WNT signaling and cancer stemness

Canonical WNT signals are transmitted through Frizzled (atypical G-protein-coupled receptor, GPCR) and LRP5/6 (LDL receptor-related proteins) to the nuclear β-catenin-TCF/LEF (T-cell factor and lymphoid enhancer factor) complex to up-regulate the expression of target genes, such as MYC, CCND1 (cyclin D1) and LGR5 (leucine rich repeat-containing GPCR 5) [1–3] (Figure 1). In contrast, noncanonical WNT signals are transmitted through Frizzled or ROR1/2 (receptor tyrosine kinase-like orphan receptors) to the PLC (phospholipase C), Rac1 and RhoA branches to activate the NFAT (nuclear factor associated with T cells), AP-1 (activator protein 1) and YAP-TEAD (Yes-associated transcriptional regulator/TEA domain transcription factor) transcription factors, respectively [1,4,5] (Figure 1).

Genetic alterations in WNT signaling molecules drive carcinogenesis in colorectal, gastric, liver, uterine and other cancers [6]. Loss-of-function alterations in the APC (adenomatous polyposis coli), AXIN1 and AXIN2 genes and gain-of-function mutations in the CTNNB1 (β-catenin) gene activate the canonical WNT signaling cascade owing to stabilization and nuclear accumulation of β-catenin [7], while loss-of-function alterations in the RNF43 (ring finger protein 43) gene can activate both canonical and noncanonical WNT signaling cascades through stabilization of plasma membrane Frizzled, LRP6 and ROR1 [8].

In the tumor microenvironment (TME), WNT ligands derived from tumor cells [9], cancer-associated fibroblasts (CAFs) [10] and immune cells [11] also activate both canonical and noncanonical WNT signaling cascades and affect the FGF (fibroblast growth factor) [1], Hedgehog [12], Notch [13] and TGFβ (transforming growth factor-β) [14] signaling cascades. The WNT signaling network in the TME regulates cancer biological behaviors, such as dormancy, drug resistance, epithelial–mesenchymal transition (EMT), immune evasion, proliferation and stemness.

Cancer stemness is defined as the self-renewal and tumor-initiation potential of cancer stem cells (CSCs) but also broadly as a cancer biological characteristic regulated by CSC signaling networks [15,16]. CSCs are involved in drug resistance, invasion, metastasis and recurrence of tumors, while evolution of CSC concepts from hierarchical organization model to cellular plasticity model have been described elsewhere. CD44, CD133 (prominin-1) and LGR5 are functional CSC markers. For example, LGR5 potentiates WNT signaling through R-spondin-mediated Frizzled stabilization [6], and Lgr5+ gastric CSCs are required for primary tumorigenesis, liver metastasis and drug resistance [17]. CD44+, CD133+ or LGR5+ CSCs show enhanced tumor-initiation potential; however, tumors can be reconstituted even after ablation of CD44+, CD133+ or LGR5+ CSCs owing to the plasticity of CSCs [18–20]. Therefore, it is reasonable to focus on CSC signaling networks rather than CSC markers for a mechanistic understanding of malignant processes, such as maintenance of dormancy, EMT and immune evasion.

Here, cancer stemness features related to the WNT signaling network, including proliferation/dormancy plasticity, epithelial–mesenchymal plasticity and immune-landscape plasticity, will be reviewed, and perspectives will be discussed.

Tumor cells present a spectrum of proliferation/dormancy-related features, such as fast cycling (proliferative), slow cycling (shallow quiescence), reversible growth arrest (deep quiescence) and irreversible growth arrest (senescence) (Figure 2). Cyclin A (CCNA1 and CCNA2), cyclin B (CCNB1, CCNB2 and CCNB3), cyclin D (CCND1, CCND2 and CCND3), cyclin E (CCNE1 and CCNE2), cyclin-dependent kinase 1 (CDK1), CDK2, CDK4, CDK6 and E2F are representative cell cycle activators [21], whereas cyclin-dependent kinase inhibitor 1A (CDKN1A, p21 or CIP1), CDKN1B (p27 or KIP1), CDKN1C (p57 or KIP2), CDKN2A (p16 or ARF), CDKN2B (p15), CDKN2C, CDKN2D, RB and TP53 (p53) are representative cell cycle inhibitors [22]. Definitions of slow-cycling cells, quiescent cells, persistent cells, and dormant cells are confusing and controversial [23]; however, it is clear that the proliferation/dormancy-related features of cells are regulated by the balance between cell cycle activators and inhibitors [24].

Canonical WNT signals induce proliferation of tumor cells via direct transcriptional activation of CCND1 and MYC [1] and subsequent MYC-dependent up-regulation of CCNA2, CCNB1, CCND2, CCND3, CCNE2, CDK4 and CDK7 and down-regulation of CDKN1A, CDKN1B and CDKN2B [21]. In contrast, DKK1 inhibits canonical WNT signaling because of competitive binding to LRP5/6 coreceptors and induces dormancy of lung and breast cancer cells [25], and noncanonical WNT signaling via the ROR2 receptor inhibits canonical WNT signaling owing to β-catenin degradation and induces dormancy of prostate cancer cells [26]. Noncanonical WNT signaling through the ROR1 receptor induces RhoA-mediated YAP stabilization and nuclear translocation [5], and YAP reprograms LGR5+ proliferating CSCs into LGR5- dormant CSCs [27]. In contrast, TGFβ induces dormancy through SMAD-mediated up-regulation of CDKN1A, CDKN1C and CDKN2B and p38 MAPK-mediated CDKN2A up-regulation [28]. The canonical WNT signaling cascade induces CSC proliferation, whereas noncanonical WNT, TGFβ and YAP signaling cascades induce CSC dormancy (Figure 2).

Proliferative CSCs are vulnerable to chemotherapy-induced DNA damage, whereas nonproliferative CSCs are resistant to therapeutic insults [23]. Chemotherapy elicits DNA damage and TP53-mediated growth arrest or apoptosis, while WNT/β-catenin signaling activation induces reactivation of dormant cells and promotes relapse [29]. Canonical WNT signaling activation also induces resistance to targeted therapies, such as anti-HER2 monoclonal antibody (mAb, trastuzumab) [30] and MEK1/2 inhibitor (trametinib) [31] therapies. In contrast, WNT5A and receptor tyrosine kinases, such as AXL, EGFR, FGFR1, PDGFRB and NGFR, are up-regulated in BRAF inhibitor (vemurafenib)-resistant melanoma cells in part owing to YAP- and AP-1-dependent transcriptional activation [32], suggesting the possible involvement of noncanonical WNT5A signaling in drug resistance. Indeed, noncanonical WNT signaling through ROR1 elicits resistance to an anti-HER2 antibody–drug conjugate (ADC, T-DM1) in the treatment of HER2+ breast cancer patients [33]. Canonical and noncanonical WNT signaling causes therapeutic resistance and subsequent recurrence through expansion of proliferative CSCs and survival of dormant CSCs.

WNT-targeted investigational drugs in clinical development are classified into (i) Pan-WNT signaling inhibitors, (ii) canonical WNT signaling inhibitors, (iii) noncanonical WNT signaling inhibitors and (iv) biologics targeting WNT receptors and signaling modulators (Table 1). Porcupine inhibitors that abrogate the secretion of WNT ligands are pan-WNT signaling inhibitors [16]; β-catenin protein–protein interaction (PPI) inhibitors and β-catenin proteolysis targeting chimeras (PROTACs) are canonical WNT signaling inhibitors [34,35]; ROR1 inhibitors are noncanonical WNT signaling inhibitors [36]; and anti-ROR1 mAbs, anti-ROR1 ADCs, ROR1-targted chimeric antigen receptor-modified T cells (CAR-Ts) and anti-PTK7 ADCs are cutting-edge biologics targeting noncanonical WNT receptors [5,16,37,38]. Porcupine inhibitors are expected to be ideal drugs to eliminate canonical WNT signaling-dependent proliferative CSCs and noncanonical WNT signaling-dependent dormant CSCs. However, because WNT signaling cascades are involved in the homeostasis of various normal organs or tissues, such as bone, gastrointestinal tract and neural tissues, the clinical therapeutic window of pan-WNT and canonical WNT signaling inhibitors might be relatively narrow [39]. To broaden the therapeutic window of WNT-targeted therapy through prevention of treatment-induced bone loss, an anti-RANKL (RANK ligand or TNFSF11) mAb (denosumab) has been added to phase II clinical trials of the porcupine inhibitor RXC004 for the treatment of cancer patients (ClinicalTrials.gov identifier: NCT04907539 and NCT04907851). In contrast, because ROR1 is an oncofetal protein that is preferentially expressed in fetal tissues and cancers [36], ROR1 inhibitors and ROR1-targeted biologics are expected to show preferable risk-benefit ratios in clinical trials. WNT-targeted therapeutics are valuable research tools in basic and translational studies but are not yet approved for the treatment of patients with cancer.

EMT and mesenchymal–epithelial transition are reciprocal processes accompanied by organized alteration of cell polarity, cell–cell contacts, cell–matrix contacts and the intracellular actin cytoskeleton [16,40]. Epithelial cells with apical-basal polarity interact via tight junctions, adherens junctions and desmosomes and spread over the basement membrane via hemidesmosomes, whereas mesenchymal cells with front-back polarity are not connected to each other via intercellular junctions and are not attached to the basement membrane and are thus motile (Figure 3). In contrast, a spectrum of hybrid epithelial–mesenchymal (hybrid E/M) cells, such as quasi-epithelial cells and quasi-mesenchymal cells, express both epithelial and mesenchymal markers and show intermediate phenotypes. Single-cell analyses of EMT processes have revealed that hybrid E/M cells go through sequential and parallel trajectories caused by context-dependent EMT signaling, chromatin modulation and transcriptional effects [41,42]. SNAI1 (Snail), SNAI2 (Slug), TWIST1, ZEB1 (zinc finger E-box binding homeobox 1) and ZEB2 are key transcription factors that reprogram epithelial cells toward hybrid E/M cells and mesenchymal cells under the control of WNT, FGF, Hedgehog, Notch, TGFβ and other signaling cascades (Figure 3).

Colorectal cancers are classified into CMS1 (14%, microsatellite instable and immune hot), CMS2 (37%, canonical WNT and MYC signaling activation), CMS3 (13%, KRAS mutation) and CMS4 (23%, TGFβ and EMT signaling activation) subtypes [43]. Canonical WNT signaling is activated in CMS2 due to genetic alterations in the APC or RNF43 gene [7], which directly up-regulates SNAI1 via transcriptional and post-translational mechanisms [44] and induces partial EMT [43]. TGFβ signaling is activated in CMS4 owing to the processing of latent TGFβ precursor in the TME by glycoprotein A repetition predominant protein (GARP or LRRC32), integrins, metalloproteases (MMP9 and MMP14), mild acidic conditions, reactive oxygen species and TSP1 (thrombospondin 1; also known as THBS1), which up-regulates multiple EMT regulators, such as SNAI1, SNAI2, TWIST1 and ZEB1, and induces partial or complete EMT [28]. Canonical WNT signaling up-regulates the PROX1 transcriptional repressor in colorectal cancer [45] and down-regulates MMP14 to inhibit TGFβ-dependent EMT and extracellular matrix (ECM) remodeling [46], whereas TGFβ1 signaling down-regulates the WNT signaling potentiator Lgr5 to repress proliferation but induce EMT and CMS2-to-CMS4 conversion in colorectal cancer organoids [47]. Counteracting the canonical WNT and TGFβ signaling cascades shapes epithelial CMS2 and mesenchymal CMS4 phenotypes, respectively.

Breast CSCs are classified into canonical WNT-dependent epithelial-like CSCs and YAP-dependent mesenchymal-like CSCs [48], and gastric cancers are classified into an epithelial subtype with canonical WNT signaling activation and a mesenchymal subtype with Hedgehog signaling activation [49]. Noncanonical WNT and Hedgehog-Smoothened signaling promotes YAP stabilization and YAP-dependent transcription [5,50], and YAP up-regulates WNT5A and ITGAV (integrin αv) to establish a noncanonical WNT-YAP signaling loop [51] and activate TGFβ signaling [14], respectively. In contrast, noncanonical WNT, TGFβ and YAP signaling converges to induce ZEB1-dependent transcription [52–54] as well as AP-1-dependent transcription [1,28,53]. ZEB1 directly down-regulates ZEB1-only target genes, such as CDH1, CLDN3, CLDN4 and LLGL2, to promote EMT, while ZEB1 interacts with AP-1 and YAP and up-regulates target genes of ZEB1/AP-1/YAP (CTGF, EDN1, NRP1, TGFBI and others) to regulate para-EMT signaling [53]. ZEB1 and YAP bind to the AXIN2 promoter to inhibit canonical WNT signaling and maintain a quiescent or dormant state [55]. Noncanonical WNT, TGFβ and YAP signaling cascades constitute a mesenchymal signaling network that counteracts the canonical WNT signaling cascade. The epithelial CSC signaling network driven by the canonical WNT signaling cascade and the mesenchymal CSC signaling network driven by the noncanonical WNT, TGFβ and YAP signaling cascades compete with each other, and their interactions fine-tune the epithelial–mesenchymal plasticity (Figure 3).

Canonical WNT signaling is involved in proliferation and epithelial (or hybrid E/M) transition, whereas TGFβ and noncanonical WNT signaling are involved in dormancy and mesenchymal (or hybrid E/M) transition, as discussed above. The shared characteristics of epithelial/proliferative and mesenchymal/dormant phenotypes suggest mechanistic overlap in epithelial–mesenchymal and proliferative/dormant plasticity. In contrast, mammary tumor cells with dominant negative β-catenin mutation retained TGFβ-induced stemness but lost EMT, chemoresistance and metastasis potential [56], which indicates that the trajectories of epithelial/proliferative and mesenchymal/dormant reprogramming are not completely consistent with each other. Malignant phenotypes individually develop during sequential and parallel EMT trajectories depending on (i) context-dependent activation of WNT, TGFβ and other signaling cascades, (ii) combinatory effects of β-catenin-TCF/LEF, SMADs, NFAT, AP-1, YAP-TEAD and EMT-related transcription factors and (iii) the state of chromatin of target loci. Therefore, large-scale and multiomics single-cell analyses of primary or metastatic tumors are needed to understand how the canonical and noncanonical WNT signaling networks control cancer stemness, EMT, dormancy, drug resistance and relapse.

Tumor cells, CAFs, endothelial cells and immune cells are organized into inflammatory or noninflammatory TMEs [12,57–61]. Immune-stimulating CAFs, leukocyte-attracting endothelial cells and proinflammatory immune cells generate an inflammatory TME with elevated CCL2 (C-C motif chemokine ligand 2), CCL5, IFNγ (interferon-γ; also known as IFNG), IL1β (interleukin-1β) and TNF (tumor necrosis factor) levels (Figure 4A). In contrast, ECM-remodeling CAFs, angiogenic endothelial cells and suppressive immune cells give rise to a noninflammatory TME with elevated TGFβ, CCL28, IL4, IL10, lactate and VEGF (vascular endothelial growth factor) levels (Figure 4B). An inflammatory TME with active immune cells (featuring a hot immune environment with strong immune surveillance) preserves antitumor immunity to suppress tumorigenesis, whereas an inflammatory TME with exhausted immune cells (featuring a hot immune environment with strong immune evasion) and a noninflammatory TME (featuring a cold immune environment with strong immune evasion) lack antitumor immunity and thus promote tumorigenesis.

PD-1 (programmed cell death 1, also known as PDCD1 or CD279) receptor and its ligand PD-L1 (PD-1 ligand 1; also known as CD274) constitute an immune checkpoint. The immunosuppressive PD-L1 ligand is up-regulated in tumor cells, CAFs, endothelial cells and monocytic myeloid-derived suppressor cells (MDSCs) [57–59,62] owing to IFNγ signaling, WNT/β-catenin-MYC signaling and hypoxia [62–65], while the PD-1 receptor is expressed on macrophages, natural killer cells, CD8+ (cluster of differentiation 8 positive) exhausted T cells and CD4+ follicular regulatory T (Treg) cells [66–68]. Anti-PD-L1 mAbs (atezolizumab, avelumab and durvalumab) and anti-PD-1 mAbs (cemiplimab, nivolumab and pembrolizumab) have been developed as immune checkpoint inhibitors (ICIs) to relieve immune exhaustion caused by PD-L1/PD-1 signaling [69,70]. Because PD-1 is expressed on both CD8+ cytotoxic T cells and CD4+ Treg cells [68], ICIs reinvigorate cytotoxic T cells to not only suppress tumorigenesis in some patients but also expand Treg cells to promote tumorigenesis in other patients.

Canonical WNT signaling affecting the β-catenin-TCF7 (Tcf-1) complex in CD8+ T and Treg cells enhances inflammation in part through up-regulation of proinflammatory cytokines, such as IFNγ and TNF [71,72]. In addition, WNT ligands, such as WNT3A, WNT5A, WNT10A and WNT10B, are up-regulated in inflamed tissues through proinflammatory signaling affecting the NF-κB (nuclear factor enhancer of immunoglobulin κ light chain of activated B cells) and STAT3 (signal transducer and activator of transcription 3) transcription factors [73–75]. Because WNT5A enhances NK-κB signaling to up-regulate IL1β and IL6 and these cytokines activate STAT3 signaling to up-regulate WNT5A, WNT signaling is incorporated into a proinflammatory feedforward loop and promotes immune surveillance. In contrast, canonical WNT signaling activation in dendritic cells (DCs) elicits immune evasion owing to IDO1 (indoleamine 2,3-dioxygenase 1)- and IL10-mediated Treg-cell expansion and cytotoxic T-cell impairment, respectively [76,77]. Canonical WNT signaling activation in tumor cells also promotes immune evasion through direct up-regulation of PD-L1 [64], the Treg-cell-attracting chemokine CCL28 [78] and the suppressive macrophage-supporting protein WISP1 (WNT-induced secreted protein 1) [79] and indirect down-regulation of the DC-attracting chemokines CCL4 [80] and CCL5 [81]. Taken together, these findings clearly indicate that WNT signaling activation induces immune surveillance as well as immune tolerance in a context-dependent manner (Figure 4C).

The response to pembrolizumab monotherapy in 1188 patients with cancer was positively correlated with a T-cell-inflamed signature (P = 3.6 × 10⁻¹²) and negatively correlated with angiogenesis, monocytic MDSCs and stroma/EMT/TGFβ signatures (P = 0.0001, 0.0001 and 0.0003, respectively) [82]. Because tumors with a hot environment with immune evasion are vulnerable to ICI-induced reinvigoration of CD8+ cytotoxic T cells, it is reasonable to select cancer patients for ICI monotherapy using the T-cell-inflamed signature [69]. TGFβ promotes generation of ECM-remodeling CAFs, M2-TAMs, monocytic MDSCs and Treg cells [61]; monocytic MDSCs produce IL10, TGFβ and VEGF [59]; and VEGF-dependent tumor angiogenesis induces a hypoxic and acidic TME, PD-L1 up-regulation and Treg-cell/MDSC expansion [58]. Because TGFβ signaling, monocytic MDSCs and tumor angiogenesis are immunosuppressive and together generate an ICI-resistant, cold environment with immune evasion (Figure 4), combination ICI therapies using angiogenesis inhibitors, monocytic MDSC inhibitors or TGFβ signaling inhibitors are promising options for cancer patients with a noninflamed environment featuring immune evasion.

The WNT expression signature was not significantly correlated with ICI resistance in the study mentioned above (P=0.10) [82], although an association between WNT-related genetic alterations and decreased T-cell inflammation was noted by others [83]. WNT-related genetic alterations occur in the CMS2 subtype of colorectal cancer, which shows canonical WNT and MYC signaling activation, as well as in the CSM1 subtype, which shows microsatellite instability (generally considered immune hot), and the CMS4 subtype, which shows stroma/EMT/TGFβ signaling activation (immune cold) [7,43]. It is not appropriate to predict ICI response based on WNT-related genetic alterations because such alterations are present in cases with the T-cell inflamed signature and stroma/EMT/TGFβ signature. Although the context-dependent immune regulation by WNT signaling cascades [71–81] and the unknown risk–benefit ratio of WNT signaling inhibitors [39] are limiting factors, phase I or II clinical trials of combination ICI therapy with WNT signaling inhibitors or WNT-related biologics (Table 1) are ongoing and aim to reveal new avenues for immunotherapy in cancer.

• Canonical WNT signaling affecting the β-catenin-TCF/LEF cascade promotes CCND1/MYC-dependent proliferation and SNAI1-dependent partial EMT and context-dependently modulates immune surveillance and immune evasion.

• Transcriptional outputs resulting from noncanonical WNT signaling depend on YAP, AP-1 and NFAT crosstalk with TGFβ signaling cascades to promote CDKN-dependent dormancy, ZEB1-dependent EMT and cold immune evasion.

• The balance between the epithelial signaling network driven by the canonical WNT signaling cascade and the mesenchymal signaling network driven by noncanonical WNT, TGFβ and YAP signaling cascades modulates cancer stemness features.

• WNT-related proliferation/dormancy plasticity, epithelial-mesenchymal plasticity and immune-landscape plasticity are promising targets for therapeutic intervention.

• Further understanding of context-dependent reprogramming trajectories and plasticity-related chromatin remodeling are needed to optimize the clinical benefits of WNT-targeted therapy.

The authors declare that there are no competing interests associated with this manuscript.

This work was supported in part by grants-in-aid from Masaru Katoh’s Fund for the Knowledge-Base Project and the Global Network Project.

Masuko Katoh and Masaru Katoh designed and wrote this manuscript.

ADC

antibody–drug conjugate

AP-1

activator protein 1

APC

adenomatous polyposis coli

CAF

cancer-associated fibroblast

CCL

C-C motif chemokine ligand

CCN

cyclin

CD8+

cluster of differentiation 8 positive

CDK

cyclin-dependent kinase

CDKN

cyclin-dependent kinase inhibitor

CSC

cancer stem cell

CTNNB1

β-catenin

DC

dendritic cell

ECM

extracellular matrix

EMT

epithelial–mesenchymal transition

FGF

fibroblast growth factor

Frizzled

atypical G-protein-coupled receptor with Frizzled-like domain

GARP

glycoprotein A repetition predominant protein or LRRC32

GPCR

G-protein-coupled receptor

ICI

immune checkpoint inhibitor

IDO1

indoleamine 2,3-dioxygenase 1

IFNγ

interferon-γ

IL1β

interleukin-1β

ITGAV

integrin αv

LGR5

leucine rich repeat-containing G protein-coupled receptor 5

LRP5/6

LDL receptor related proteins 5 and 6

mAb

monoclonal antibody

MDSC

myeloid-derived suppressor cell

NF-κB

nuclear factor enhancer of immunoglobulin κ light chain of activated B cells

NFAT

nuclear factor associated with T cells

NK

natural killer

PD-1

PDCD1 or CD279

PD-L1

PD-1 ligand 1 or CD274

PLC

phospholipase C

PPI

protein–protein interaction

PROTAC

proteolysis targeting chimera

RNF43

ring finger protein 43

ROR1/2

receptor tyrosine kinase-like orphan receptors 1 and 2

SNAI1

Snail

STAT3

signal transducer and activator of transcription 3

TAM

tumor-associated macrophage

TCF/LEF

T-cell factor 7 (TCF7 or Tcf-1) TCF7-like 1 (TCF7L1 or Tcf-3) TCF7-like 2 (TCF7L2 or Tcf-4) and lymphoid enhancer factor 1 (LEF1 or Lef-1)

TGFβ

transforming growth factor-β

TME

tumor microenvironment

TNF

tumor necrosis factor

Treg

regulatory T

VEGF

vascular endothelial growth factor or VEGFA

WISP1

WNT-induced secreted protein 1

YAP-TEAD

Yes-associated transcriptional regulator and TEA domain transcription factor

ZEB1

zinc finger E-box binding homeobox 1

ZNRF3

zinc and ring finger 3