1 (Figure 1), thereby leading to its shedding, and promotes the rebinding of the shed L1 ectodomain to L1 molecules on the cell surface and enhances the metastatic potential of human CRC cells [32].

Immunohistochemical analysis of human CRC tissue revealed that the more differentiated areas of the tumor and the normal colonic epithelium do not express L1 [19]. L1 is exclusively expressed at the invading edge of human CRC tissue (Figure 2) in the membrane of cells that display strong nuclear β-catenin staining, indicative of a highly active β-catenin-TCF transactivation [19]. These results were recently confirmed and extended to show that, while L1 is not required for adenoma initiation, it plays multiple roles in cancer propagation, liver metastasis, and chemoresistance [33]. This study also demonstrated that L1 is not expressed in the homeostatic intestinal epithelium, but its expression is required for CRC organoid formation and metastasis initiation and growth. Finally, L1

expression was shown to be crucial for the regrowth occurring during wound healing in the intestine following injury [33]. Taken together, these results are reminiscent of the important roles played by L1 in the dynamic cellular processes occurring in nerve cells during brain development (i.e., axonal growth, pathfinding, and fasciculation) [34].

**Figure 2.** L1 is exclusively expressed at the invasive front of human colorectal cancer (CRC) tissue in cells expressing β-catenin in their nuclei. (**A**) Immunohistochemical staining of human CRC tissue for L1. Note the preferential localization of L1 in invasive areas of the tumor (black arrowheads), but not in the inner more differentiated areas of the tumor. (**B**) In contrast to L1 localization, a serial tissue section stained with anti β-catenin antibody displays a uniform staining of the same CRC tissue area. (**C**) Enlarged picture of the boxed area in (**A**) showing the membranal localization of L1. Single CRC cells invading into the stroma could also be seen (red arrowheads). (**D**) Magnified picture of the boxed area shown in (**B**) localizing β-catenin staining in both the cytoplasm and nuclei of CRC tissue cells and in the nuclei of single invasive cells (red arrowheads) at the tumor tissue edge [19]. Scale bar: (**A**,**B**) 375 μm, (**C**,**D**) 75 μm.

#### **4. Mechanisms and Downstream Targets of L1 Signaling**

Numerous studies have addressed the mechanisms underlying the downstream signaling of L1. In neuronal cells, neurite outgrowth was shown to involve the MAPK pathway by increasing the expression of MAP2 [35]. In addition, the involvement of PI3K, ERK, and Rac-1 was also implicated in L1 signaling [36–39]. More recent studies have shown that, in CRC cells [40] and in pancreatic cancer cells [41], the signaling downstream of L1 involves the NF-κB pathway. According to this model (Figure 3), the L1 signaling pathway includes the activation of the cytoskeletal protein ezrin by phosphorylation on Thr567 (by ROCK), which leads to the re-localization of ezrin from filopodia to L1 in the membrane domain (Figure 3) and requires Tyr1151 on the L1 cytodomain (Figure 3B). Point mutations in Tyr1151 abolish the tumorigenic and metastatic capacities conferred by L1 [40]. In the

next step, the IκB–NFκB complex is recruited to this multimolecular assembly, which enhances IκB phosphorylation and its degradation by the proteasome, thereby releasing NF-κB from IκB, which enables the migration of NF-κB into the nucleus and the activation of NF-κB target genes (Figure 3C). In support of this model, the activated (phosphorylated) p65 NF-κB subunit was detected in the nuclei of CRC tissue cells in invasive areas of the tumor together with L1 and ezrin expression in the membrane and cytoplasm of the same cells [40]. In addition, blocking NF-κB signaling in CRC cells expressing elevated L1 expression abolishes the properties conferred by L1 including enhanced growth and motility, tumorigenesis, and metastasis [40].

**Figure 3.** An NF-κB-ezrin signaling pathway is involved in L1 signaling in CRC cells. (**A**) The

cytoskeletal protein ezrin is recruited to the cytoplasmic tail of L1 after it is activated by ROCK phosphorylation. The binding of activated ezrin to L1 involves Tyr1151 on the L1 cytoplasmic tail. (**B**) The L1-activated ezrin complex recruits the cytoplasmic NF-κB–IκB complex and leads to increased phosphorylation of IκB. (**C**) Elevated IκB phosphorylation results in its increased degradation by the proteasome and the release of NF-κB from the complex followed by NF-κB migration into the nucleus and transcriptional activation of target genes [40].

#### **5. Genes Induced or Suppressed by L1 That A**ff**ect CRC Progression**

In the next step, we wished to determine the genes induced, or suppressed, by L1 in CRC cells via an NF-κB-dependent mechanism using cDNA microarrays and compared these gene expression patterns to those of a large set of human CRC tissue samples [42]. A rather unexpected result of these analyses was the finding that, among the genes whose expression was suppressed by L1 overexpression (and by NF-κB signaling), and was also suppressed in human CRC tissue samples, was the well-known oncogene c-KIT [42]. Reconstitution of c-KIT expression in human CRC cell lines overexpressing L1 resulted in the inhibition of the pro-metastatic properties promoted by L1 in these cells [42]. The mechanism underlying this anti-metastatic effect conferred by c-KIT also involves the NF-κB pathway, but in this case, NF-κB plays an inhibitory role by suppressing the expression of SP-1, a key transcription factor of the c-KIT gene. The inhibition of SP-1 expression resulted in decreased c-KIT levels. In addition, the reduction in c-KIT also promoted an elevation in E-cadherin levels, the growing of cells in flat epithelial-like colonies, and the inhibition of SLUG (a key transcription factor of the EMT process), suggesting a mesenchymal to epithelial conversion (MET) [42]. While these dramatic effects of c-KIT on metastasis and cell motility indicated a tumor suppressive effect played by c-KIT [42], the proliferation in vitro and in vivo (in mice) of c-KIT overexpressing CRC cells showed that c-KIT enhances tumorigenesis, thus pointing to distinct modes of action of c-KIT in early versus late phases of tumor progression. A similar result was also reported for the key oncogene c-myc [43], thus further arguing that separate pathways mediate the tumorigenic and metastatic processes by these oncogenic molecules.

Further insight into the nature of genes induced by L1 in CRC cells by the NF-κB-ezrin pathway and their role in CRC tumorigenesis was provided by the discovery of insulin like growth factor receptor 2 (IGFBP-2) among these genes [44]. IGFBP-2 overexpression mimics the effects conferred by L1 on cell proliferation, motility tumorigenesis and metastasis, and the suppression of IGFBP-2 levels in L1-overexpressing cells blocked these properties conferred by L1 in CRC cells [44]. Interestingly, IGFBP-2 forms a molecular complex with L1, further supporting the important role played by these molecules in CRC progression. A most significant finding regarding the possible role of IGFBP-2 in CRC was derived from immunohistochemical analyses of CRC tissue samples to detect the localization of IGFBP-2. We detected IGFBP-2 at increased levels throughout the human CRC tissue samples, co-localizing with the activated p65 NF-κB subunit [44]. Most importantly, in the adjacent normal colonic mucosa, IGFBP-2 was exclusively localized at the bottom of the colonic crypts (Figure 4A). Because cells in the colonic crypts, especially at the crypts bottom, contain the colonic stem cells [45], which are believed to also be progenitors of the developing human CRC [45], we have further investigated this relationship between L1-induced genes and the colonic stem cell signature genes.

**Figure 4.** Genes induced by L1 in CRC cells include intestinal stem cell signature genes. (**A**) Insulin like growth factor receptor 2 (IGFBP-2) staining of CRC tissue revealed strong staining of the tumor tissue throughout the tissue, while in the adjacent normal mucosa, IGFBP-2 staining was exclusively confined to the bottom of colonic crypts (black arrowheads). (**B**) The intestinal stem cell signature gene secreted modular calcium-binding matricellular protein-2 (SMOC-2) was detected at the bottom of colonic crypts in normal colonic mucosa (red arrowheads). (**C**) In CRC tissue, SMOC-2 was localized in more differentiated areas of the tumor with stronger staining of invasive areas of the tumor (blue arrowhead) [44,46]. Scale bar: (**A**) 250 μm, (**B**) 50 μm, (**C**) 100 μm.

#### **6. Colonic Stem Cell Signature Genes Induced by L1 in CRC Cells**

The human intestinal epithelium contains invaginating crypts that harbor, at their base, the intestinal stem cells that express the Lgr5 molecular marker [20]. This epithelium is the most frequently regenerating tissue in the body and the lifetime of intestinal epithelial cells is less than a week. The stem cells fuel a continuous generation of all differentiated colonic cell types and are believed to be the progenitors of human CRC cells [47]. Among the genes induced by L1-ezrin-NF-κB signaling, we detected (in addition to IGFBP-2) the secreted modular calcium-binding matricellular protein-2 (SMOC-2). SMOC-2 is known as a representative of the group of Lgr5<sup>+</sup> intestinal stem cell signature genes in mice [48]. We found that the induction of SMOC-2 in human CRC cells was necessary for the pro-tumorigenic properties conferred by L1 [46]. SMOC-2 overexpression could mimic the increase in cell proliferation under stress, motility, tumorigenesis, and liver metastasis and confers a more mesenchymal phenotype characterized by suppression of E-cadherin levels and an increase in the EMT-promoting transcription factor SNAIL. These properties of SMOC-2 overexpressing in CRC cells involve signaling by integrin linked kinase (ILK) [46]. In addition, we found an increase in the intestinal stem cell signature gene Lgr5 in CRC cells overexpressing SMOC-2, L1, or the p65 subunit of NF-κB [46]. Most significantly, we detected SMOC-2 exclusively at the base of normal colonic epithelial crypts (Figure 4B) and a preferential increase in its expression at invasive areas of human CRC tissue (Figure 4C).

In the next step, we identified clusterin (CLU) as a gene induced by L1 that is also expressed at increased levels in Lgr5<sup>+</sup> intestinal stem cells of the mouse [49]. CLU is a secreted highly glycosylated protein that was implicated in playing a role in a variety of human tumors and is considered to be a marker for CRC development [50]. Similar to IGFBP-2 and SMOC-2, CLU overexpression induces CRC motility and tumorigenesis, but CLU does not promote experimental liver metastasis, implying the involvement of additional factors. However, the suppression of CLU in L1-overexpressing cells dramatically reduced their metastatic potential [49]. The mechanism of L1-mediated increase in CLU does not involve the NF-κB pathway, but rather a STAT-1-mediated elevation in the expression of the transcription factor SP-1 that activates the CLU gene promoter [49].

In the search for key intestinal/colonic stem cell compartment signature genes that could be activated by L1-mediated signaling in CRC, we turned to analyze the expression of ASCL2 in L1-overexpressing CRC cells. ASCL2 is a basic-helix-loop-helix transcription factor, a target gene of Wnt/β-catenin signaling and is restricted to Lgr5<sup>+</sup> basal crypt cells in both mice and humans [48]. A recent study identified ASCL2 as the key transcriptional regulator that is induced at a very early stage during regeneration of the intestinal stem cell compartment following injury [51]. We found that overexpression of L1 in CRC cells induces the expression and nuclear accumulation of ASCL2, a decrease in E-cadherin levels, and increased levels of β-catenin in the nucleus, together with elevated β-catenin-TCF transactivation of Wnt/β-catenin target genes [52]. This downregulation of E-cadherin expression, the increase in the accumulation of nuclear β-catenin, and the transactivation of Wnt/β-catenin-TCF target genes were also reported in breast cancer cells [53]. This suggests that the replacement of E-cadherin-mediated adhesions by L1 in CRC cells is a more general characteristic of cancer cells. In addition, we found that the overexpression of ASCL2 in CRC cells could mimic the effects conferred by L1 on cell proliferation, motility, tumorigenesis, and liver metastasis (including an elevation in the intestinal stem cell signature genes Lgr5, OLFM4, and SMOC-2), while ASCL2 suppression in L1-transfected CRC cells blocked these properties conferred by L1 [52]. We detected ASCL2 in invasive areas of human CRC tissue in cells expressing increased levels of L1 (but not in normal colon mucosa) [52], indicating that L1 and ASCL2 cooperate in promoting CRC progression.

#### **7. Genes A**ff**ected by Point Mutations in the L1 Ectodomain That Regulate CRC Development**

As inherited mutations in the L1 ectodomain were shown to affect the adhesive properties of L1 and are associated with severe human brain developmental diseases [25–29,54], we searched for genes induced by L1 that are affected by specific point mutations in the L1 ectodomain and examined their role in CRC development. All the known ectodomain point mutants of L1 that we analyzed lost their ability to confer the tumorigenic and metastatic properties in CRC cells [55]. Among the genes that are specifically affected by the L1/H210Q mutation, but not by other L1 mutations in the L1 ectodomain, we identified the membrane-associated neutral endopeptidase, neprilysin (CD10) [55]. We found that the induction of CD10 by L1 that is blocked by the L1/H210Q mutation is required for the pro-tumorigenic and metastatic capacities conferred by L1 [55]. As with several other L1-induced genes, CD10 expression was dependent on an NF-κB-ezrin signaling pathway and we identified L1 and CD10 in cells localized in invasive areas of CRC tissue, suggesting that the two molecules act together in promoting the invasive properties of CRC cells [55]. The identification of genes that are specifically affected by such L1 ectodomain point mutations could provide additional targets for CRC diagnosis and therapy.

#### **8. Secreted Factors That Promote the Tumorigenesis Induced by L1 Overexpressing CRC Cells**

Because a great number of genes that are induced by L1 overexpression in CRC cells are coding for either membrane bound proteins and are exposed to the extracellular milieu, or proteins secreted into the culture medium (see above, IGFBP-2, CLU, neprilysin, SMOC-2), we conducted a proteomic analysis of the secretome from L1 expressing CRC cells. Among the proteins whose levels were increased by L1 expression in CRC cells, we studied the role of the aspartate protease cathepsin D (CTSD), a lysosomal and secreted protein, because numerous studies reported on its association with the development of cancer in various tumors [56–61]. The levels of RNA, protein, and secreted CTSD protein were increased in response to L1 expression, and this induction of CTSD was necessary for L1-mediated CRC progression and liver metastasis [62]. The overexpression of CTSD in CRC cells, in the absence of L1, could confer increased proliferation, motility, tumorigenesis, and liver metastasis in these cells [62]. Enhancing Wnt/β-catenin signaling increased the levels of CTSD, suggesting its involvement in regulating CTSD expression. CTSD was detected in more invasive areas of the tumor in both epithelial cells and the adjacent stromal compartment, but not in normal mucosa, supporting a role for CTSD in L1-mediated CRC progression [62].

Another protein whose level is elevated in the secretome of CRC cells is the ubiquitin-like interferon induced gene 15 (ISG15), which operates much like ubiquitin by conjugating to target proteins (ISGylation) [63]. We found that increased ISG15 levels were required for L1-mediated CRC progression because suppression of ISG15 expression blocked the L1-mediated increase in CRC cell motility, tumorigenesis, and metastasis [63]. The induction of ISG15 was dependent on proper L1–L1 mediated adhesions, as point mutations in the L1 ectodomain abolished its ability to induce the expression of ISG15 [63]. The induction in ISG15 by L1 was dependent on the NF-κB pathway and ISG15 was detected in CRC tumor tissue and in the adjacent stroma, but not in normal colonic mucosa, suggesting that ISG15 could serve as a therapeutic target for CRC treatment [63].

#### **9. Conclusions**

L1, a cell adhesion receptor and a target gene of Wnt/β-catenin signaling, is a key perpetuator of CRC development and metastasis. L1 is not expressed in normal homeostatic colonic mucosa, but is induced at the invasive front of CRC tissue in cells expressing the Lgr5 intestinal stem cell marker as well as during regeneration of the intestinal/colonic tissue following injury. In addition, L1 was reported to contribute to the generation of an immunosuppressive tumor microenvironment [64] and promotes chemoresistance [33,65,66]. The studies summarized above point to the numerous genes that are induced (and suppressed) during CRC progression following L1 expression. Because the level of L1 expression was shown to be a powerful prognostic factor for indicating poor survival in a variety of cancer types, L1 is considered a promising target for cancer therapy that involves blocking L1 antibodies in combination with cytostatic drugs and/or radio-immunotherapy [67–70]. The downstream target genes of L1-mediated CRC progression described here could mimic the effects conferred by L1 on the motile, tumorigenic, and metastatic properties of CRC cells. Targeting these downstream effectors of L1-mediated signaling could provide additional approaches to CRC diagnosis and therapy.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Angiogenesis-Related Functions of Wnt Signaling in Colorectal Carcinogenesis**

#### **Aldona Kasprzak**

Department of Histology and Embryology, Poznan University of Medical Sciences, Swiecicki Street 6, 60-781 Pozna ´n, Poland; akasprza@ump.edu.pl; Tel.: +48-61-8546441; Fax: +48-61-8546440

Received: 28 October 2020; Accepted: 1 December 2020; Published: 2 December 2020

**Simple Summary:** Angiogenesis belongs to the most clinical characteristics of colorectal cancer (CRC) and is strongly linked to the activation of Wnt/β-catenin signaling. The most prominent factors stimulating constitutive activation of this pathway, and in consequence angiogenesis, are genetic alterations (mainly mutations) concerning *APC* and the β-catenin encoding gene (*CTNNB1*), detected in a large majority of CRC patients. Wnt/β-catenin signaling is involved in the basic types of vascularization (sprouting and nonsprouting angiogenesis), vasculogenic mimicry as well as the formation of mosaic vessels. The number of known Wnt/β-catenin signaling components and other pathways interacting with Wnt signaling, regulating angiogenesis, and enabling CRC progression continuously increases. This review summarizes the current knowledge about the role of the Wnt/Fzd/β-catenin signaling pathway in the process of CRC angiogenesis, aiming to improve the understanding of the mechanisms of metastasis as well as improvements in the management of this cancer.

**Abstract:** Aberrant activation of the Wnt/Fzd/β-catenin signaling pathway is one of the major molecular mechanisms of colorectal cancer (CRC) development and progression. On the other hand, one of the most common clinical CRC characteristics include high levels of angiogenesis, which is a key event in cancer cell dissemination and distant metastasis. The canonical Wnt/β-catenin downstream signaling regulates the most important pro-angiogenic molecules including vascular endothelial growth factor (VEGF) family members, matrix metalloproteinases (MMPs), and chemokines. Furthermore, mutations of the β-catenin gene associated with nuclear localization of the protein have been mainly detected in microsatellite unstable CRC. Elevated nuclear β-catenin increases the expression of many genes involved in tumor angiogenesis. Factors regulating angiogenesis with the participation of Wnt/β-catenin signaling include different groups of biologically active molecules including Wnt pathway components (e.g., Wnt2, DKK, BCL9 proteins), and non-Wnt pathway factors (e.g., chemoattractant cytokines, enzymatic proteins, and bioactive compounds of plants). Several lines of evidence argue for the use of angiogenesis inhibition in the treatment of CRC. In the context of this paper, components of the Wnt pathway are among the most promising targets for CRC therapy. This review summarizes the current knowledge about the role of the Wnt/Fzd/β-catenin signaling pathway in the process of CRC angiogenesis, aiming to improve the understanding of the mechanisms of metastasis as well as improvements in the management of this cancer.

**Keywords:** colorectal cancer; Wnt/beta-catenin signaling; angiogenesis; anti-angiogenic therapy

#### **1. Introduction**

The Wnt/Frizzled (Fzd)/β-catenin signaling pathway plays a significant role in physiology and pathology (including carcinogenesis) [1–4]. Since the pioneering mouse model genetic studies and *Drosophila* as well as the discovery of the first mammalian Wnt gene (1982), the role of Wnt signaling

was mostly implied in cell growth regulation during embryonic development and maintenance of adult tissue structure [5–7]. Aberrant activation of the Wnt/β-catenin signaling pathway as well as its interactions with other pathways is characteristic for various types of carcinogenesis [6,8,9].

The elements of canonical Wnt signaling include both a range of extracellular factors (e.g., Wnts) and cytoplasmic proteins (e.g., β-catenin). Wnt ligands, which consist of more than 19 cysteine-rich secreted glycoproteins, mediate cell–cell communication and adhesion, while β-catenin acts as the main downstream effector of the pathway in a target cell [4,9–12]. The Wnt protein binding cell surface receptor complex is composed of two molecules, the Fzd protein and the single-pass transmembrane molecule, low-density lipoprotein-related protein 5/6 (LPR5/6). There are also several other transmembrane molecules that function as alternative Wnt receptors (e.g., the retinoic acid receptor (RAR)-related orphan receptor (ROR) and related to receptor tyrosine kinase (RYK)) [9]. In turn, there are also Wnt isoforms with the ability to activate the Wnt/β-catenin-independent signaling (e.g., Wnt/calcium and the Wnt/planar cell polarity pathways). Moreover, a number of secreted proteins regulating Wnt signaling have been identified (e.g., Dickkopf (DKK) family proteins, Fzd-related Proteins (FRPs), and Wnt Inhibitory Factor-1 (WIF-1)) [6,9].

It was long suggested that the Wnt/Fzd/β-catenin signaling pathway regulates the development of blood vessels in physiological and pathological conditions due to the presence of Wnt ligands (e.g., Wnt-2, -5a, -7a, and -10b), Wnt receptors (e.g., Fzd-1, -2, -3, and -5), and Wnt inhibitors (e.g., FRP-1 and -3) in vascular cells [13]. Descriptions of the biological activity of several identified human Wnt isoforms are already the subject of a number of excellent reviews [1,4,6,9,14].

Wnt/β-catenin signaling plays an especially important role in the carcinogenesis of the organs of the gastrointestinal tract in which this pathway takes part in the regulation of embryonic development as well as the homeostasis of adult tissues [1,8,15–17]. This group includes colorectal cancer (CRC), the third most commonly diagnosed tumor as well as the second leading cause of cancer-related deaths worldwide [18].

The most common clinical CRC characteristics include high levels of angiogenesis, metastasis, and chemoresistance [19]. In CRC etiology, the decisive role is attributed to the genetic changes (especially mutations of tumor suppressor genes and/or proto-oncogenes) occurring in different stages of carcinogenesis (e.g., mutation of the Adenomatous Polyposis Coli (APC) gene during the initiation, and Kirsten Rat Sarcoma Virus (KRAS, K-Ras) gene mutation during the progression of the tumorigenesis) [20]. Currently (2020), a classical *APC-KRAS-TP53* progression model, described by Fearon and Vogelstein in the 1990s [21], has been confirmed, proving that APC mutations have the highest odds of occurring early, followed by *KRAS*, loss of 17p and Tumor Protein 53 (TP53), and SMAD family member 4 (SMAD4) gene mutations [22]. Inactivating mutations of *APC* leads to constitutive activation of Wnt/β-catenin signaling and tumor development. The CRC is therefore considered a prototype example of an oncogenic function of the Wnt/β-catenin signaling [6,8,20].

The key component of the Wnt signaling is the cytoplasmic protein β-catenin, serving two important cellular functions. In the cytoplasm, it participates in a so-called destruction complex (DC), together with Axin, APC, and a two serine-threonine kinases: glycogen synthase kinase 3α/β (GSK3α/β) and casein kinase 1 α/δ (CK1 α/δ). Phosphorylation of the β-catenin N terminus represents a pre-requirement for recognition by E3-ubiquitin ligase β-TrCP, with its subsequent degradation in proteasomes. The second important cellular function of β-catenin in epithelial cells is the formation of intercellular junctions of *zonulae adherens* type, together with other catenins (α and γ) and E-cadherin. Activation of the canonical Wnt signaling inhibits β-catenin phosphorylation and protein degradation. Stabilization and cytoplasmic accumulation of β-catenin leads to its transport to the cell nucleus, resulting in the indirect regulation of transcription by the binding of sequence-specific Lymphoid Enhancer Factor/T cell Factor (LEF/TCF) DNA binding factors that upregulate target genes [9,23]. A recent meta-analysis of transcriptomic studies suggests that LEF/TCF-specific transcriptional regulation of Wnt target genes in CRC is relevant for tumor progression and metastasis [24]. It is worth noting that a subset of β-catenin transcriptional targets is LEF/TCF-independent [25].

Hence, particular actions of Wnt/β-catenin signaling can be regulated through interactions with various molecular partners including the molecules of adherent junction (E-cadherin), DC elements (axin/conductin, APC, GSK3α/β, CK1 α/δ, and β-TrCP) as well as LEF/TCF family transcription factors [9].

The *APC* and catenin β1 (β-catenin) encoding gene (*CTNNB1*) mutations are observed in familial adenomatosis polyposis and 60–90% of sporadic CRC [8,26]. Recently, splice alterations in intronic regions of *APC* and large-frame deletions in *CTNNB1* have been described, increasing Wnt/β-catenin signaling oncogenic alterations to 96% of CRC [27]. Mutations of *APC* encompassing at least two β-catenin downregulating motifs are significantly more frequent in microsatellite unstable (MSI-H) than in microsatellite stable (MSS) CRC [28]. However, the functional effects of *APC* and *CTNNB1* mutations might differ, sparking the search for other factors influencing the action of the Wnt/β-catenin signaling pathway, especially in the context of CRC treatment.

Several lines of evidence argue for the use of angiogenesis inhibition in the treatment of CRC. In the context of this paper, components of the Wnt pathway with anti-angiogenic activity are among the most promising targets for CRC therapy [6,8,20].

This review summarizes the current knowledge about the role of the Wnt/Fzd/β-catenin signaling pathway in the process of CRC angiogenesis for a better understanding of the mechanisms of metastasis as well as improvements in the management of this cancer.

#### **2. Wnt**/β**-Catenin Signaling and Colorectal Cancer–General Comments**

The link between hyperactivation of the Wnt/Fzd/β-catenin signaling and the development of colorectal cancer has been long recognized [2,19,20,29,30]. The activated Wnt/β-catenin signaling promotes CRC cell invasion and migration in vitro, subcutaneous tumor growth, angiogenesis, and liver metastases in vivo [31].

The activation of the Wnt canonical pathway causes inhibition of β-catenin phosphorylation as well as the absence of its degradation. Its stabilization and accumulation in the cytoplasm facilitate the transport of β-catenin to the cell nucleus. In the cell nucleus, β-catenin forms a complex with LEF/TCF and intensifies the expression of various target genes associated with proliferation, differentiation, migration, and angiogenesis [2,15,17,30]. In CRC progression and angiogenesis, simultaneous hyperactivation of Wnt/β-catenin signaling and inhibition of the phosphatidylinositol 3' kinase (PI3K)/Akt pathway promote nuclear accumulation of β-catenin and the Forkhead box 03 protein (FOXO3a), respectively, promoting metastasis by regulating a panel of target genes [2]. Recently, a total of 13 target genes, highly functionally correlated with β-catenin, were identified to be significantly altered in CRC [30].

Evaluation of Wnt signaling activity in CRC became a basis to indicate molecular subtypes of this cancer. Hence, based on different responses to epidermal growth factor receptor (EGFR)-targeted therapy (cetuximab), six CRC subtypes were characterized and associated with distinctive anatomical regions of the colon crypts (phenotype), with location-dependent differentiation states and Wnt signaling activity [32]. In another molecular characterization of CRC, four consensus molecular subtypes (CMSs) were indicated including CMS2 ("canonical" subtype) (37%), which is characterized as epithelial and chromosomally unstable with marked Wnt and Myc signaling activation [33]. The most recent classification, among CRC intrinsic subtypes (CRIS), indicates CRIS-D, and to a lesser extent, CRIS-E as subtypes with high Wnt activity and a bottom crypt phenotype [34].

#### **3. Typical Features of Angiogenesis in Solid Tumors (Including Colorectal Cancer (CRC))**

Angiogenesis is one of the key mechanisms of tumor development and is critical for invasive tumor growth and metastasis [31,35–37]. The notion that "sustained angiogenesis" is one of the six key processes enabling malignant growth [38], tumor progression, and is one of the commonly accepted indicators of prognosis, is still valid [39]. This process (interchangeably called neoangiogenesis) enables new blood vessel formation through sprouting and splitting from the pre-existing ones. Hence, cancer-focused research currently indicates two major types of angiogenesis: sprouting and nonsprouting (intussusceptive), dependent or independent of endothelial cell (EC) proliferation, respectively [40]. Other authors have reported six mechanisms of vascularization observed in solid tumors. These include, apart from the above-mentioned, recruitment of endothelial progenitor cells (EPCs), vessel co-option, vasculogenic mimicry (VM), and lymphangiogenesis [41]. In CRC, the two main types of angiogenesis (sprouting and nonsprouting) are most commonly described, with the addition of VM [42–44]. The "mosaic" vessels have also been reported in the xenograft of human colon adenocarcinoma cells (LS174T) and in human CRC tissues in which both ECs and tumor cells form the lumen. Potential mechanisms of mosaic vessel formation are discussed [45].

Among the cells participating in neoangiogenesis/neovascularization in CRC, EPCs, and ECs co-opted from surrounding vessels [41,46,47] as well as cancer stem cells (CSCs) are all indicated [40].

The process known as VM is based on the formation of vascular channels without ECs. It is carried out through transdifferentiation of colorectal CSCs (CRCSCs) to form vascular-tube structures (mimic the function of vessels) that facilitate tumor perfusion independently of tumor angiogenesis [40,43,44]. VM formation in CRC is promoted by the Zinc Finger E-box Binding Homeobox 1 (ZEB1) protein. Its silencing resulted in VM inhibition and vascular endothelial (VE)-cadherin downregulation in colon cancer cells (HCT116) [48]. Canonical Wnt signaling also participates in VM. It was demonstrated that in VM-positive CRC samples, the expression of Wnt3a and nuclear expression of β-catenin is increased compared to VM-negative samples. In in vitro (HT29 cells) studies as well as in the mouse xenograft model, the tube-like structure formation was confirmed with the mechanism of overregulated Wnt3a participation in this process explained (through increased expression of vascular endothelial growth factor receptor type 2 (VEGFR-2) and VE-cadherin) [11].

The best-known molecular pathway driving tumor vascularization (including CRC) is the hypoxia-adaptation mechanism. When the tumors grow to 0.2–2.0 mm in diameter, they become hypoxic and hindered in growth in the absence of angiogenesis. During the angiogenic switch, pro-angiogenic factors predominate and result in a transition from a vascularized hyperplasia to vascularized tumor, and eventually, to malignant tumor progression [46,49–51]. Pro-angiogenic proteins are produced by the tumor and stromal cells and include (i.e., VEGF, transforming growth factor (TGF), basic fibroblast growth factor (bFGF), and platelet-derived growth factor (PDGF)) [35,52,53]. The two latter growth factors are indispensable in the maintenance of the angiogenic process [35].

The best-studied pro-angiogenic factor in solid tumors is VEGF, which is important for sprouting angiogenesis as well as the recruitment of circulating EPCs to tumor vasculature [46,47,54]. Several members of the VEGF family have been described, namely the VEGF-A, B, C, D, E and placental growth factor (PlGF, PGF) [50]. These factors bind specific receptors present on the EC surface (VEGFR-1, VEGFR-2, VEGFR-3, neuropilin-1 and -2), which dimerize and activate the intracellular tyrosine kinases (TKs), conducting the angiogenesis promoting signals [41]. VEGF-dependent tumor angiogenesis appears to activate inverse and reciprocal regulation of both VEGFR-1 and VEGFR-2. The VEGFR-1 signaling is required for EC survival, while VEGFR-2 regulates capillary tube formation [55].

Increased production of VEGF follows for upregulation of the hypoxia-inducible transcription factor 1 (HIF-1) complex [56,57]. In turn, other factors regulate the HIF-1 complex. An increase in HIF1α expression was reported to be invoked by overexpression of Sine Oculis Homeobox Homolog 4 (SIX4) via Akt signaling. SIX4 also intensified VEGF-A expression by coordinating with HIF-1α in CRC, promoting angiogenesis and tumor growth both in vivo and in vitro [57]. Other pro-angiogenic genes, activated through HIF-1 binding to hypoxia response sequence element (5'-CGTG-3') in their promoters, are PDGF and TGF-α, activation of which results in blood vessel remodeling and angiogenesis [53,58]. Other HIF-1 target genes with proven roles in colon carcinoma cell invasion include vimentin, keratins 14, -18, -19, fibronectin 1, matrix metalloproteinase 2 (MMP-2), urokinase-type plasminogen activator receptor (uPAR), cathepsin D, and autocrine motility factor (AMF) [58]. In turn, HIF-1α and HIF-2α were proven to play different, or even opposing, roles in canonical Wnt signaling in colon cancer cells. Hence, while HIF-1α silencing negatively affected the stability and transcriptional activity of β-catenin,

HIF-2α knockdown did not affect β-catenin level, increasing the transcriptional activity of this protein by inducing its nuclear transport.

Participation of the Wnt/β-catenin axis in CRC angiogenesis is a complex process. It was proven that β-catenin induces VEGF-A expression (mRNA and protein) in human colon cancer cells, underlining the importance of this protein in early and stepwise events of CRC neoangiogenesis [59,60]. Furthermore, VEGF expression positively correlates with cytoplasmic β-catenin expression in tumor cells as well as with tumor progression in vivo [61]. In turn, while VEGFR-1 (Flt-1) is considered specific for ECs, it is also present and functional in different CRC cell lines [62]. Moreover, the study of Ahluwalia et al. reported strong expression of not only VEGF but also VEGFR-1 and VEGFR-2 in human CRC specimens as well as in in vitro studies (HCT116 and HT29 cells). This indicates an autocrine mechanism of action of cancer cell produced VEGF, independent of its primary function in the induction of angiogenesis [63]. Other studies indicate that in CRC, VEGF is secreted through a K-ras/PI3K/Rho/ROCK/c-Myc axis [64]. There are also reports of Wnt signaling promotion by K-ras activation as well as the cooperation of these signaling pathways in the CRC angiogenesis process [59].

In CRC cells, non-endothelial interactions between both VEGF receptor type 1 and 2 (VEGFR-1, VEGFR-2) and the Wnt/β-catenin pathway have also been reported [10,65]. Naik et al. showed that VEGFR-1 is a positive regulator of the Wnt/β-catenin pathway, functioning in a GSK3β-independent manner [65]. Inhibition of VEGFR-1 action by RNA interference (RNAi) or TK inhibitors (TKIs) in Wnt-addicted CRC cells leads to cell death via direct disruption of the Wnt/β-catenin "survival" signaling [10,65].

An interesting model of the regulating influence of Wnt signaling on cancer metabolism and angiogenesis through pyruvate dehydrogenase kinase 1 (PDK1), as a direct Wnt target gene, was demonstrated by Pate et al. They reported that Wnt/β-catenin signaling directs a metabolic program of glycolysis in colon cancer cells (as a common cancer phenotype called the Warburg effect) and affects the tumor microenvironment through increased vessel development [66].

When it comes to mechanisms of CRC angiogenesis regulated through Wnt/β-catenin signaling, a growing number of factors promoting or inhibiting this process are described [67–69].

#### **4. Factors Promoting CRC Angiogenesis via Wnt**/β**-Catenin Signaling**

Many described factors promote angiogenesis through Wnt/β-catenin signaling pathway regulation. These include Wnt pathway components and non-Wnt signaling biologically active molecules such as chemoattractant cytokines (chemokines) [70] and various enzymatic proteins including transcription factors [71–78]. The components of the Wnt pathway include agonists (e.g., B cell Lymphoma 9 protein (BCL9)) [67,79,80] as well as antagonists such as the DKK-4 (also called the Dickkopf Wnt signaling pathway inhibitor 4) [81]. An increase in DKK-4 mRNA production was observed in CRC tissues, with elevated ectopic expression of the DKK-4 protein intensifying cell migration and invasion. Moreover, conditioned media from DKK-4 expressing cells also promoted the migrative abilities of CRC as well as the formation of capillary-like tubules of human primary microvascular ECs [81].

It needs to be noted that the activity of many classical pro-angiogenic factors (e.g., VEGF-A, MMPs, inducible nitric oxide synthase (iNOS), and chemokines) is usually regulated by at least two signaling pathways (e.g., PI3K/Phosphatase and the Tensin Homolog Deleted on Chromosome Ten (PTEN)/Akt pathway and canonical Wnt/β-catenin downstream signaling). Hence, aberrant Wnt/β-catenin signaling, along with the production of nitric oxide (NO), can positively regulate tumor angiogenesis [68].

The BCL9 protein, a transcriptional Wnt/β-catenin cofactor, is the angiogenesis promoting element of the Wnt pathway in CRC [80]. An β-catenin independent function of the BCL9 was also proven, correlating with poor prognosis subtype of the CRC [82]. In the past, it has been underlined that BCL9 intensifies β-catenin-mediated transcriptional activity, independently of Wnt signaling component mutations. BCL9 knockdown enhanced the survival of the xenograft mouse model of CRC and attenuated the expression of pro-angiogenic factors (e.g., CD44, and VEGF), which resulted in a reduction of tumor metastasis and angiogenesis [67]. Hence, BCL9 is a coactivator of the β-catenin-mediated transcription that is highly expressed in tumors, but not in the physiological cells of their origin. The mechanism of BCL9 action in Wnt signaling is based on its direct binding to a unique BCL9-β-catenin binding domain [79], corresponding to its Homology Domain 2 (HD2), which contains a single amphipathic α-helix [83].

(C-X-C motif) ligand 8 (CXCL8) (also known as interleukin (IL)-8) is one of the proinflammatory chemokines produced by CRC cells at the tumor invasion front. It promotes angiogenesis through VEGF-A upregulation and cell invasion via the Akt/GSK3β/β-catenin/MMP-7 pathway, by upregulating the anti-apoptotic B-cell lymphoma protein 2 (Bcl-2) [70]. Participation of stromal cell-derived factor 1 (SDF-1) and its receptor (C-X-C chemokine receptor type 4 (CXCR4, fusin, CD184)) was also proven in the mechanisms of CRC progression. In vitro studies confirmed that stromal cell-derived factor 1 (SDF-1) induced CXCR4-positive CRC cell invasion and epithelial-mesenchymal transition (EMT) via activation of the Wnt/β-catenin signaling [84].

DEAH box protein 32 (DHX32), one of the RNA helicases, also belongs to the group of angiogenesis promoting enzymes. This transcriptional regulator enhanced the expression of VEGF-A in CRC cells, interacting and stabilizing β-catenin. Thus, the study showed that DHX32 overexpression was associated with angiogenesis in CRC as well as poor outcomes of human CRC patients [72]. Another factor, overexpression of which influences the aggressive phenotype, angiogenesis, chemoresistance, and metastasis of CRC cells, is gankyrin (PSMD10). It is a regulatory subunit of the 26S proteasome complex. A unique pathway participates in the regulation of the above-mentioned processes by gankyrin, namely the PI3K/GSK3β/β-catenin (a cross-talk between the PI3K/Akt and Wnt/β-catenin canonical signaling pathways) [19]. In turn, Cheng et al. proved the stimulating influence on CRC progression and metastasis exhibited by Uba2, a vital component of small ubiquitin-like protein SUMO-activating enzyme, occurring through the regulation of Wnt signaling and EMT enhancement [85].

A positive influence on CRC angiogenesis is also attributed to tissue transglutaminase 2 (TGM2). It was reported that silencing of TGM2 inhibited angiogenesis and suppressed the expression of MMP-2, MMP-9, Wnt3a, β-catenin, and cyclin D1 [75]. Similar results were obtained by other authors, describing a decrease in both stemness and angiogenesis through TGM2 inhibition [86]. Similarly, in the case of the Casitas B-lineage lymphoma (c-Cbl) gene encoding CBL protein, which plays a role as an E3 ubiquitin-protein ligase, it was proven that mutant *C-Cbl-Y371H* resulted in augmented Wnt/β-catenin signaling, increasing Wnt gene expression, angiogenesis, and CRC growth. Furthermore, for the regulation of nuclear β-catenin and angiogenesis, phosphorylation of c-Cbl Tyr371 is also required [73].

High aggressiveness and intense angiogenesis were also attributed to the HCT-116 CRC cells stably overexpressing Akt. In these cells, an increased expression of EMT-related transcription factors was noted including β-catenin. Akt/HCT-116 xenografts were highly aggressive and angiogenic (with high microvessel formation and increased expression of Factor VIII) compared to the pCMV/HCT-116 xenografts. Additionally, the tumors were characterized by the nuclear localization of β-catenin and lower expression of E-cadherin [71].

Among the transcription factors, interesting correlations can be observed between Wnt/β-catenin signaling and Zink Finger Transcription Factor Spalt (Sall)-like Protein 4 (SALL4). In a study of the SALL4 gene promoter, a consensus TCF/LEF-binding site within a region of 31 bp was described, possibly playing a role in the stimulation of Wnt/β-catenin signaling in various cancers (including CRC) through direct β-catenin biding and oncogene action [76,78]. In CRC cells, co-expression and correlation between SALL4 and β-catenin expression was described, promoting lymph node metastasis and advanced CRC clinical stage [78]. Recently, it was also demonstrated that SALL4 participates in the process of human umbilical vein ECs (HUVECs) angiogenesis, modulating VEGF-A expression [87].

When it comes to other transcription factors, it was demonstrated that the Forkhead Box Q1 protein (FOXQ1) protein is overexpressed in CRC and correlates with stage of tumor and lymph node

metastasis. Small iRNA knockdown experiments on the SW480 cell line weakened the aggressive potential of cancer, downregulating angiogenesis, invasion, EMT, and resistance to drug-induced apoptosis through the inhibition of nuclear translocation of β-catenin. It was also demonstrated that the expression and action of FOXQ1 were promoted by TGF-β1. Hence, CRC progression via angiogenesis was enabled by the co-operation of two signaling pathways: Wnt and TGF-β1 [77].

The influence of several plant-based compounds on CRC angiogenesis [88] as well as the connection between the activity of such compounds and Wnt/β-catenin signaling in CRC, were also investigated [89]. These compounds include the water solutions of *Aloe vera* extracts (with two active components: aloin and aloesin). It seems that the action of active *Aloe Vera* components on angiogenesis and tumor growth depends on the activity of more than one signaling pathway. It was proven that aloin promotes activation of the Wnt/β-catenin signaling as well as inhibits the Notch signaling pathway in CRC cells only in the presence of Wnt3a. In turn, aloesin directly activates Wnt signaling and inhibits the Notch pathway in a Wnt3a independent manner [89]. These results are contradictory to previous reports describing the inhibiting influence of aloin on CRC angiogenesis via signal transducer and activator of transcription protein 3 (STAT3) activation [88].

#### **5. Factors Inhibiting CRC Angiogenesis via Wnt**/β**-Catenin Signaling**

Furthermore, various factors inhibiting angiogenesis CRC via Wnt/β-catenin signaling were also described. These include antagonists of Wnt (e.g., DKK-1 genes) [90]. DKK-1 protein expression in CRC tissues was downregulated during the CRC adenoma-carcinoma sequence, correlating with the downregulation of VEGF expression and decreased microvessel density. Overexpression of DKK-1 in CRC cells in vitro (HCT116) inhibited the formation of tube-like structures and downregulated VEGF expression in HUVECs. Xenografts of DKK-1 overexpressing CRC cells have decreased microvessel density (MVD) and VEGF expression vs. the control cells [90].

Angiogenesis inhibiting factors also include tumor suppressors (e.g., tumor necrosis factor α (TNFα)-induced protein 8 like 2 (TIPE2, TNFAIP8L2)) [91]. TIPE2 plays a role in immune homeostasis and is associated with carcinogenesis on many tumors [92]. The study by Wu et al. on human rectal adenocarcinoma demonstrated that the expression of this protein was higher in tumor tissues compared to the control. However, TIPE2 overexpression increased cell apoptosis through downregulation of Wnt3a, phospho-β-catenin, and GSK3β expression in rectal adenocarcinoma cells. It was proven that TIPE2 knockdown promoted the growth of this tumor through angiogenesis modulation. The participation of TIPE2 in the regulation of proliferation, migration, invasion, and, consequently, angiogenesis involves the Wnt/β-catenin and TGF-β/Smad2/3 signaling pathways [91].

A relation between re-expression of type 1 cyclic guanosine monophosphate (cGMP)-dependent protein kinase (PKG) in metastatic colon carcinoma and reduced tumor angiogenesis was also described. In vivo studies confirmed reduced levels of VEGF in PKG-expressed tumors compared with tumors that were derived from parental SW620 cells. Moreover, PKG expression was associated with reduced levels of β-catenin in comparison with the parental cells. Administration of exogenous PKG in SW620 cells also inhibited the expression of β-catenin and resulted in a decrease of TCF-dependent transcription [93].

Another molecule inhibiting β-catenin mRNA production and promoter activity is the scaffold/matrix attachment region binding protein 1 (SMAR1). Effects inhibiting the Wnt/β-catenin signaling activity were obtained by recruiting histone deacetylase-5 to the β-catenin promoter, resulting in decreased CRC cell migration and invasion as well as indirectly inhibiting cancer progression and angiogenesis. Moreover, smaller tumor size in in vivo NOD-SCID mice correlated with the suppression of β-catenin [94].

Protein kinase C-α (PKCα) can also function as a Wnt/β-catenin inhibitor, participating in RORα phosphorylation, hence inhibiting transcriptional activity of β-catenin. The key mechanism of that Wnt/β-catenin signaling inhibition is Wnt5a/PKCα-dependent phosphorylation on SER<sup>35</sup> of RORα. Reduction of RORα phosphorylation in >70% of CRC cases appears clinically important, together with

a significant correlation of this reduction and PKCα phosphorylation in tumor samples compared to normal tissue specimens [95]. It was also proven that PKCα also phosphorylates β-catenin itself, leading to its physiological degradation in proteasomes [96]. Recent in vitro (DLD-1 cells incubated with PKCα activators) and in vivo (C57BL/6J mice) studies with knocked-out *PRKCA* (gene encoding mouse PKCα) confirmed that this kinase exerts an anti-tumor (anti-growth, stimulating cell death) effect on cancer cells [97].

Similarly, an inhibitory influence of certain plant-based compounds (known and used in traditional Chinese medicine) on angiogenesis is described. The research indicates the inhibiting function of sporamin (a Kunitz-type trypsin inhibitor, found in sweet potato (*Ipomea batatas*)) on the number and mass of tumor nodules formed in the abdominal cavity via reduction of β-catenin (mRNA and protein) and VEGF concentration in the liver of mouse xenografted with LoVo CRC cells [97]. Another such compound is Tanshinone IIA (Tan IIA, TSA), the active lipophilic component of a Chinese *Salvia miltiorrhiza Bunge* plant. The mechanism of its action in normoxic and hypoxic microenvironment conditions is based on the inhibition of TGF-β secretion via inhibition of HIF-1α, which drives angiogenesis by promoting β-catenin nuclear translocation and TCF/LEF activation [98].

A significant influence on Wnt/β-catenin signaling and downregulation of the key genes: TCF4 (transcription factor 7-like 2, TCF7L2), cyclin D1, and c-Myc in CRC are also exerted by emodin (the anthraquinone-active substance) [99,100]. This active component of the roots and bark of several plants regulates the expression of key components of Wnt signaling, namely β-catenin and TCF7L2 as well as several downstream targets of this pathway. Additionally, two new targets of emodin action, the p300 Wnt co-activator (downregulated), and the HMG-box transcription factor 1 (HBP1) repressor (upregulated) were indicated in CRC cell lines [99]. Recent research confirmed these observations through the demonstration of EMT and tumor growth inhibition. After emodin administration, a decrease in the expression of MMPs (MMP-7 and MMP-9), VEGF, N-cadherin, Snail, and β-catenin was observed together with an increase in E-cadherin mRNA expression [100].

A recent study (2020) indicated the inhibitory influence of 6-Gingerol (6-G) on mouse CRC tumorigenesis and angiogenesis, with the participation of the Wnt/β-catenin signaling [101]. The use of ginger (*Zingiber o*ffi*cinale*) extract and 6-G in therapy against cancers (including CRC) is very well known in medicine (reviewed in [102]). After 6-G exposition, downregulation of various oncogenic proteins' expression was demonstrated, including Wnt3a and β-catenin. Inhibition of angiogenesis occurred through the downregulation of the concentration of VEGF, Angiopoietin-1 (ANG-1), FGF, and growth differentiation factor 15 (GDF-15) in the colon of benzo[a]pyrene and dextran sulfate sodium (DSS)-exposed mouse [101].

Furthermore, the inhibitory action of Raddeanin A (RA), an active oleanane type triterpenoid saponin and a major compound isolated from *Anemone raddeana Regel* was also described in CRC, influencing invasion and metastasis of this cancer's cells. This process occurred via nuclear-factor kappa B (NF-κB) and STAT3 signaling pathways. However, the main signaling pathway associated with RA action seems to be the PI3K/Akt (reviewed in [103]). Inhibition of cell proliferation and tumor growth occurs through the downregulation of canonical Wnt/β-catenin and NF-κB signaling pathways. In the mechanism of Wnt pathway downregulation, suppression of phosphorylated lipoprotein-related protein 6 (p-LPR6), Akt inactivation, the release of GSK3β inhibition, and attenuation of β-catenin expression were noted [104]. It was proven that RA inhibits HUVEC proliferation, motility, migration, and tube formation as well as reduces angiogenesis in the chick embryo chorioallantoic membrane. As an anti-tumor plant-based compound, RA also inhibits angiogenesis in vitro (HCT-15 cell line) as well as in preclinical models in vivo. Mechanism of its action in CRC is based on the modulation of VEGF-mediated phosphorylation of VEGFR-2 as well as downstream focal adhesion kinase (FAK), phospholipase C γ1 (PLCγ1), Src, and Akt kinases [105].

Pan et al. demonstrated the inhibitory action of aloin (derived from *Aloe barbadensis Miller* (*Aloe vera*) leaves) on angiogenesis, mainly occurring through the inhibition of the STAT3 signaling pathway. Aloin inhibited HUVEC proliferation, migration, and tube formation in vitro as well as activation

of VEGFR-2 and STAT3 phosphorylation in ECs. After aloin administration in SW620 CRC cells, a downregulation of antiapoptotic (Bcl-xL), pro-proliferative (C-Myc), and angiogenic factors (e.g., VEGF) was also observed. Moreover, reduced tumor volumes and weight were noted in vivo (mice xenograft model) [88].

The activity of other plant-derived compounds as potential therapeutic targets will be discussed in further sections of this review.

In Table 1, a summary of pro- and anti-angiogenic activity of chosen factors influencing the Wnt/β-catenin signaling in CRC is presented.


**Table1.**Alistofknownproandanti-angiogeneticfactorsandtheirinfluenceontheregulationoftheWnt/β-cateninsignalingpathwayincolorectal(CRC)





SER—Serine; SMAR1—Scaffold/Matrix Attachment Region Binding protein 1; STAT3—Signal Transducer and Activator of Transcription Protein 3; Tan IIA/TSA—Tanshinone IIA; TCF—T cell Factor; TCF7L2—Transcription Factor 7-like 2; TGF-β—Tumor Growth Factor beta; TGM2—Tissue Transglutaminase 2; TIPE2 (TNFAIP8L2)—Tumor Necrosis Factor α (TNFα)-induced protein 8 like 2; VEGF (R)—Vascular Endothelial Growth Factor (Receptor).

#### **6. Cellular Components of Tumors in Angiogenesis-Related Functions of Wnt**/β**-Catenin Signaling in CRC**

Cells active in CRC angiogenesis mediated by Wnt/β-catenin signaling (interacting with vascular ECs) are tumor colorectal cells [109], CRC stem cells [40,64,110,111] and CRC-associated fibroblasts [35,64,106] (Figure 1). The group of cells crucial in the process of angiogenesis and metastasis promotion includes those directly associated with blood vessels, namely progenitor ECs (EPCs) [35,112], tumor-associated ECs (TECs) [37], pericytes [113], and platelets [35,53].

**Figure 1.** Angiogenesis-related functions of the Wnt/β-catenin signaling pathway in colorectal cancer (CRC). Schematic overview of the main components of the canonical and non-canonical Wnt signaling overexpressed (↑) in the main cells of the tumor (i.e., CRC cells, CRC stem cells, and cancer-associated fibroblasts (CAFs)). Various extracellular factors (e.g., Wnts) and cytoplasmic proteins (e.g., β-catenin) secreted by these cells play a stimulating (arrows) or regulating (dotted lines) role in angiogenesis. There are several other pro-angiogenic factors (e.g., VEGF, IL-6, Norrin) that interact with the Wnt pathway components to enhance angiogenesis in CRC. CRC stem cells can directly transdifferentiate into tumor endothelial cell (TECs) to form vascular-tube structures (vasculogenic mimicry). In the sprouting angiogenesis and vasculogenic mimicry, Wnt pathway-related mechanisms are well described. In turn, the role of Wnt signaling in mosaic vessel formation in CRC is poorly understood (for details see text). Abbreviations: APC—Adenomatous Polyposis Coli gene; CTNNB1—catenin β1 (β-catenin) gene; DKK—Dickkopf-related Protein; ECs—Endothelial Cells; Fzd4,8—Frizzleds 4,8 proteins; G-CSF—Granulocyte Colony-stimulating Factor; HIF-1α—Hypoxia-inducible Factor 1 α; IL-6—Interleukin 6; LRP5—Low-density Lipoprotein-related Protein 5; PGF—Placental Growth Factor; TGM2—Tissue Transglutaminase 2; VE-cadherin—Vascular Endothelial cadherin; VEGF (R)—Vascular Endothelial Growth Factor (Receptor).

#### *6.1. Tumor Cells*

β-catenin produced by the tumor directly induces VEGF production and an increase in vessel density, which was proved in the *Min*/+ mouse model. Levels of VEGF-A (mRNA and protein) upregulated by 250–300% were observed in an in vitro model, using transfection of normal colon epithelial cell line NCM460 with activated β-catenin. The relation between β-catenin and regulation of VEGF-A expression was also proven on colon cancer cell lines (HCT116, SW620), which indicates the participation of β-catenin in angiogenesis initiation. A positive correlation was demonstrated between the upregulation of VEGF-A expression and *APC* mutational status [60].

Wnts are not the only ligands of the Fzd receptors. Norrin, a non-Wnt ligand, binds selectively to Fzd4 and stimulates Wnt signaling [9]. The norrin/Fzd4 interactions are modulated via the regulation of Fzd4 expression by Wnt2 [114]. Norrin produced by colon cancer cells increases EC growth and motility in a tumor microenvironment [114,115]. In turn, ECs in the microenvironment of colorectal tumor comprise all of the components of the Norrin signaling pathway. Hence, this signaling pathway has an important role in CRC tumor microenvironment angiogenesis [115].

In CRC cells, aberrant expression of E-cadherin/β-catenin complex can be observed as well as that of other angiogenesis markers such as Syndecan-1, platelet (Endothelial) cell adhesion molecule 1 [P(E)CAM-1, CD31], and endoglin (CD105), all involved in tumor progression and prognosis. Moreover, endoglin expression in tumor cells was positively correlated with E-cadherin, β-catenin, and Syndecan-1 [116].

It was demonstrated that exosomes derived from hypoxic CRC cells promote angiogenesis. These exosomes, enriched with Wnt4, promoted the proliferation and migration of ECs through Wnt4-induced β-catenin signaling. It was proved that Wnt4 increased nuclear translocation of β-catenin in ECs. Furthermore, an increase in tumor size and angiogenesis via CRC cell-derived exosomes was also confirmed in an animal in vivo model [109].

#### *6.2. Colorectal Cancer Stem(-Like) Cells (CRCSCs)*

CSCs of human CRC are unique cell types able to maintain tumor mass, modify the tumoral microenvironment by expressing angiogenic factors and enhanced neovascularization, and survive outside of the primary tumor at metastatic sites [40,64,110,111]. These cells play an important role in tumor vasculogenesis through their ability for transdifferentiation into human colorectal carcinoma ECs as well as to generate functional blood vessels [110]. Moreover, they also play a role in VM [40]. Surface markers of CSCs have been characterized, and their role in angiogenesis of all gastrointestinal cancers (including CRC) has been discussed in great detail in recent reviews [40,111]. Furthermore, it seems that CRCSCs cooperate with pericytes during angiogenesis initiation in CRC [113].

The mutual relations between CRCSCs and the canonical Wnt/β-catenin signaling pathway are also described in the case of CRC. This signaling pathway is a master regulator of a balance between stemness and differentiation in several adult stem cell niches including colon CSCs population in intestinal crypts of Lieberkühn. The colon-crypt base is characterized by high activity of Wnt signaling, especially in the bottom third of the crypts (where CSCs reside) due to signals from the stromal microenvironment cells [1,32]. In HCT116 and HT29 sphere models, Wei et al. demonstrated the promotion of proliferation, migration, and tube formation of EPCs via VEGF secretion by spheroid cells [112]. The malignancy in CRC spheroid cells (with high CSC characteristics) was associated with increased expression of TGM2 (TG2), β-catenin, VEGF, and EMT features [86]. Many new canonical Wnt signaling gene targets on CRCSCs were also identified as components of the stem-like subtype signature described by the authors [32].

#### *6.3. Cancer-Associated Fibroblasts (CAFs)*

Cancer-associated fibroblasts, as a major component of tumor stroma, play an underestimated role in the development and progression of various solid tumors (including CRC) [117,118]. Activated CAFs isolated from CRC produce IL-6, which induces angiogenesis mainly through intensification of VEGF-A expression in these cells [119]. Among the pro-angiogenic Wnt signaling components highly enriched in colorectal cancer CAFs is the Wnt2 protein [106–108,120]. Initially, overexpression of this protein was demonstrated in CRC cells, with a knockdown of Wnt2 downregulating Wnt/β-catenin target gene expression. Furthermore, the pro-proliferative properties of this protein were also observed [121]. The role of CAFs, as the main source of Wnt2 in CRC, was first demonstrated by Kramer et al. [107]. CAF-derived Wnt2 activates canonical signaling in APC/β-catenin wild-type colon cancer cells (but not in cells with *APC*/*CTNNB1* mutations) in a paracrine manner. Fzd8, a putative Wnt2 receptor, was identified on CAFs. It was demonstrated that Wnt2 activates autocrine canonical Wnt signaling in primary fibroblasts, which was connected to the pro-migrative and pro-invasive phenotype. These studies indicate the major role of Wnt2 in the promotion of growth, invasion, and CRC metastasis in vivo [107]. Further research of this group demonstrated that Wnt2 intensifies EC migration and invasion. However, induction of the canonical Wnt pathway was only observed in a small number of cells. In turn, in the CRC xenograft model, Wnt2 overexpression led to enhanced vessel density and tumor volume. A correlation of Wnt2 levels was observed with the expression of vascular markers as well as an increase in pro-angiogenic functions of many proteins (e.g., ANG-2, IL-6, granulocyte colony-stimulating factor (G-CSF), and placental growth factor (PGF)). Three of them (IL-6, G-CSF, and PGF) have a major part in angiogenesis intensification via increased Wnt2. Hence, the authors proved the key role of Wnt2 in the formation of the active CAF phenotype in CRC, associated with the maintenance of pro-angiogenic secretome and extracellular matrix (ECM) remodeling signals [106]. The research of Aizawa et al. demonstrated that gene sets related to the Wnt signaling were highly expressed in CAFs (with Wnt2 specifically expressed). The authors observed Wnt2-induced cancer cell migration and invasion in CRC and confirmed the correlation between Wnt2 expression and clinicopathological data (including venous invasion) in CRC in vivo studies [108].

#### *6.4. Tumor-Associated (Vascular) Endothelial Cells (TECs, TVECs)*

In physiology, ECs are responsible for the formation of a semi-permeable barrier, a process enabled by the structure of intercellular connections as well as the presence of VE-cadherin (cadherin 5/CD144) and β-catenin, linking the VE-cadherin junction complex to the cytoskeleton [122,123]. The factors destroying intercellular connections in ECs also play a role in angiogenesis induction. Temporary and reversible damage of the VE-cadherin/β-catenin junctional complex was observed as a result of the activity of some inflammatory agents (e.g., histamine) [122]. A decrease in VE-cadherin expression, release of β-catenin from the complex, induction of nuclear accumulation of β-catenin, and an increase in MMP-7 mRNA expression in HUVECs were also observed after application of recombinant matrilysin (MMP-7) [124].

In tumor (including CRC) blood vessels, structural and functional changes can be observed, connected to alterations in leukocyte trafficking. It was demonstrated that VE-cadherin expression and downstream activation of the Akt/GSK3β/β-catenin signaling caused an increase in the expression of the chemokine (C-C motif) ligand 2 (CCL2) and CXCL10, which facilitate CD8+ T cell transmigration into tumor parenchyma. Restoration of proper EC junctions not only inhibits vascular leak, but also regulates immune cell infiltration into tumors [125]. The endothelial Wnt/β-catenin signaling also participates in angiogenesis through differentiation and sprouting of ECs, remodeling as well as arterio-venous specification [126,127].

The blood vessels produced within the tumor are lined by TECs, characterized by abnormal proliferation and apoptosis [35]. TECs exhibit many altered phenotypes compared with normal ECs and produce several "angiocrine factors", which promote tumor progression. One of these factors is biglycan, which is produced in highly metastatic tumors including CRC. Stages and mechanisms of tumor metastasis involving TECc as well as elements of the stromal microenvironment (cells, extracellular matrix) are well described in the literature [128]. In the case of diabetes-complicated CRC and liver metastasis, results of a recent study indicate that the expression of biglycan is particularly intense in the myxomatous stroma. Induction of its production in vitro (HT29 cells) is regulated by high sugar concentration, fatty acids, and insulin. In turn, the co-culture with mesenchymal stem cells (MSCs) resulted in enhanced stemness and EMT phenotype [129].

In the case of CRC, in contrast to normal ECs, TECs originate not only from EPCs but also from the differentiation of CSCs. It was shown that CRC cells (HCT116 line) can transform into TECs under hypoxia conditions via a VEGFR-2-dependent mechanism. These cells expressed EC markers and formed tube-like structures in vitro [130]. Characterization of CRC blood supply and the role of TECs in this type of cancer, depending on its stage and immune remodeling, can both be found in a recent review [37]. Recent studies on tumor vascular ECs (referred as TVECs) purified from CRC tissues using iTRAQ-based quantitative proteomics analysis, among several groups of differentially expressed proteins (DEPs) and signaling pathways, also indicated proteins important in angiogenesis (e.g., HIF1 and PI3K/Akt signaling pathway-related proteins) were upregulated in TVECs compared with the controls [131]. The role of EPCs in CRC angiogenesis was also emphasized, with these cells exhibiting the potential to increase the tumorigenic capacity of CRC spheroid cells through angiogenesis, making them responsible for CRC progression [112].

#### **7. Tissue Expression and Serum Levels of Wnt**/β**-Catenin Signaling Molecules–Diagnostic and Prognostic Role in CRC**

When it comes to the expression of Wnt signaling components in CRC tissues, nuclear localization of β-catenin is described in the invasive front, in close proximity of the tumor microenvironment cells (known as the β-catenin paradox) [17,26,132,133]. Such localization mostly concerns isolated, scattered tumor cells [26]. Moreover, a correlation is described between nuclear β-catenin at the invasive front of the primary tumor and liver metastases [132,133]. Nuclear accumulation of β-catenin in neoplastic cells and the blood vessels was even considered as the most powerful predictor of liver metastasis in CRC [133]. However, there are also studies of rectal cancer, which did not detect any correlation between the nuclear overexpression of β-catenin and distant metastases or disease-free survival (DFS) [29]. Apart from its localization in the invasive front, a more heterogenous distribution of β-catenin can be observed intracellularly, both in cell membranes and in the cytoplasm [15,17,26,133,134]. Serafino et al. used a multiparametric analysis of IHC expression and subcellular localization of Wnt/β-catenin upstream (e.g., β-catenin, E-cadherin) and downstream signaling components (e.g., C-Myc, cyclin D1) in an animal model (rats) of chemically-induced CRC and human samples obtained from patients with inflammatory bowel diseases (IBD) or at sequential stages of sporadic CRC. A similar trend of β-catenin expression was noted in human and rat samples, reaching maximal values of nuclear β-catenin upregulation or membranous β-catenin downregulation in high grade dysplasia vs. normal mucosa. In advanced CRC from humans, membranous β-catenin was predominant vs. nuclear β-catenin. In their conclusions, the authors state that the crucial components of the Wnt pathway could be important markers for diagnosis, prevention, and therapy in IBD and sporadic CRC, and also possess a predictive value for responsiveness to Wnt-targeting therapy [134].

It was also noted that the cytoplasmic levels of β-catenin increased in response to hypoxia [60]. Dilek et al. showed nuclear expression of β-catenin in only 26.1% of rectosigmoid tumors, also reporting positive correlation between cytoplasmic β-catenin expression and VEGF [61]. The presence of the high Wnt signaling activity observed in tumor cells localized in the closest proximity to stromal myofibroblasts suggests a significant influence of the tumor microenvironment in further promotion of the nuclear translocation of β-catenin [15,16]. However, the prognostic role of nuclear β-catenin for distant metastases in rectal cancers is still a matter of discussion [29]. Nuclear localization of β-catenin at the invasive front of CRC appears to be important in early stages of colorectal carcinogenesis. However, there is not yet a consensus on the prognostic significance of such an expression pattern. It was stated that mutations in *APC* and *CTNNB*, while crucial for constitutive Wnt pathway activation, are not sufficient for nuclear β-catenin accumulation and full action of this signaling pathway [15,16].

Another protein of the Wnt pathway, increased expression of which in tumor cells is important for the initiation of inhibition of CRC angiogenesis process, is Wnt2 [108,120,121]. Positive expression of this protein was mainly demonstrated in stromal cells (CAFs), with little presence in cancer cells themselves. In turn, expression in CAFs positively correlates with clinicopathological data (depth of tumor, lymph node metastasis, TNM stage, venous invasion, and recurrence) [108]. Zhang et al. showed a significant positive correlation between tissue expression of Wnt2, collagen type VIII (COL8A1) (i.e., produced in ECs) and worse survival outcomes in CRC patients. Hence, Wnt2 and COL8A1 were deemed as independent factors of poor CRC prognosis. Moreover, high levels of Wnt2 expression were connected to ECM receptor and focal adhesion pathways [120]. Apart from higher β-catenin (mRNA, protein) in CRC tissues compared to the control, a correlation with elevated expression of CXCR4 was observed. Furthermore, a correlation between CXCR4 expression and low E-cadherin, high N-cadherin, and high vimentin was also noted, suggesting links between the SDF-1/CXCR4 pathway and Wnt/β-catenin signaling [84].

When it comes to the role of serum concentration of Wnt signaling components, as markers of CRC angiogenesis, it was proven that serum VE-cadherin was about fourfold higher in CRC patients compared with the controls, but it was not correlated with the VEGF level and any clinicopathological data (sex, age, tumor site, lymph node metastasis, grade, the subtype of CRC). Hence, the authors suggest that these proteins can be considered as independent markers of CRC angiogenesis [135].

#### **8. Wnt**/β**-Catenin Signaling and Other Signalizing Partners in CRC Angiogenesis**

The number of known Wnt/β-catenin signaling components and other pathways interacting with Wnt signaling, regulating angiogenesis, and enabling CRC progression continuously increases [35,66,68,86,106,136,137]. The pro-angiogenic pathways include Akt [71], PI3K/GSK3β [19], RAS-extracellular signal-regulated kinase (ERK) [20,138], PI3K/Akt/I kappa B kinase (IKK), PI3K/Akt/FOXO3a [2,136], PI3K/PTEN/Akt [68], cAMP/protein kinase A [137], SDF-1/CXCR4 [84], Norrin [115], Notch and VEGF-A/VEGFR-2 [127], miR-27a-3p/RXRα [139], ECM receptor, and focal adhesion [120] signaling pathways.

In turn, anti-angiogenic signaling pathways interacting with Wnt signaling are TGF-β1 [77,98], HIF-1α/β-catenin/TCF3/LEF1 [98], and the protein kinase C-α (PKCα) signaling pathways [74,95,96,140].

#### **9. The Role of Non-Coding RNAs in Angiogenesis via Wnt Signaling in CRC**

MicroRNAs (miRNAs, MiRs) and long noncoding RNAs (lncRNAs) are two major families of non-protein-coding transcripts [141]. This group also includes circular RNAs (circRNAs), which are closed-loop RNAs formed by covalent bonds containing exons and introns [142]. The latter are generated via alternative back-splicing, which connects the terminal 5' and 3' ends of the single-stranded mRNA [143].

#### *9.1. MicroRNAs (miRNAs, miRs)*

MiRNAs are the most commonly studied form of non-coding RNAs, responsible for modulating up to 60% of protein-coding gene expression [144]. An increasing number of studies concerns the clarification of the role of micro-RNAs in CRC progression (including angiogenesis) via alteration of different signaling pathways including Wnt/β-catenin signaling (reviewed in [51,145]).

Notably, increased expression of β-catenin in CRC tissues of mice (C57BL/6Apc(min/+) and human CRC cells positively correlated with significantly upregulated miR-574-5p. This miRNA changed the expression of β-catenin and p27 (Kip1 protein) as well as intensified the migration and invasion of cancer cells. Furthermore, in CRC tissues, miR-574-5p was negatively correlated with the expression of RNA binding protein Quaking (Qki) (associated with developmental defects in vascular tissues) [146].

Another study, among 26 deregulated miRNAs in an APC-inducible cell line, identified members of the miR-17-92 cluster that were inhibited by APC. In this process, the stabilized form of β-catenin (as a result of APC mutation) bound to and activated the miR-17-92 promoter. The main mechanism by which APC exerted its tumor suppressor activity was the reduction of miR-19a, the most important member of the miR-17-92 cluster. Therefore, the expression of miR-19a correlated with the level of β-catenin in the CRC samples, and was associated with an aggressive stage of cancer [147]. MiR-92a exhibits oncogene functions, being upregulated in chemoresistant CRC cells and tissues as well as intensifies Wnt/β-catenin signaling through Kruppel-like factor 4 (KLF4), GSK3β, and DKK-3. miR-92a expression was enhanced by IL-6/STAT, directly targeting its promoter. The authors also proved that increased miR-92a resulted in increased Wnt signaling and promotion of stem-like phenotypes of CRC cells [148].

Upregulation of miR-452 in ~70% CRC tissue samples vs. normal tissues was also reported, correlating with the clinical data. This MiR-452 promotes nuclear relocalization of β-catenin and the expression of target genes (e.g., C-Myc and cyclin D1). In turn, in vitro and xenograft mice models showed that MiR-452 can activate Wnt/β-catenin signaling and promotes an aggressive CRC phenotype through direct regulation of the 3' untranslated region (3'UTR) of GSK3β. The miR-452 promoter is affected by the same transcription factors (TCF/LEF family of transcription factors). The authors conclude that a miR-452-GSK3β-TCF4/LEF1 positive feedback loop has an important role in CRC initiation and progression (including angiogenesis) [149].

Other MiRs promoting CRC proliferation, migration, invasion, and suppression of apoptosis in vitro, and in vivo include miR-27a-3p. This molecule acts through downregulation of nuclear receptor retinoid x receptor alpha (RXRα). On the tissue level, an increased expression of this MiR was demonstrated, correlating negatively with RXRα, and positively with various clinical (clinical-stage, distant metastasis, patients' survival) and histological data (tumor differentiation). The authors also noted that RXRα negatively regulates the expression of β-catenin by its ubiquitination in CRC [138]. This confirms earlier observations of the aberrant expression of β-catenin, upregulated by suppression of RXRα [150] as well as direct interactions between RXRα and β-catenin, which suppress β-catenin transcription and protein expression in CRC cells [151].

MiR-224 [152] or epigenetic silencing of miR-490-3p [153] also promotes the aggressive CRC phenotype through activation of Wnt/β-catenin signaling. Direct regulative effects of MiR-224 on the 3'UTR of GSK3β and secreted Frizzled-related protein 2 (SFRP2) genes was demonstrated, leading to the activation of Wnt signaling and nuclear localization of β-catenin. Furthermore, ectopic miR-224 expression enhanced CRC proliferation and invasion [152].

On the other hand, miR-490-3p inhibits β-catenin and suppresses cell proliferation as well as lowers cell invasiveness by repressing EMT. Its direct target was identified as the protooncogene frequently rearranged in advanced T-cell lymphoma 1 (FRAT1) protein, which is linked with nuclear accumulation of β-catenin. Furthermore, hypermethylation of the miR-490-3p promoter downregulated the expression of this miR in CRC cells. The authors conclude that alterations in the miR-490-3p/FRAT1/β-catenin pathway can play an important role in CRC progression (including angiogenesis) [153].

Antagonistic action in transactivation of Wnt signaling is also exhibited by ectopic miR-29b expression. This miR acts through downregulation of β-catenin coactivators (TCF7L2, Snail, BCL9L) in colon cancer cells (SW480). It binds the 3'UTR of BCL9L, lowering its expression and reducing nuclear translocation of β-catenin. As a consequence, MiR-29b inhibits anchorage-independent cell growth, promotes EMT reversal, and reduces the ability of CRC cell-conditioned medium to induce in vitro tube formation in ECs [154].

#### *9.2. Long-Non Coding RNAs (lncRNAs)*

Other commonly investigated molecules taking part in different stages of CRC progression (including angiogenesis) also include long non-coding RNAs [31,155–157]. These conserved, small non-coding RNAs, made up from 21–25 nucleotides, act as negative regulators of gene expression. In the context of angiogenesis, they are also known as "angiomiRs", directly or indirectly influencing this process (reviewed in [53]). Among this group of molecules, the Wnt/β-catenin signaling activating ability is attributed to lncRNA SLCO4A1-AS1. This molecule promotes β-catenin stabilization, impairing β-catenin-GSKβ interactions, and inhibiting its phosphorylation [156].

In turn, inhibition of tumorigenesis and progression (including angiogenesis and metastasis) in CRC is caused by lncRNA-CTD903 [155] and lncRNA-APC1 [158]. In CRC tissues, strong upregulation of CTD903 expression compared with adjacent normal tissues was observed. Furthermore, in the CTD903 knockdown model in CRC cell lines (RKO and SW480), both cell invasion and migration increased with EMT characteristics as well as reduced adherence ability. Downregulation of this lncRNA resulted in Wnt/β-catenin activation with increased transcription factors expression (e.g., Twist, Snail) [155], whereas overexpression of lncRNA-APC1 was sufficient to inhibit CRC cell growth, metastasis, and tumor angiogenesis by suppressing exosome production. Moreover, the results showed the oncogenic role of CRC-derived exosomal Wnt1, which acts in an autocrine manner through non-canonical Wnt signaling [158].

Inhibition of the Wnt signaling is also mediated by upregulation of lncRNA growth arrest specific 5 (lncRNA GAS5). This type of lncRNA plays a pivotal role in the prevention of angiogenesis, inhibiting invasion and CRC metastasis [31]. Other types of lncRNAs involved in Wnt signaling in CRC metastasis (e.g., colon cancer associated transcript 1/2 (CCAT-1/2), CASC11, PVT1, Wnt-regulated lincRNA-1 (WiNTRLINC1), PCAT1, and CCAL) are presented in recent reviews [157].

#### *9.3. Circular RNAs (circRNAs)*

One circRNA, namely circular decaprenyl-diphosphate synthase subunit 1 (PDSS1) was upregulated in CRC tissue compared to the control samples. All experiments showed that circPDSS1 is linked with local and distant metastasis as well as poor prognosis in CRC patients. Moreover, it was reported to stimulate angiogenesis in CRC via Wnt/ β-catenin signaling. Knockdown experiments resulted in attenuated migratory ability and angiogenesis in CRC cells. The authors noted a downregulation of Wnt/β-catenin signaling proteins including β-catenin, GSK3β, C-Myc, MMP-9, and cyclin D1 protein levels in CRC transfected with sh-cicrPDSS1 [159].

The main types of non-coding RNAs in CRC angiogenesis regulated by Wnt/β-catenin signaling-mediated mechanisms are summarized in Table 2.


*Cancers* **2020**, *12*, 3601

Frizzled-related

 Protein 2;

TCF7L2—Transcription

 Factor 7-like 2; 3'UTR—3' Untranslated

 region of gene.

#### **10. Anti-Angiogenic Therapy in CRC**

Treatment of CRC patients, especially those affected by metastatic CRC (mCRC), still poses a major challenge and requires significant treatment personalization. Different forms of anti-angiogenic therapy have been attempted, taking into account the mechanisms of CRC angiogenesis, in which a major role is played by the VEGF pathway. There have been approaches based on the application of anti-angiogenic small-molecule TKIs (e.g., sorafenib, sunitinib, vatalanib, or tivozanib), with or without chemotherapy. Furthermore, monoclonal antibodies have also been used, both anti-VEGF pathway and EGFR targeting (cetuximab and panitumumab). The effectiveness of typical anti-VEGF-R TKIs (regorafenib, famitinib, axitinib, and apatinib) turned out to greatly vary in mCRC treatment. The first, most effective multikinase inhibitor of angiogenic (including VEGFR-1, -2, -3), stromal and oncogenic receptor TK, was regorafenib [160–162]. This drug evoked the most significant effects in cases of advanced, refractory disease [161], especially in anti-angiogenic-naïve patients with chemotherapy-refractory mCRC. The therapy with regorafenib showed antitumor activity in 59 CRC patients in a single-center, single-arm phase IIb study [162]. The most recent open-label, single center, single-arm, phase 3 study indicates clinical effectiveness of another multikinase inhibitor, lanvatinib, in the therapy of unresectable mCRC patients, especially refractory or intolerant to classical chemotherapy, anti-VEGF therapy, and anti-EGFR therapy (tumor with wt-*RAS* expression) [163]. Furthermore, promising results have also been reported for another highly-selective anti-VEGFR-1, -2, and -3 small molecule, fruquintinib, which improved both overall survival (OS), and progression-free survival (PFS) in mCRA patients compared with the placebo. This TKI was approved by the China Food and Drug Administration (CFDA) (2018) for mCRC patients after at least two standard anticancer therapies [164].

However, anti-angiogenic CRC therapies (also those combined with other forms of treatment) are not fully effective, being a matter of discussion in many excellent reviews [36,160,161,165]. Many individual variations have been observed in response to anti-angiogenic factor therapies, sparking the search for new compounds and/or identification of susceptibility markers [160,161,166]. An analysis of the profile of the expression of genes important for an effective response to cetuximab (anti-EGFR-targeted agent) therapy in 80 CRC tumors allowed for the identification of six clinically relevant CRC subtypes. Each of those subtypes showed differing degrees of "stemness" and Wnt signaling [32]. Furthermore, there has been a perspective for the improvement of efficacy and more targeted treatment in the form of studies on host genetic markers (reviewed in [166]).

There are currently a few anti-angiogenic agents approved by the U.S. FDA for mCRC treatment: anti-VEGF/VEGF-R agents (e.g., bevacizumab, ziv-aflibercept, regorafenib, ramucirumab), anti-EGFR agents (e.g., cetuximab, panitumubab), or immune-check-point inhibitors (e.g., pemprolizumab, nivolumab, ipilimubab). However, bevacizumab is the only anti-angiogenic compound for the first-line treatment of mCRC (from 2004) [36,165]. The subgroup analysis from the CONCUR trial suggests that regorafenib treatment prior to targeted therapy (including bevacizumab) may improve clinical outcomes [162].

#### *Wnt*/β*-Catenin Signaling as a Potent Therapeutic Target in CRC-Associated Angiogenesis*

Apart from anti-angiogenic therapy based on the VEGF pathway, Wnt/β-catenin signaling is among the pathways offering potential sites for targeting [140,165,167,168]. Most studies aiming to establish the most efficient anti-Wnt/β-catenin therapy concerned a better understanding of the mechanisms regulating APC signaling and/or factors downstream of APC that control β-catenin stability and/or co-transcriptional activity [74,140,168,169]. The confirmed factors inhibiting the Wnt/β-catenin signaling pathway include examples of potential CRC therapeutic factors, most of them exhibiting anti-tumor activity [83]. Their effects are mainly exerted through the inhibition of cell proliferation/migration/invasion, cancer progression delay as well as the prevention of CRC metastasis [83,140,165]. Furthermore, some drugs can be used to eliminate chemotherapy-resistance [167].

The most commonly mentioned existing anti-angiogenic drugs targeting the Wnt/β-catenin pathway in CRC include non-steroidal anti-inflammatory drugs (NSAIDs) (e.g., sulindac and celecoxib), which can "bypass" many carcinogenic effects, also regulating the increased expression of PTEN and GSK3β, inhibition of Akt (and β-catenin), and MMPs as well as iNOS activation, all of which induce cancer cell apoptosis [68,167]. Other anti-inflammatory drugs (e.g., artesunate and aspirin) caused a marked reduction in preneoplastic changes in a rat model. Both drugs also downregulated Wnt/β-catenin signaling and reduced the levels of angiogenic markers like VEGF and MMP-9. These drugs inhibited cellular proliferation and resulted in pro-apoptotic effects [170].

Apart from NSAIDs, vitamin A and D derivatives also showed efficacy in the disruption of a number of signaling pathways (e.g., Erk and PI3K/Akt) including Wnt. This fact, together with the introduction of new generations of their derivatives, creates a perspective for potential new interesting clinical trials [171–174]. Vitamin D3 metabolites, which generally inhibit growth and induce differentiation of cancer cells, have been found to also exert anti-proliferative effects on CRC cell lines (LoVo, HT29, and HCT116) and clinical samples [171]. There are reports stating that the active form of vitamin D3 and its analogs inhibit proliferation, angiogenesis, migration/invasion, and induce differentiation and apoptosis in malignant cell lines including CRC cells (reviewed in [175]). One of the newfound anti-tumor effects of 1,25(OH)2D3 in human CRC occurs through the DKK-1 gene induction [176] and DKK-4 gene downregulation, both considered as novel mechanisms of Wnt signaling inhibition [81]. The use of protein-vitamin D-pectin nano-emulsion (NVD) induces cytotoxicity in CRC cells in a dose- and time-dependent manner. This compound inhibits the growth of CRC cells (HCT116 and HT29) through the regulation of proteins responsible for the G2 phase of the cell cycle (cyclins A, B1, E2, and decrease in Cdc25c) as well as encourages apoptosis. In the context of Wnt/β-catenin signaling, NVD causes a decrease in expression of β-catenin (mRNA, protein), Akt, and survivin genes in vitro as well as in vivo (mice xenograft model). NVD administration in CRC cells decreases PI3K and Akt phosphorylation as well as inhibits β-catenin production. Hence, the inhibitory effects of vitamin D derivatives on CRC cells also depend on blocking the Wnt/β-catenin signaling and its downstream targets (e.g., survivin) [173]. Therefore, NVD, as a Wnt/β-catenin inhibitor, has the potential to stop tumor invasion and metastasis processes (including angiogenesis). It was also proven that calcitriol (1α,25-dihydroxyvitamin D3), as an active vitamin D metabolite, inhibits the tumor-promoting properties of patient-derived CAFs, also modulating many types of immune cells expressing vitamin D receptor (VDR) [177]. Other mechanisms and factors increasing the anti-proliferative action of vitamin D derivatives in CRC cells (SW480) have also been examined. These include cytochrome P450 family 24 subfamily A member 1 (CYP24A1), overexpression of which can be observed in CRC. It was recently proven that CYP24A1 inhibition induces translocation of β-catenin from the nucleus to the cell membrane in SW480 cells, intensifying the inhibitory effect of 1,25(OH)2D3 on C-Myc. Methylation of this factor increased the anti-tumor effects of vitamin D in CRC [172].

In turn, when it comes to vitamin A and its derivatives, it is worth noting that the pathways directing β-catenin for proteasome degradation (in addition to p53/Siah-1/APC and Wnt/GSK3β/APC) include the RXR-mediated pathway [178,179]. It was proven that retinol decreases the levels of β-catenin and increases ubiquitinated protein in three all-trans retinoic acid (ATRA)-resistant human CRC cells (HCT-116, WiDr, and SW620). Retinol treatment lowered the transcription of the TOPFlash reporter and mRNA levels of the endogenous β-catenin target genes (cyclin D1 and C-Myc). Hence, the potential influence of retinol on colon cancer cell growth inhibition occurs through an increase in β-catenin degradation in proteasomes with the use of the RXR-mediated pathway [180]. The research of the same group confirmed that retinol administration to ATRA-resistant human CRC cells increased β-catenin and RXRα protein interactions, inducing β-catenin transport to the degradation location in the cytoplasm [179].

A growing number of novel agents targeting the Wnt pathway are subjected to clinical trials including specific small molecules [165,168]. This group includes G007-LK and G244-LM, specific tankyrase inhibitor compounds, which reduce Wnt/β-catenin signaling through the prevention of Axin degradation, resulting in the promotion of β-catenin destabilization [169]. As β-catenin is considered the primary cause of dysregulated Wnt signaling, the action of a range of its direct inhibitors and knockdown strategies was examined. However, as of today, none of them have been introduced into oncological practice [74,140]. The usefulness of current approaches to target anti-Wnt therapy against CRC is the subject of recent reviews [181]. Furthermore, ongoing clinical trials 1-2 employ novel agents affecting this signaling pathway (with Wnt as targets), for example, Wnt-974, Foxy-5, and LGK-974 [165,167].

It was proven that aberrant activation of Wnt/β-catenin signaling mediates resistance of CRC cells to irradiation and 5-FU-based chemotherapy. Higher levels of active β-catenin and increased TCF/LEF reporter activity were observed in SW1463 cells that evolved radiation resistance. It was also demonstrated that inhibition of β-catenin (via siRNAs or small-molecule inhibitor of β-catenin transcription, XAV-939), sensitized CRC cells to chemoradiotherapy [182]. Other studies in in vitro and mice xenograft tumor models showed that PI3K/Akt signaling inhibition leads to nuclear β-catenin and FOXO3a accumulation (both promoting metastasis). It was proven that nuclear β-catenin confers resistance to the FOXO3a-mediated apoptosis induced by PI3K and Akt inhibitors (API-2), with this effect reversed by XAV-939 [2].

Recently, previous observations on some Wnt/β-catenin signaling inhibitors and downstream targets involving PKCα came back to light, indirectly related to the progression of CRC and angiogenesis. It was proven that PKCα is rarely mutated in CRC samples, hence its function might be activated with no side effects for the intestinal epithelium. Additionally, PKCα activation results in increased cell death and is drug-inducible. According to the authors of the study, there are ongoing phase II clinical trials on the application of natural PKCα activators (found in the Bryozoan species *Bugula neritina*) for CRC treatment [74]. The use of a stabilized form of BCL9 α-helix (SAH-BCL9) is also suggested in potential therapy, as its administration caused dissociation of the native β-catenin/BCL9 complex as well as suppressed tumor growth and angiogenesis in the mouse xenograft model of the Colo320 CRC cell [83].

Moreover, a growing number of publications have documented the action of anti-Wnt/β-catenin signaling plant based compounds (particularly those used in traditional Chinese medicine) However, anti-angiogenic actions linked to Wnt signaling are only attributed to some of them, for example, Raddeanin [103–105] and Tanshinone IIA [39,98,183] (Table 1). Other naturally occurring compounds that inhibit Wnt signaling include thymol, derived from *Thymus vulgaris L* [184,185]. One of the mechanisms of this factor's action in CRC in vitro (HCT116 and LoVo cells) as well as in vivo is the prevention of EMT, invasion, and metastasis through the inhibition of Wnt/β-catenin signaling [186]. In turn, in the case of Radix *Tetrastigma hemsleyani* flavone (RTHF), it was proven that this compound causes downregulation of β-catenin activation and downstream protein expression (Lgr5, C-Myc, and cyclin D1). It also decreased the size of tumors in vivo in mice through the inhibition of pro-proliferative properties of the Wnt pathway [187]. Another plant-based polyphenol compound extracted from the root of *Curcuma longa*, is curcumin. This phytochemical also shows an anti-inflammatory, anti-oxidant, and anti-cancer activity [188]. In studies of CRC cells (SW480) as well as in the xenograft tumor model, Dou et al. proved anti-tumor activity of curcumin via inhibition of cell proliferation by suppression of the Wnt/β-catenin pathway. It was also noted that overexpression of miR-130a could abolish the anti-tumor activity of curcumin [189]. Recently, a study of another CRC cell line (SW620 cells) reported an inhibitory influence of curcumin on cell viability as well as the promotion of apoptosis. At the same time, an increase in the expression of Caudal Type Homeobox-2 (CDX2) and decreased β-catenin nuclear translocation were observed. In turn, the expression of downstream proteins of Wnt/β-catenin signaling (Wnt3a, C-Myc, survivin, and cyclin D1) was reduced. Furthermore, it was reported that the inhibitory action of Wnt/β-catenin in these cells occurred due to CDX2 restoration [190]. The isobavachalcone, a flavonoid extracted from *Psoralea corylifolia*, also inhibits growth and colony formation of CRC tumor cells as well as the induction of apoptosis through the inhibition of the AKT/GSK3β/β-catenin pathway have been noted [191]. Promising study results in anti-Wnt/β-catenin

signaling therapies also concern berberine (and its synthetic 13-arylalkyl derivatives) [192,193], an isoquinoline alkaloid present in several plants including *Coptis sp.* and *Berberis sp.* [194]. Special attention was given to its anti-tumor function, mediated by the inhibition of β-catenin transcriptional activity and weakening of anchorage-independent growth (decrease in E-cadherin expression) [192]. It was proven that berberine inhibits the function of β-catenin by direct binding to a unique RXRα region that contains the Gln275, Arg316, and Arg371 residues. As a result, a promotion of this receptor's interaction with nuclear β-catenin occurs, leading to c-Cbl mediated degradation of β-catenin, hence the inhibition of cell proliferation. Moreover, human CRC xenograft in nude mice also demonstrated the inhibition of tumor growth in an RXRα-dependent manner [193].

The basic drawbacks of anti-angiogenic and anti-Wnt signaling targeted therapies have been presented in several reviews [54,165]. These include the costs of treatment, extra adverse events, crossover, and bypass mechanisms between different signaling pathways and drug resistance as well as varying efficacy among patients [165]. The main challenges and complexities associated with creating the perfect therapeutic agents targeting the Wnt/β-catenin signaling pathway in CRC have been summarized by others [140].

Direct CRC angiogenesis inhibition mechanisms based on Wnt/β-catenin signaling are only described in a small number of existing or potential therapeutics. Nevertheless, previously mentioned results of studies (Sections 2–4) demonstrate a tight interaction of Wnt signaling with angiogenesis markers in CRC. It can therefore be assumed that the inhibition of upstream and/or downstream targets of Wnt signaling, apart from downregulating cell proliferation/migration/invasion, hence tumor growth and metastasis, is also a statement of angiogenesis inhibition in the tumor. More detailed information on the therapeutics targeting the Wnt/β-catenin signaling pathway in CRC can be found in existing works focused solely on this topic [140]. Table 3 summarizes the selected existing drugs and several agents under investigation for different Wnt/β-catenin targets in CRC with an indication of their influence on angiogenesis.


Selectedclassesofexisting/potentialanti-Wnt/β-cateninsignalingtherapeuticswithanti-angiogeniceffectsincolorectal

C α; RXR—retinoid

 X receptor;

PORCN—Porcupine;

PTPRK—Receptor-type

sFRP2—secreted

 Frizzled related protein 2;

Tyrosine-protein

 Phosphatase kappa; TCF4—Transcription

 Factor 4, T-cell Factor-4;

RSPO1-4—Wnt

 agonists of the R-spondin family; RTHF—Radix

TOPFlash—TCF

 Reporter Plasmid; WIF1—Wnt Inhibitory Factor 1.

*Tetrastigma hemsleyani* flavone;

#### **11. Final Remarks and Future Perspectives**

Angiogenesis belongs to the most clinical characteristics of CRC and is strongly linked to the activation of Wnt/β-catenin signaling. The most prominent factors stimulating constitutive activation of this pathway, and in consequence angiogenesis, are genetic alterations (mainly mutations) concerning *APC* and the β-catenin encoding gene (*CTNNB1*), detected in a large majority of CRC patients. These mutations lead to an intensification of CRC cell proliferation, migration, and invasion in vitro as well as tumor growth, angiogenesis, and distant metastases in vivo. In addition to the mutations mentioned, there are more and more genetic and epigenetic biomarkers used to determine CRC diagnosis, prognosis, and response to therapy, as summarized in excellent reviews [202]. There are also potential clinical applications of liquid biopsy biomarkers in CRC including circulating tumor cells, circulating tumor DNA, miRNAs, lncRNAs, and proteins from blood and body fluids, and their genomic and proteomic analyses (reviewed in [203]).

Wnt/β-catenin signaling is involved in the basic types of vascularization (sprouting and nonsprouting angiogenesis) and vasculogenic mimicry as well as the formation of mosaic vessels. In vascular cells, expression of Wnt ligands, Wnt receptors, and Wnt inhibitors has been reported. The main type of angiogenesis with the participation of Wnt signaling is currently assumed to occur through the hypoxia-adaptation mechanism mediated by VEGF-signaling and upregulation of the HIF-1 complex. β-catenin itself induces the expression of VEGF in colon cancer cells in the early steps of CRC neoangiogenesis. Furthermore, tissue VEGF expression positively correlates with the cytoplasmic expression of β-catenin in tumor cells and tumor progression in vivo. In turn, the influence of HIF-1α (increasing) and HIF-2α (decreasing) on β-catenin levels/transcriptional activity in CRC cells remains much more varied. Moreover, non-endothelial interactions between both VEGF receptor types (VEGFR-1, VEGFR-2) and Wnt/β-catenin signaling have also been reported. It was confirmed that VEGFR-1 positively regulates Wnt signaling in a GSK3β-independent manner. In contrary to the previous paradigm, the presence of both VEGF receptor types was also demonstrated on tumor CRC cells, suggesting the possibility of autocrine VEGF action.

Factors regulating angiogenesis with the participation of Wnt/β-catenin signaling include different groups of biologically active molecules, namely selected molecules belonging to Wnt family proteins (e.g., Wnt2, DKK, BCL9) as well as various factors outside the Wnt family (e.g., DHX32, gankyrin, Uba2, CXCL8, SALL4, FOXQ1, bioactive compounds of plants, etc.).

A direct influence of several pro-angiogenic factors (e.g., BCL9, SALL4) on Wnt signaling has been demonstrated (binding β-catenin) in the angiogenesis process. Other factors promoting angiogenesis (e.g., DHX32, gankyrin, Uba2, AKT) regulate Wnt signaling through β-catenin stabilization and increase Wnt gene expression as well as the intensification of EMT-related transcription factor expression (including β-catenin). This regulation results in EC migration and the formation of capillary-like tubules of human microvascular ECs. The opposite effects are evoked by the anti-angiogenic factors through the inhibition of production and transcriptional activity of β-catenin (e.g., TIPE2, SMAR1, PKG, PKCα, sporamin, emodin, 6-Gingerol, raddeanin A). Recently, an increasingly important role in Wnt signaling involving CRC angiogenesis is attributed to non-coding RNAs. A number of these molecules activate (e.g., miR-574-5p, miR-17-92, miR-92a, miR-452, miR-27a-3p, miR-224, lncRNA SLCO4A1-AS1, and circPDSS1), while other inhibit Wnt signaling (e.g., miR-490-3p, miR-29b, lncRNA-CTD903, lncRNA APC1, and lncRNAGAS5).

The active cellular components of CRC-related angiogenesis consist of tumor cells, CRC stem cells, and cancer-associated fibroblasts (CAFs) as well as cells directly linked to blood vessels (EPCs, TECs, pericytes). Moreover, complex intercellular interactions have been reported in tumors during angiogenesis. CRC cells produce β-catenin (mRNA and protein), which intensifies VEGF expression and increases vessel density. The norrin protein produced by cancer cells binds to Fzd4, regulating EC proliferation and motility. In turn, norrin/Fzd4 interactions are modulated via regulation of Fzd4 expression by Wnt2. Furthermore, exosomes enriched in Wnt4 produced by CRC cancer cells promote angiogenesis by increasing ECs proliferation and migration via Wnt signaling. Both tumor CRC cells

and CAFs (main source) produce the Wnt2 protein, which plays a major role in the initiation and maintenance of the CRC angiogenesis process. Wnt2 expression in CAFs correlates with a number of clinicopathological data (including venous invasion) of CRC patients. Wnt2 intensifies EC migration and invasion, enhanced vessel density, and tumor volume. Wnt2 expression positively correlates with the expression of vascular markers and an increase in pro-angiogenic function of many proteins (e.g., IL-6, G-CSF, and PGF). When it comes to CRC stem cells, high Wnt activity is mostly present in the bottom third of the crypts (where CSCs reside). These cells have the ability of transdifferentiation into human TECs as well as the generation of functional blood vessels.

The list of Wnt/β-catenin signaling components and pathways interacting with Wnt signaling, regulating angiogenesis, and conditioning CRC progression, continuously increases.

As β-catenin is considered as a primary cause of dysregulated Wnt signaling in CRC as a consequence of APC/*CTNNB1* mutations, there are ongoing studies on the action of a number of inhibitors of β-catenin itself as well as knockdown strategies. However, no results of such research have yet been introduced into CRC oncological practice, due to the relatively low effectiveness as well as significant intestinal toxicity. Small molecules blocking Wnt signaling in CRC also include tankyrase inhibitors (G007-LK and G244-LM). There are several clinical trials (phase 1/2, phase 1, and phase 2) on the use of novel Wnt targeting agents in CRC (e.g., Wnt-974, LGK-974, Foxy-5). Positive anti-angiogenic effects, disrupting Wnt/β-catenin signaling have been demonstrated for a number of NSAIDs (e.g., sulindac, celecoxib, artesunate, and aspirin) and vitamin A and D derivatives. Furthermore, many natural plant-derived compounds used in traditional Chinese medicine inhibits Wnt/β-catenin signaling and, directly or indirectly, CRC angiogenesis (e.g., RA, thymol, RTHF, curcumin, IBC, Tan IIA, and berberine). As for now, the available results mostly concern in vitro and mouse in vivo models.

#### **12. Conclusions**

As the reviewed literature shows, the role of aberrant Wnt/β-catenin signaling in CRC-related angiogenesis is undisputed. These activities mostly occur due to canonical APC/β-catenin pathway activation in tumor colorectal cells, CRC stem cells, cancer-associated fibroblasts and tumor ECs, intensification of β-catenin expression, and translocation to the nucleus as well as positive correlations with other typical pro-angiogenic factors (e.g., VEGF, VEGRs). Furthermore, the role of a number of active polypeptides, proteins, and non-coding RNAs is indicated in this process. However, when it comes to anti-angiogenic CRC treatments based on targeting the Wnt/β-catenin signaling, studied inhibitors of this pathway are still mostly in preclinical stages, with only a few compounds reaching phase 1 or 2 clinical trials. Individualized targeted CRC therapeutic strategies should take into account the newest findings of molecular biology, explaining the role of direct tumor cell interactions, and all pro- and anti-angiogenic factors acting on this type of signaling as well as other related pathways. An especially large number of publications in the last five years focusing on the role of Wnt/β-catenin signaling in cancer progression (including CRC) has certainly resulted in a better understanding of the mechanisms of metastasis as well as improvements in the management of this cancer.

**Author Contributions:** The author worked on the information compilation, analysis and manuscript writing. The author has read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** My sincere apologies to researchers whose primary articles could not be cited due to length constraints of this review. I wish to thank Monika Swierczewska for her assistance in the artwork. ´

**Conflicts of Interest:** The author declares no conflict of interest.

#### **Abbreviations**



#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
