You are on page 1of 17

ReVIeWS

Mutations and mechanisms of WNT


pathway tumour suppressors in cancer
Jeroen M. Bugter , Nicola Fenderico and Madelon M. Maurice ✉

Abstract | Mutation-​induced activation of WNT–β-​catenin signalling is a frequent driver event in


human cancer. Sustained WNT–β-​catenin pathway activation endows cancer cells with sustained
self-​renewing growth properties and is associated with therapy resistance. In healthy adult stem
cells, WNT pathway activity is carefully controlled by core pathway tumour suppressors as well
as negative feedback regulators. Gene inactivation experiments in mouse models unequivocally
demonstrated the relevance of WNT tumour suppressor loss-​of-​function mutations for cancer
growth. However, in human cancer, a far more complex picture has emerged in which missense or
truncating mutations mediate stable expression of mutant proteins, with distinct functional and
phenotypic ramifications. Herein, we review recent advances and challenges in our understanding
of how different mutational subsets of WNT tumour suppressor genes link to distinct cancer types,
clinical outcomes and treatment strategies.

Stem cell
Nearly 40 years ago, the first WNT gene was discovered1. from driver mutations and predict whether and how these
An undifferentiated cell that This discovery unveiled WNT1 as the gene responsible mutations generate vulnerabilities that can be exploited
divides indefinitely to renew for driving mammary tumorigenesis in mice upon inser- by cancer treatment18. Furthermore, mutations that hit
itself and produce progenitors tion of the mouse mammary tumour virus (MMTV)1. a single pathway may display noticeable tissue or cancer
that differentiate into
In the following two decades, most of the key components subtype specificity, or show cooperative effects with
specialized cell types to
replenish lost cells in the of the WNT pathway were unravelled. Concurrently, mutations that deregulate alternative signalling pathways
tissue during homeostasis these discoveries signified prominent roles for WNT sig- in the same tumour, further enhancing complexity and
and repair. nalling in the control of embryogenesis as well as cancer impeding a mechanistic understanding of how genetic
development2. Later work uncovered WNT signalling as alterations contribute to cancer growth.
Missense mutations
Single base pair substitutions
a primary driving force for adult stem cell self-​renewal This Review focuses on recent advances in our under-
in the genome that result in a and cell-​fate specification in the intestine, liver, skin and standing of how various classes of mutations in WNT
change in the codon, leading to many other organs 3–7. During tissue homeostasis, pathway tumour suppressors contribute to cancer deve­
the encoded protein bearing a WNT pathway activity is tightly controlled by various lopment. An overview is provided of how different WNT
single amino acid substitution.
tumour suppressors that function either as core compo­ pathway mutations link to tissue specificity and clini-
Nonsense mutations nents or act in negative feedback loops8–14. Mutation or cal progression of disease. We discuss consequences of
Single base pair substitutions silencing of WNT tumour suppressor genes has been loss-​of-​function (LOF) mutations in preclinical models and
in the genome that result in an identified as a frequent occurrence in a wide variety of human patients and describe how the investigation of
early stop codon, leading to a
cancer types15. In line with these observations, persistent other classes of patient-​derived mutations have unveiled
prematurely truncated protein
variant or nonsense-​mediated
WNT–β-​catenin pathway activation was found to endow novel mechanisms by which WNT pathway tumour sup-
mRNA decay. cancer cells with self-​renewing growth properties and, pressor mutations drive cancer growth and progression.
furthermore, is linked to therapy resistance16,17. Finally, we highlight the need for a more comprehensive
Large-​scale sequencing efforts have made it clear understanding of the common principles by which WNT
that genetic alterations in cancer are highly diverse pathway mutagenic events drive carcinogenesis to enable
and complex. Whereas mutations in oncogenes often improved applications of precision cancer medicine and
Oncode Institute and display mutational hot spots that bring about protein the design of more effective therapies.
Department of Cell Biology, overactivity, the landscape of mutations in tumour
Center for Molecular Medicine, suppressor genes usually comprises a mixture of dele- WNT–β-​catenin signalling
University Medical Center
Utrecht, Utrecht, Netherlands.
tions and missense mutations, nonsense mutations and WNT-​mediated receptor activation can direct signalling
✉e-​mail: m.m.maurice@ frameshift mutations that are found scattered over the along distinct pathways that are roughly divided into
umcutrecht.nl whole length of the gene, complicating functional inter- β-​catenin-​dependent and β-​catenin-​independent signal-
https://doi.org/10.1038/ pretation. Thus, these observations have posed a chal- ling routes. Herein, we focus on the β-​catenin-​dependent
s41568-020-00307-​z lenge to unambiguously distinguish passenger mutations pathway that is prominently linked to the development

naTure RevIeWs | CANceR volume 21 | January 2021 | 5


Reviews

a Off On
LRP5
FZD WNT
or LRP6

DVL DVL
CK1 GSK3β CK1 APC AXIN1
GSK3β
GSK3β β-TrCP
AXIN1 AXIN1 CK1
β-Catenin β-Catenin Ub
APC APC
P P P P P P Ub β-Catenin
β-TrCP β-Catenin
Ub
Destruction
complex

β-Catenin • AXIN2
• RNF43
TCF TCF • ZNRF3
Nucleus • LGR5

b Negative feedback WNT potentiation


RSPO
RNF43 WNT RSPO WNT
or ZNRF3

1 Ub DVL
Ub LGR APC AXIN1
AXIN2 Ub
2 Endosome GSK3β
CK1
CK1 GSK3β
AXIN1/2 β-Catenin
β-Catenin
APC
P P P β-Catenin
β-Catenin
TCF β-Catenin TCF
Lysosomal Lysosomal
degradation degradation
Frameshift mutations
Mutations by which one or
c WNT-independent WNT-hypersensitive
multiple base pairs are inserted
in or deleted from the genome WNT
resulting in a change of the
reading frame of the gene,
leading to a truncated protein
variant with an incorrect amino DVL
acid sequence downstream
CK1 GSK3β APC AXIN1
of the frameshift or nonsense- DVL β-TrCP GSK3β
mediated mRNA decay. AXIN1/2 CK1
β-Catenin
Passenger mutations APC
P P P β-Catenin
Mutations that do not endow
β-TrCP
a competitive advantage but β-Catenin
coincide with a driver mutation. β-Catenin β-Catenin β-Catenin
Typical cancer genomes β-Catenin β-Catenin
comprise 102–106 passenger β-Catenin β-Catenin
mutations.
TCF TCF
β-Catenin β-Catenin
Driver mutations
Mutations that endow tumour
cells with a growth or survival
advantage over cells without of cancer16,19. A central event in this signalling cascade is it is ubiquitylated by the E3 ligase SKP1-​CUL1-​F-​box
this mutation. On average,
the regulated proteolytic turnover of the transcriptional protein (SCF)–β-​transducin repeat-​containing protein
cancer genomes comprise
four or five driver mutations. co-activator β-​catenin. In resting cells, the pool of cyto- (β-​TrCP) and dispatched to the proteasome to complete
solic β-​catenin is kept low by the activity of a multisub- its degradation10,20–27 (Fig. 1a).
Loss-​of-​function (LOF) unit destruction complex, composed of the tumour WNT-​mediated stimulation of cells is initiated by
mutations suppressors AXIN1 and adenomatosis poly­posis coli the formation of a trimeric complex between WNT and
Genomic mutations resulting
in an (partially) inactive gene
(APC) and the kinases casein kinase 1 (CK1) and glyco­ its receptors, Frizzled (FZD) and low-​density lipopro-
product. They often involve gen synthase kinase 3β (GSK3β). This protein complex tein receptor-related protein 5 (LRP5) or LRP6, at the
loss of heterozygosity. captures and phosphorylates β-​catenin, after which cell surface. Receptor activation mediates recruitment

6 | January 2021 | volume 21 www.nature.com/nrc


Reviews

◀ Fig. 1 | WNT signalling pathway regulation and mutation routes to WNT-independent receptor (LGR) family and RNF43 or ZNRF3, RSPO
or WNT-hypersensitive growth states in cancer. a | Left: in the absence of WNT promotes the clearance of RNF43 and ZNRF3 from
ligands, cytosolic β-​catenin is continuously degraded by a destruction complex the cell surface13,40,41 (Fig. 1b). Additionally, a subset of
comprising the oligomerizing scaffold proteins adenomatous polyposis coli (APC) RSPO family members promotes clearance of RNF43
and AXIN1 and the associated kinases glycogen synthase kinase 3β (GSK3β) and casein
and ZNRF3 via an LGR-​independent mechanism that
kinase 1 (CK1). The kinases phosphorylate β-​catenin, marking it for ubiquitylation by
the E3 ligase SKP1-​CUL1-​F-​box protein (SCF)–β-​transducin repeat-​containing protein
involves binding to heparan sulfate proteoglycans
(β-​TrCP) and subsequent proteasomal degradation. Right: WNT proteins bind and (HSPGs)42–45. Thus, RSPO activity allows WNT recep-
activate their cognate receptor Frizzled (FZD) and co-​receptor low-​density lipoprotein tors to accumulate at the cell surface and induce high
receptor-related protein 5 (LRP5) or LRP6. Next, the cytosolic protein Dishevelled (DVL) levels of β-​catenin-​mediated transcription required for
is recruited to the plasma membrane by the activated receptors, where it promotes stem cell maintenance.
multimerization of the receptor complex and recruits components of the destruction
complex, ultimately leading to inhibition of β-​catenin proteolysis. As a consequence, WNT tumour suppressor mutations
cytosolic β-​catenin accumulates and translocates to the nucleus where it promotes Driver routes for β-​catenin activation
the transcription of WNT target genes in association with T cell factor/lymphoid The primary output of driver mutations in WNT pathway
enhancer-​binding factor (TCF/LEF) transcription factors. b | Left: WNT target gene
tumour suppressors is unwarranted β-​catenin-​mediated
negative feedback regulators ring finger protein 43 (RNF43) and zinc and ring finger
protein 3 (ZNRF3) ubiquitylate FZD proteins and mark these receptors for endocytosis transcription, which activates a gene programme that
and lysosomal degradation (step 1). In addition, RNF43 facilitates the restoration of promotes stemness, proliferation and survival 16,19.
destruction complex activity (step 2). WNT target gene negative feedback regulator Excessive β-​catenin activity can be achieved by two
AXIN2 helps the formation of new destruction complexes. Whether and how restoration major mutational driver routes. Mutational inactivation
of destruction complex activity involves relocation from the receptor complex to the of the β-​catenin destruction complex is the archetypal
cytosol remains unknown (dashed arrow). These events lead to attenuation of WNT mode of WNT pathway activation in cancer, exemplified
signalling. Right: in stem cells, R-​spondins (RSPOs) bind to their cognate receptors by inactivating mutations in APC, AXIN1 and AXIN2 or
leucine-​rich repeat-​containing G protein-​coupled receptor 4 (LGR4), LGR5 or LGR6 activating mutations in β-​catenin (encoded by CTNNB1)
and the extracellular domain of RNF43 or ZNRF3. This complex facilitates RNF43 and itself  8,46–48. These events are generally assumed to drive
ZNRF3 clearance from the plasma membrane, thereby potentiating WNT signalling.
a state of WNT-​independent tumour growth (Fig. 1c).
c | Molecular mechanisms of aberrant WNT signalling pathway activation mediated
by cancer mutations. Tumour suppressor mutations (left) abrogate destruction complex More recently, misregulation of WNT receptor abun-
activity to drive WNT-​independent growth or (right) abrogate negative feedback at the dance emerged as a second route to carcinogenesis.
receptor level to drive WNT-​hypersensitive growth. P, phosphorylation; Ub, ubiquitin. This oncogenic route is initiated by loss of RNF43 and
ZNRF3, thus eradicating the negative feedback loop that
normally drives WNT receptor endocytosis in adult
of the effector protein Dishevelled (DVL), which in turn stem cells14. Of note, RNF43- and ZNRF3-​depleted cells
recruits AXIN1 and the other core destruction complex have lost their requirement for RSPO and thus display
components to the plasma membrane28–35. Together, aspects of niche-​independent growth, a hallmark of
these events abrogate destruction complex-​mediated cancer cells14,49. An alternative pathway by which cancer
degradation of β-​catenin, allowing it to accumulate cells may avoid the activities of RNF43 and ZNRF3 is
and translocate to the nucleus to bind the T cell factor/ by acquired RSPO overexpression50,51. As a consequence,
lymphoid enhancer-​binding factor (TCF/LEF) family of these cancers display an overabundance of cell surface
transcription factors and drive the transcription of WNT WNT receptors, which mediates ligand hypersensi­
target genes36,37 (Fig. 1a). tivity and promotes WNT-​dependent tumour growth13,14
To prevent excessive signalling, WNT-induced (Fig. 1c).
res­ponses are balanced by numerous negative feedback
regulators. The WNT target gene AXIN2 (also known as Tissue specificity
conductin), the RNA product of which is commonly used Although genetic alterations in the WNT pathway occur
as a marker for WNT pathway activity, helps to increase in most cancer types, mutations in individual WNT sig-
cytosolic destruction complex activity and downregulate nalling components show a substantial level of tissue
β-​catenin-​mediated transcription11,12,38. Furthermore, specificity. Examination of the frequencies of mutations
induced expression of the membrane-​bound E3 ligases in individual WNT pathway components across 33 spo-
ring finger protein 43 (RNF43) and zinc and ring finger radic cancer types listed in The Cancer Genome Atlas
protein 3 (ZNRF3) mediates ubiquitylation of WNT (TCGA) reveals that endometrial, skin and stomach
receptors, which drives their internalization and lyso- cancer collect mutations in various core components,
somal degradation, thereby attenuating the sensitivity whereas other cancer types are more selective (Fig. 2a).
of cells to incoming WNTs13,14. Additionally, RNF43 In colorectal cancer (CRC; n = 594), the largest fraction
was found recently to perform a secondary suppressor of sporadic tumours accumulates APC mutations (67%),
role, preventing β-​catenin-​mediated transcription by followed by lower fractions of RNF43 (8%), CTNNB1
promoting destruction complex activity, through an (6%) and AXIN2 (5%) mutations. By contrast, liver
incompletely understood mechanism that involves cancer cases (n = 372) preferentially acquire CTNNB1
regu­latory interactions of the RNF43 cytosolic tail with (25%) and AXIN1 (8%) mutations, whereas pancreatic
CK1 (ref.39) (Fig. 1b). However, within the adult stem cell cancers (n = 184) favour mutations in RNF43 (6%) and
niche, negative feedback by RNF43 and ZNRF3 is coun- adrenocortical cancer (n = 92) links to mutations in
teracted locally by secreted proteins of the R-​spondin ZNRF3 (20%) or CTNNB1 (15%). Of note, the contri-
(RSPO) family13. By forming a complex with members bution of WNT pathway alterations to cancer develop-
of the leucine-​rich repeat-​containing G protein-​coupled ment might be larger than reflected in TCGA database.

naTure RevIeWs | CANceR volume 21 | January 2021 | 7


Reviews

a b APC RNF43 ZNRF3

APC Colorectal adenocarcinoma 66.55%


Uterine corpus endometrial carcinoma 14.34% CMS1
Skin cutaneous melanoma 14.25%
CMS2
Stomach adenocarcinoma 13.38%
Bladder urothelial carcinoma 7.04% CMS3
RNF43 Uterine corpus endometrial carcinoma 15.28% CMS4
Stomach adenocarcinoma 10.20%
Colorectal adenocarcinoma 7.90% c APC RNF43 ZNRF3
Pancreatic ductal adenocarcinoma 5.95%
ZNRF3 Adenocortical carcinoma 20.43%
Uterine corpus endometrial carcinoma 6.23%
Skin cutaneous melanoma 5.35%
AXIN1 Liver hepatocellular carcinoma 7.77%
Uterine corpus endometrial carcinoma 6.42%
Deep deletion
AXIN2 Uterine corpus endometrial carcinoma 7.36% AXIN1 AXIN2 CTNNB1
Colorectal adenocarcinoma 5.04% In-frame
mutation
CTNNB1 Liver hepatocellular carcinoma 25.47%
Missense
Uterine corpus endometrial carcinoma 25.47%
mutation
Adenocortical carcinoma 15.05%
Skin cutaneous melanoma 6.68% Truncating
Stomach adenocarcinoma 6.68% mutation
Colorectal adenocarcinoma 5.71% Gene fusion

Fig. 2 | Frequencies and types of WNT pathway mutation in human cancer types. Data source: Memorial Sloan
Kettering Cancer Center (MSKCC) cBioPortal196. a | Frequency of mutation or deep deletion of the indicated WNT pathway
genes per The Cancer Genome Atlas (TCGA) study. Frequencies >5% are shown. b | Consensus molecular subtypes (CMSs)
of TCGA colorectal cancer samples that have a mutation in adenomatous polyposis coli (APC), ring finger 43 (RNF43) or zinc
and ring finger 3 (ZNRF3). Samples without a CMS label were left out. CMS classification was derived from the Colorectal
Cancer Subtyping Consortium (CRCSC)66. c | Distribution of deep deletions, in-​frame mutations, missense mutations,
truncating mutations or gene fusions found in all TCGA Pan-​Cancer studies for APC, RNF43, ZNRF3, AXIN1, AXIN2 and
CTNNB1 (encoding β-​catenin).

For instance, the number of identified RNF43 muta- mutations in sporadic CRCs58,59 (Fig. 2a). Molecular
tions appears under-​represented owing to incomplete characterization of tumour lesions at various stages of
Promoter hypermethylation calling of frameshift mutations that display similarity development indicated that APC mutations are an early
A common mechanism of to technical DNA slippage errors52. Furthermore, WNT and perhaps rate-​limiting event60. However, recent work
gene silencing, mediated by
path­way regulators may be downregulated by promoter suggests that pre-​existing oncogenic KRAS mutations,
methylation of CpG islands
in the promoter region. This hypermethylation, as shown for AXIN1 in a large frac- commonly found in the ageing human colon, might
mechanism is frequently tion of non-​small-​cell lung cancer (NSCLC) cases (43%, mark preferred initiation sites for APC mutations61.
misused by cancer cells to 29/67)53, as well as for ZNRF3 in BRAF-​mutant CRC Clearly, driver mutations in KRAS, TP53, SMAD4 and/or
turn off the expression of (72%, 36/50)54. PIK3CA cooperate with APC mutations to drive the step-
(tumour suppressor) genes.
wise progression from adenoma to carcinoma60,62,63.
Familial adenomatous Tumour subtypes and prognosis Besides alterations in WNT, epidermal growth factor
polyposis Although WNT tumour suppressor mutations all (EGF), transforming growth factor-​β (TGFβ) or PI3K
(FAP). An autosomal dominant converge on enhancing β-​catenin activity, emerging signalling pathways that each provide a selective growth
inherited condition that
evidence indicates that the functional consequences advantage, these CRCs are characterized by chromosomal
involves extensive polyp
formation in the colon and of different WNT pathway mutations are not equal. instability (CIN)60. Although this conventional sequence of
a high risk of colorectal Large-​s cale cancer genome profiling revealed that mutational events explains the genetic basis of a size­able
cancer, caused by a specific mutations in WNT pathway components may fraction of CRCs, a much larger collection of mutated
heterozygous germline occur either early or late during cancer development, genes were revealed with the advances of cancer genome
mutation in adenomatous
polyposis coli (APC). A
associate with selective cancer subtypes and differen- sequencing, thereby disclosing the heterogeneous nature
second-​hit mutation or loss tially correlate with clinical prognosis. Below, we discuss of CRC64,65.
of heterozygosity precedes recent insights obtained for two major cancer types. Through a large-​scale, international effort, patients
malignant polyposis. with CRC (n = 3,962) were subclassified recently into
Colorectal cancer. The prominent role of APC as a four consensus molecular subtypes (CMSs), each with
Chromosomal instability
(CIN). Genomic instability gatekeeper gene was first revealed in 1991 by multiple distinguishing biological characteristics: CMS1 (14%),
that involves frequent research groups who collectively showed that inactivat- hypermutated subgroup with microsatellite instability
mis-​segregation of ing APC mutations cause familial adenomatous polyposis (MSI) and strong immune activation; CMS2 (37%),
chromosomes during (FAP), a syndrome characterized by extensive intesti- canonical epithelial subgroup with marked WNT and
cell division, resulting
in gain or loss of entire
nal polyp formation and a high risk of CRC develop- MYC activation; CMS3 (13%), epithelial subgroup with
chromosomes or sections ment46,55–57. These findings were substantiated by the evident metabolic dysregulation; and CMS4 (23%), dis-
of them. simultaneous identification of a large number of APC playing prominent TGFβ activation, stromal invasion

8 | January 2021 | volume 21 www.nature.com/nrc


Reviews

Microsatellite instability
and angiogenesis66. Strikingly, WNT pathway mutations usually higher in left-​sided CRC tumours76,77. Indeed,
(MSI). A state of hypermutation in CRC segregate to different CMSs (Fig. 2b). Although physiological WNT signalling levels also display regional
due to impaired mismatch APC mutations are prevalent in CMS2 and coincide with differences along the healthy intestinal tract77, which
repair of the genome. Most CIN, RNF43 and ZNRF3 mutations as well as RSPO might be related to the developmental history of the
common in colorectal, gastric
and endometrial cancer.
family member gene fusions are found strongly enriched adult colon, with left and right regions originating from
within MSI-​linked hypermutated tumours of the CMS1 the embryonic hindgut and midgut, respectively78,79.
Mismatch repair subgroup52,67,68. Thus, a relationship exists between the Importantly, the primary tumour site differentially cor-
(MMR). A DNA repair pathway type of WNT pathway mutation and the biological responds with prognosis. Left-​sided tumours mostly
that repairs DNA replication
context of the tumour. metastasize to the liver and lung, for which resection and
errors, such as base–base
mismatches or erroneous
In particular, RNF43 mutations and, to a lesser extent, chemotherapy is a potential cure, whereas right-​sided
insertions or deletions. ZNRF3 mutations associate with the serrated pathway of tumours give rise to peritoneal metastases that remain
CRC development that originates from pre-​malignant difficult to treat and lead to poor survival65,80,81.
Sessile serrated adenoma serrated polyps enriched for activating BRAF muta- In summary, driver mutations in APC and RNF43
(SSA). Pre-​malignant
flat (or sessile) polyps
tions and mismatch repair (MMR) gene mutations are representative of CRC subsets with distinct charac-
predominantly observed (mainly mutL homologue 1 (MLH1))67,69,70. Contrary teristics. An overview of the biological and molecular
on the right side of the colon. to APC mutations, the acquisition of RNF43 mutations characteristics of APC-​mutant and RNF43-​mutant CRC
These polyps comprise the is considered a late event that drives the progression of subsets is provided in Table 1.
main precursor lesion of
sessile serrated adenoma (SSA) to carcinoma67,71. Notably,
serrated carcinogenesis.
unlike APC-​mutant tumours, BRAF-​mutant, MSI-​linked Hepatocellular carcinoma. Hepatocellular carcinoma
Left-​sided CRC tumours do not display prominent nuclear β-​catenin (HCC) is the sixth most frequent cancer worldwide
(Left-​sided colorectal cancer). accumulation72. Thus, these findings suggest that RNF43 and the third leading cause of cancer-related death82.
Tumours originating distal to mutations generally associate with lower levels of WNT HCC accounts for around 90% of patients with liver
the splenic flexure (descending
colon, sigmoid colon and
pathway activation when compared with APC muta- cancer and is characterized by frequent overactiva-
rectum). Typically associated tions. Recent work showed that RNF43 mutations are tion of WNT–β-​catenin signalling. The most promi-
with higher levels of WNT also enriched in signet-​ring cell carcinoma, a rare type nent genetic alterations in the WNT pathway in HCC
signalling. of early-​onset CRC that displays aggressive behaviour comprise activating mutations in CTNNB1 (28–40%,
and associates with poor prognosis73. Distinct from 34/123 and 18/45) and LOF mutations in AXIN1 (11%,
Right-​sided CRC
(Right-​sided colorectal cancer).
APC-​mutant tumours, these signet-​ring cell tumours 13/123 and 5/45)83,84. In line with their tumour sup-
Tumours originating proximal do not display prominent activation of EGF or PI3K pressor role, AXIN1 mutations frequently asso­ciate
to the splenic flexure (caecum, pathways and only show low expression levels of conven- with loss of heterozygosity (LOH)48. As a large fraction
ascending colon and transverse tional WNT target genes, further emphasizing a distinct of human hepatocytes are polyploid, which may pro-
colon). Typically associated
with lower levels of WNT
pathogenic mechanism73. tect these cells from tumour suppressor loss85, it is
signalling. Strikingly, APC and RNF43 mutations also associate tempting to speculate that AXIN1 mutations are likely
with distinct primary tumour locations within the acquired in a diploid state. Contrary to APC mutations
Loss of heterozygosity colon. Whereas APC mutations are predominantly in CRC, CTNNB1 and AXIN1 mutations are usually
(LOH). A frequent event in
found in left-​sided CRC tumours, RNF43 mutations are not present in pre-​malignant lesions of the liver but
cancer, involving loss of one
allele of a gene in the tumour
enriched in right-​sided CRC tumours65,74,75. In line with instead associate with later stages of tumour develop-
genome, after which a single the RNF43-​mutant profiles described above, right-​sided ment48,86,87. Concordantly, WNT pathway activation
allele of the (mutated) gene tumours commonly harbour BRAF mutations and becomes apparent only in the late stages of HCC and is
remains. display MSI-​linked hypermutation. Yet, even when thus considered a late step in the progression towards
Polyploid
correc­ted for MSI status, RNF43 mutations correlate hepatocarcinogenesis88.
A condition in which a cell with right-​sided CRC tumours, further supporting a Strikingly, CTNNB1 and AXIN1 mutations asso­
contains more than two regional preference for WNT pathway alterations74. The ciate with distinct HCC subtypes that display divergent
complete sets of mole­cular basis for this preference remains unknown, clinical and pathological features. CTNNB1 mutations,
chromosomes.
although a possible link may be the differential require- directly mediating stabilization and nuclear accumula-
ment for oncogenic WNT signalling levels that are tion of β-​catenin, associate with a ‘non-​proliferative’ class

Table 1 | Molecular and clinical features of APC-​mutant and RNF43-​mutant colorectal cancer
Feature APC-​mutant CRC Refs RNF43-​mutant CRC Refs
Histological classification Tubular adenoma 93
Serrated adenoma 69,70

Genetic subtype CIN, MSS 65,192


Mainly MSI-​H 52,67

CMS Enriched in CMS2 66


Enriched in CMS1 66

Sidedness Enriched in left-​sided CRC 64,65


Enriched in right-​sided CRC 65,74,75

Co-​current mutations KRAS, TP53 60,62–64


BRAF, ZNRF3 54,67,69

Order of events Early and rate-​limiting driver event in 60,62


Later event in serrated 52,67,193

traditional adenoma sequence carcinogenesis


Familial disorders FAP 46,55–57
Familial SSA 67,194,195

APC, adenomatous polyposis coli; CIN, chromosomal instability; CMS, consensus molecular subtype; CRC, colorectal cancer; FAP,
familial adenomatous polyposis; MSI-​H, microsatellite instability high; MSS, microsatellite stable; RNF43, ring finger 43; SSA, sessile
serrated adenoma; ZNRF3, zinc and ring finger 3.

naTure RevIeWs | CANceR volume 21 | January 2021 | 9


Reviews

Hypomorphic
of HCC, encompassing chromosomally stable tumours In agreement with this assumption, progressive loss of
A mutation resulting in that exhibit markers of differentiation and tend to retain APC regulatory repeat sequences due to incremental
partial loss of function hepatocyte-​like markers89. These tumours express a clas- shortening of the protein corresponds with a stepwise
of a gene, reducing but sical, canonical WNT activation gene expression signa- increase in WNT signalling in mouse embryonic stem
not abolishing protein
ture and generally associate with a better prognosis83. (ES) cells, human CRC cells and Drosophila embryos95–98.
function.
By contrast, AXIN1 mutations are confined to a ‘prolife­ Various lines of evidence indicate that APC variants
rative’ class of clinically aggressive HCC tumours, hall- truncated within the ‘mutation cluster region’ (MCR),
marked by poor differentiation, overexpression of cell ranging from codon 1286 to 1581, are best suited to
cycle markers and CIN89. In addition, AXIN1-​mutant generate a defined level of WNT signalling that is most
HCC tumours show activation of a diffe­rent tran­ optimal for tumour growth. The presence of monoal-
scriptional programme in comparison with CTNNB1 lelic or biallelic APC MCR truncations associate with
mutations83,84,90. the most severe forms of polyposis in patients with FAP,
Taken together, although mutations in CTNNB1 and whereas truncating mutations shifted towards the amino
AXIN1 both converge on the inhibition of β-​catenin terminus or C terminus correspond with more attenu-
turnover, their functional roles in liver carcinogenesis ated forms59,99. Furthermore, in sporadic CRC, 76% of
are markedly different. The molecular basis for this APC-​mutant cases carry at least one allele comprising
difference remains unresolved. a MCR truncation76,100. Finally, preclinical mouse mod-
els of FAP also revealed that the most severe intesti-
Molecular mechanisms of mutations nal polyposis occurs in heterozygous MCR-​truncated
In the traditional view, mutation-​induced biallelic inacti- Apc1322T/+ mice, followed by the commonly used APC
vation of tumour suppressors induces the inappropriate multiple intestinal neoplasia (ApcMin/+) mouse that car-
activation of signalling pathways that drive cell survival ries a shorter form, truncated at codon 850 (refs101–104).
and proliferation91. Accordingly, tumour suppressor Attenuated polyposis forms were observed for longer,
roles are typically studied using knockout approaches. hypomorphic APC truncations at codon 1572 or 1638 that
However, with the advances of cancer genome sequenc- retain more β-​catenin regulatory domains105,106 (Fig. 3a).
ing in human patients, a far more complex picture has Notably, in these mouse models, polyposis primarily
emerged. Mutations in tumour suppressor genes are develops in the small intestine, whereas in patients with
highly heterogeneous, giving rise to a range of errone- FAP and sporadic CRC the large intestine is the pre-
ous proteins with unknown activity that are not captured ferred site for polyp formation. The molecular basis for
well by knockout models. In addition, this heterogeneity this difference remains to be resolved but may relate to
poses a challenge to unambiguously distinguish passen- differences in physiological WNT gradients between the
ger mutations from driver mutations and to under- human and mouse intestinal tract77.
stand the underlying cancer-​promoting mechanisms. Which molecular features of MCR-truncated
Moreover, besides loss of their growth suppressive ability, APC allow for a specific level of β-​catenin-​mediated
mutant tumour suppressors may acquire novel functions trans­cription? First, loss of all three AXIN-binding
that promote tumorigenesis92. Herein, we discuss recent Ser-​Ala-Met-​Pro (SAMP) repeats appears to be a key
insights into how patient-​derived mutations alter WNT determinant for inactivation of the β-​catenin destruction
pathway tumour suppressor function in cancer. complex97,107. Despite lacking these conventional AXIN
binding sites, SAMP-​depleted APC truncations still form
Destruction complex mutations a complex with AXIN, likely by employing an alternative
APC truncating mutations in CRC generate a ‘just-​right’ binding site located within the Armadillo repeat (ARM)
level of WNT signalling. The tumour suppressor acti­ domain97,98,108–110 (Box 1). A second necessity is retention
vity of APC primarily depends on its interactions with of up to three β-​catenin-​binding 20Rs, through combined
β-​catenin and AXIN through multiple, short repeat biallelic APC truncations95,97. Although these remaining
sequences located in the carboxy-​terminal half of its 20Rs still promote β-​catenin phosphorylation, they do
long, intrinsically disordered region (Box 1). Although not facilitate β-​catenin destruction in APC-​depleted
CRCs usually carry biallelic APC mutations, APC func- cells108. Instead, these 20Rs are thought to confer residual
tion is typically not fully inactivated. In the vast majority regulatory activity by mediating cytoplasmic β-​catenin
of cases, the 5′ half of the gene is hit by nonsense or retention97. Finally, loss of the conserved catenin inhibi­
frameshift mutations that mediate expression of prema- tory domain (CID; or ‘sequence B’), located adjacent to
turely truncated APC protein variants (Fig. 2c). Notably, the second 20R (20R2) (Box 1), is an additional require-
mutations affecting both APC alleles are interdepend- ment for driving high basal levels of WNT pathway activ­
ent. In patients with FAP, germline mutations removing ity, as shown in intestinal organoids98. The underlying
all β-​catenin-​binding 20 amino acid repeats (20Rs) are mechanism needs better characterization but seems to
followed by somatic second-​hit mutations that retain involve a loss of interaction with α-​catenin, which pre-
at least one 20R93–95. Conversely, when the germline vents stable interactions of APC with phosphorylated
variant retains one or more 20Rs, the second hit will β-​catenin111. In addition, diminished binding of the E3
lose all 20Rs, either by upstream truncation or by allelic ligase β-​TrCP by CID-​deleted APC might further impede
loss93–95. These findings have led to the ‘just-​right’ signal­ β-​catenin ubiquitylation and proteolysis98,112. Moreover,
ling model, in which both APC alleles are selected to CID removal may promote recruitment of the enzyme
retain sufficient β-​catenin regulatory activity to generate ubiquitin-​specific processing protease 7 (USP7), which
an optimal window of signalling for tumour growth94. further enhances β-​catenin stability by de-​ubiquitylation98.

10 | January 2021 | volume 21 www.nature.com/nrc


Reviews

Box 1 | AXIN1-mediated and APC-​mediated organization of the β-​catenin destruction complex


The β-​catenin destruction complex is a central node of WNT signalling regulation27. The main activity of the complex is the
capturing and phosphorylation of β-​catenin, leaving a ‘recognition tag’ for subsequent ubiquitylation and proteasomal
degradation. The main organizer of the complex is the AXIN1 scaffold protein that interacts with all core components
of the complex and coordinates the activity of the kinases casein kinase 1 (CK1) and glycogen synthase kinase 3β (GSK3β).
AXIN1 employs its amino-​terminal RGS domain to interact with adenomatous polyposis coli (APC)10,23,199,200 and carries short
binding motifs for CK1 and GSK3β in its centrally located intrinsically disordered region10,26,201. The DIX domain located at
the AXIN1 carboxy terminus mediates self-​polymerization and nucleates the assembly of highly dynamic, spherically shaped
cytosolic puncta110,202,203. The role of APC is less well understood and primarily depends on its interactions with β-​catenin
and AXIN1 through multiple, short repeat sequences located in its long, intrinsically disordered region. For β-​catenin binding,
APC employs four clustered 15 amino acid repeat motifs (15Rs) and seven more dispersed 20 amino acid repeat motifs (20Rs)
that gain affinity for β-​catenin upon phosphorylation22,204. The second 20R (20R2) does not bind β-​catenin but cooperates
with an adjacent region called the catenin inhibitory domain (CID; or ‘sequence B’) to perform an essential but incompletely
understood step in the turnover of phospho-​β-​catenin95,97,110,111. Key interactions between APC and the AXIN1 RGS domain
are mediated by three Ser-​Ala-​Met-​Pro (SAMP) repeat motifs interspersed with 20R4 to 20R7 (refs199,201,205). A second, more
dynamic interaction with AXIN1 involves the APC N-​terminal Armadillo repeat (ARM) domain that interacts with the central
intrinsically disordered region of AXIN1 in a phosphorylation-​dependent manner110. The N terminus of APC carries
oligomerization activity206,207. The multimerizing properties of both AXIN1 and APC have led to the view that destruction
complex activity depends on the assembly of supermolecular non-​membrane-​bound compartments, often referred to as
biomolecular condensates143,208.
The figure shows the functional protein domains of APC, its close homologue APC2 as well as AXIN1 and AXIN2: ARM
mediates interaction with the cytoskeleton, protein phosphatase 2A (PP2A) and AXIN; 15R is for β-catenin binding; 20R is for
β-​catenin binding; CID is essential for promoting β-​catenin ubiquitylation; SAMP mediates AXIN interaction; basic region
mediates microtubule binding; end-​binding protein 1 (EB1) binding motif; mutation cluster region (MCR) of APC, which is
frequently hit by truncating mutations in human cancer; Tankyrase binding motif (TNKS); RGS mediates APC interaction;
and DIX mediates self-​polymerization. Binding motifs for GSK3β, β-catenin, Dishevelled (DVL) and CK1 are indicated.

AXIN1 or AXIN2 β-Catenin MCR β-Catenin


Oligomerization 15R 20R Basic region EB1
APC ARM
CID SAMP-AXIN1 or AXIN2
0 400 800 1,200 1,600 2,000 2,400 2,843
amino
acids
AXIN1 or AXIN2 β-Catenin
Oligomerization 20R Basic region EB1
APC2 ARM
CID SAMP-AXIN1 or AXIN2
0 400 800 1,200 1,600 2,000 2,303
amino
acids

APC AXIN1 or AXIN2, DVL


GSK3β β-Catenin CK1
AXIN1 TNKS RGS DIX

0 200 400 600 862


amino
acids

APC AXIN1 or AXIN2, DVL


GSK3β β-Catenin CK1
AXIN2 TNKS RGS DIX

0 200 400 600 843


amino
acids

Thus, MCR-​truncated APC balances loss of β-​catenin forms truncated after the CID display attenuated intes-
degradation with residual regulatory activity to generate tinal polyposis but acquire multiple tumours in extra-​
an optimal dosage of β-​catenin-​mediated transcription to intestinal tissues instead (for example, in the breast, liver
drive tumour growth. and bone)106,113 (Fig. 3a). In agreement, the fraction of
CID-​retaining APC truncations within TCGA database
Relevance of less prevalent truncated APC forms. APC is lowest in CRC (around 10% of 557 APC-​mutant cases)
truncations downstream of MCR have milder effects on but increases for other malignancies, such as stomach
WNT signalling, occur at a lower frequency in human adenocarcinoma (40% of 33) and uterine endometrioid
cancer and are less well studied. However, several lines of carcinoma (40% of 58), indicating that these cancer types
evidence indicate that these variants also perform tumo­ might select for lower WNT signalling levels (Fig. 3b,c).
rigenic driver roles. Mice expressing hypomorphic APC Notably, residual suppressor activity of mutant APC is

naTure RevIeWs | CANceR volume 21 | January 2021 | 11


Reviews

a Phenotype Refs
Multiple intestinal adenomas 101,102
Min
Severe intestinal adenomas 103,104

lower nuclear β-catenin


1322T
Extra-intestinal adenomas 106
1572T
Multiple intestinal adenomas 105,107

destabilized APC protein


1638N

MCR
Oligomerization 15R 20R Basic region EB1
APC ARM
CID SAMP
0 400 800 1,200 1,600 2,000 2,400 2,843
amino
99% sporadic CRC, acids
both alleles <1,581
76% sporadic CRC,
at least one allele >1,286

b 20R1–3 CID SAMP1 c


Uterine Uterine
endometrioid endometrioid
carcinoma carcinoma
Stomach Stomach
adenocarcinoma adenocarcinoma
Skin cutaneous Skin cutaneous
melanoma melanoma
Bladder Bladder
urothelial urothelial
carcinoma carcinoma CID lost
Colorectal Colorectal CID
adenocarcinoma adenocarcinoma retained
0 1,000 2,000 3,000 0 0.2 0.4 0.6 0.8 1.0
Position of APC truncations Fraction of APC truncations

Fig. 3 | Correlation of APC truncation position and CID domain loss with tumour phenotype. a | Top: truncating
adenomatous polyposis coli (APC) mutation positions and corresponding tumour phenotypes in mouse models. Truncating
APC mutations are plotted against APC protein structure101–106. Bottom: the mutation cluster region (MCR) covers the
majority of APC mutations found in human colorectal cancer (CRC)76. b | Amino acid position of truncating (nonsense and
frameshift) mutations in APC shown for five cancer types with >5% frequency of APC mutations197. Positions of the first three
β-​catenin-​binding 20 amino acid repeats (20Rs) are indicated in purple. The first AXIN1 and AXIN2 binding Ser-​Ala-​Met-​Pro
(SAMP) repeat is indicated in pink. This repeat is lost in 99% of CRCs. The catenin inhibitory domain (CID) is indicated in
green. Medians are indicated by a red line. c | Truncating APC mutations in The Cancer Genome Atlas (TCGA) cancer types
from part b plotted for the fraction of mutations that retain or lose an intact CID. ARM, Armadillo repeat; EB1, end-​binding
protein 1; 15R, 15 amino acid repeat motif.

concentration-​dependent and may be further modulated similar structural organization to APC but lacks the
by alterations in protein stability105. In line with this, β-​catenin-​binding 15 amino acid repeat (15R) region116,117
a partial reduction in APC protein levels can sensitize (Box 1). Whereas APC2 itself displays weak β-​catenin
colon cells to WNT stimulation109. Furthermore, cooper- down­regu­lating activity, its efficiency is enhanced by
ation with driver mutations of other pathways may alter heterodimerization with APC116–118. In fact, APC2 bind-
WNT dosage effects and tissue preference, as shown for ing to MCR-​truncated APC promotes residual levels of
hypomorphic Apc1572T mice that acquire a prominent β-​catenin degradation in CRC cells118. Although this
intestinal polyposis phenotype upon co-​introduction activity cannot fully compensate for the loss of APC
of SMAD4 mutations106. Lastly, it is noteworthy that function in the colon 117, a large fraction of CRCs
APC mutations may activate alternative driver path- (43–95%, 29/66 and 97/102) display APC2 promoter
ways in cancer, such as Hippo–Yes-​associated protein hypermethylation, indicating that higher levels of
(YAP) signalling and CIN114,115. Although these events β-​catenin activity acquired by loss of APC2 expression
are outside the scope of this Review, they evidently will might provide a growth advantage119,120.
contribute to APC mutation-​driven cancer development. The APC2-​dependent regulation of the WNT path-
way displays tissue specificity. For instance, deletion of
Role of APC2 activity in APC-​mutant cancer. The either Apc or Apc2 from the mouse mammary epithe-
severity of APC mutations is balanced by expression lium did not affect tissue homeostasis, whereas simulta-
of its functional homologue APC2 that displays a neous deletion of Apc and Apc2 led to excessive levels of

12 | January 2021 | volume 21 www.nature.com/nrc


Reviews

stabilized β-​catenin and mammary tumour formation121. NSCLCs130. As RNF146 is capable of degrading both
Thus, in contrast to the colon, APC2 appears to com- AXIN paralogues, upregulation of this E3 ligase might
pensate for APC loss in breast tissue. Whether and how be more potent than individual mutations in AXIN1
APC2 expression balances WNT signalling thresholds in or AXIN2.
other tissues thus deserves further investigation.
Consequences of AXIN2 mutation. Although AXIN2-​
Consequences of AXIN1 loss of function. The mutational mediated negative feedback appears insufficient to fully
spectrum of AXIN1 across all cancer types shows that compensate for AXIN1 loss125,128, somatic mutations in
deep deletion occurs in 24% (57/240) of cases (Fig. 2c). AXIN2 are prevalent in various cancer types (Fig. 2a).
Owing to its key organizing role in destruction com- In addition, human germline mutations in AXIN2 are
plex assembly (Box 1), loss of AXIN1 is envisioned to associated with predisposition to lung cancer131,132, breast
induce WNT-​independent β-​catenin activation in cells. cancer133,134 and CRC133,135–137. Evaluation of the mode of
Indeed, Axin deletion corresponds with prominent action by which AXIN2 variants drive tumorigenesis is
WNT-​mediated phenotypes in Caenorhabditis elegans limited. In one example, a recurring truncated AXIN2
and Drosophila and induces embryonic lethality in mutant showed partial LOF upon ectopic expression
mice122–124. To study the potential cancer driver role of in cell lines138. Given the common co-​occurrence of
AXIN1 loss in somatic cells, mouse models for condi- AXIN2 mutations with other WNT pathway mutations
tional disruption of Axin1 expression were developed in cancer, weak pathway activation by AXIN2 LOF may
and, owing to the prevalence of AXIN1 mutations in act synergistically with these mutations to increase
HCC83, evaluated for the effects on liver homeosta- signalling levels or provide drug resistance138–140. Further
sis90,125. Loss of Axin1 in the adult liver induced an acute characterization of the molecular mechanisms of AXIN2
increase in hepatocyte proliferation and, at 12 months mutations and their cooperativity with other genetic
of age, 40–55% (27/67 and 5/9) of mice developed liver alterations is required to gain a better understanding
tumours that histologically resemble human HCC90,125. of how these mutations contribute to the initiation and
Contrary to Apc deletion or activating Ctnnb1 muta- progression of cancer growth.
tions, Axin1-​deleted hepatocytes and tumours displayed
no or only a weak increase in the expression of con- Mechanisms of AXIN1 missense mutations: beyond loss
ventional WNT target genes125,126. However, the Axin1 of function. Despite the useful knowledge gained from
loss-​induced gene signature showed strong similarity studying AXIN1 deletion, the largest fraction of genetic
to the poor-​prognosis ‘proliferative’ subclass of HCC alterations of AXIN1 in human cancer comprise mis-
in which AXIN1 mutations are commonly found83,90. sense mutations (Fig. 2c) that are found scattered over
Characteristic features comprised enhanced levels of the entire length of the gene, complicating functional
Notch and YAP signalling and a strong induction of fetal predictions. In-​depth analysis of missense mutations
gene expression90. in the N-​terminal RGS domain recently revealed that
An evaluation of the consequences of AXIN1 loss the ensuing mutant AXIN1 proteins may acquire novel
in other cell types and tissues is limited. A CRISPR functional properties to drive tumour growth (Box 1).
knockout screen performed in human small intestinal Mechanistically, missense driver mutations were found
organoids under conditions of WNT and RSPO with- to structurally destabilize the RGS domain and medi-
drawal promoted selective outgrowth of organoids ate formation of small-​scale aggregates that rewire the
depleted of either APC or AXIN1 (ref.127). Furthermore, AXIN1 interactome to promote basal WNT pathway
long-​t erm treatment of a WNT-​h ypersensitive, activation141 (Fig. 4). Furthermore, insertion of mis-
RSPO3-​overexpressing CRC line with potent WNT sense RGS-​mutant AXIN1 in the endogenous locus
secretion inhibitors induced outgrowth of a resistant promoted tumour growth in vivo in Drosophila, con-
clone that acquired biallelic frameshift deletions in firming the cancer driver activity. The role of aggrega-
AXIN1 (ref.128). Thus, these results confirm that AXIN1 tion as the underlying mechanism was validated by the
loss mediates increased levels of WNT signalling in introduction of secondary mutations that suppressed
intestinal organoids. In addition, induction of the nega­ aggregation and rescued tumorigenic effects141. Thus,
tive feedback regulator AXIN2 was apparently not these findings indicate that tumour suppressor mis-
sufficient to counteract the overall increase in WNT sense mutations may contribute to cancer growth via
signalling in these cells128. unforeseen mechanisms.
Lastly, whereas AXIN1 mutations are detected in However, the relevance of most other missense muta­
only a small fraction of CRCs (3%, 19/549)64, depletion tions remains unknown. Owing to the role of AXIN1 as
of AXIN1 might occur through alternative mechanisms. a hub protein as well as its ability to self-​polymerize142,
For instance, transcriptional upregulation of the E3 multiple modes of interference with these functions
ubiquitin ligase RNF146 promotes CRC proliferation by mutations can be envisioned. Mechanistic insight
and survival via ubiquitylation-​mediated degradation will help classify the large number of reported muta-
of AXIN1 and activation of β-​catenin-​mediated tran­ tions according to their mode of action. In addition,
s­cription129. Overexpression of RNF146 occurs in 6% as destruction complex activity depends on the abso-
(13/205) of CRC cases and correlates with poor pro­g­ lute levels and ratio of APC and AXIN1 proteins143,
nosis129. Similarly, RNF146 overexpression was identi- evaluation of molecular and cellular consequences
fied to drive proliferation and invasion via enhanced of mutant AXIN1 at the endogenous level will be
WNT–β-​c atenin signalling in a large fraction of essential.

naTure RevIeWs | CANceR volume 21 | January 2021 | 13


Reviews

Wild-type AXIN1
Puncta In human cancer, the mutational spectrum of RNF43
is dominated by truncating and missense mutations,
whereas ZNRF3 is more frequently hit by missense
mutations and deletions (Fig. 2c). Functional analysis of
APC different RNF43 mutational classes indicates that LOF
β-TrCP can be achieved by various mechanisms (Fig. 5a–d). First,
AXIN1 RGS DIX mutations that truncate RNF43 upstream of the E3 ligase
CK1 β-Catenin GSK3β RING domain will cause either loss of protein expres-
P P P sion or expression of non-​functional protein (Fig. 5b).
A prominent example is the R117Afs*41 RNF43 muta-
tion that accounts for 8–12% (4/51 and 6/50) of RNF43
mutations in CRC and 4% (2/58) of RNF43 mutations in
endometrium cancer52. Second, missense mutations
in the extracellular domain, as reported in pancreatic
RGS missense mutant AXIN1 Nano-aggregates cancer145, can mediate RNF43 retention in the endo-
plasmic reticulum (ER), likely by misfolding and ER
quality control-​mediated retention and clearance149.
Owing to their inability to reach the plasma membrane,
AXIN1 DIX these mutants fail to downregulate WNT receptors.
CK1 GSK3β
RGS Furthermore, ER-​trapped RNF43 mutants may harbour
β-Catenin dominant negative activity and thus have increased
β-Catenin β-Catenin tumorigenic capacity compared with gene deletions149
(Fig. 5c). A third mechanism comprises depletion of
Fig. 4 | Tumorigenic mechanism of AXIN1 missense mutations. Top: wild-​type AXIN1 mutant RNF43 via nonsense-​m ediated mRNA decay
recruits and organizes the main components of the β-​catenin destruction complex to (NMD), which may be induced following the gain of
facilitate efficient β-​catenin degradation. AXIN1 self-​polymerizes via its DIX domain premature termination codons (PTC)150 (Fig. 5d). In sup-
to form large biomolecular condensates that are visible as spherically shaped cytosolic port of this, NMD inhibition stabilized the mRNA of
puncta in the cell. Immunofluorescent microscopy image shows wild-​type AXIN1 several trunca­ting RNF43 variants in CRC cell lines151.
overexpressed in HeLa cells. Bottom: cancer missense mutations (red stars) in the
However, the role of NMD has been debated, particu-
AXIN1 RGS domain disrupt its structure and promote the formation of small-​scale AXIN1
larly for the RNF43 G659Vfs*41 hotspot mutation that
aggregates that form a dysfunctional complex that fails to form puncta and displays
impaired capacity to degrade β-​catenin. Immunofluorescent microscopy image shows is highly prevalent in colorectal, endometrium and
AXIN1 RGS missense mutant overexpressed in HeLa cells. Fluorescently labelled AXIN1 stomach MSI-​high cancer subsets52,152. Because the
is shown in white, nucleus of the cell (DAPI staining) is shown in blue. APC, adenomatous RNF43 G659Vfs*41 protein appears fully functional
polyposis coli; β-​TrCP, β-​transducin repeat-​containing protein; CK1, casein kinase 1; in overexpression assays153,154, NMD-​mediated mRNA
GSK3β, glycogen synthase kinase 3β; P, phosphorylation. Immunofluorescence images depletion was proposed as a mode of action. Indeed,
are original, unpublished images from the authors’ examination of cells. patient-​derived or genetically engineered human colon
organoids harbouring RNF43 G659Vfs*41 combined
Mutations in WNT receptor regulators with ZNRF3 inactivation acquired RSPO-​independent
RNF43 and ZNRF3 loss-​of-​function mutations. Gene growth properties, suggesting LOF effects67,155. Also, in
deletion experiments in mice revealed that the tumour line with NMD susceptibility, RNF43 mRNA levels were
suppressor roles of RNF43 and ZNRF3 are redundant decreased in G659Vfs*41-bearing human tumours69.
in the intestine and that simultaneous inactivation of However, these findings were challenged recently
both genes is a requirement for tumour growth14,144. by a study in which RNF43 G659Vfs*41 mRNA was
In line with these findings, human serrated CRCs often found stably expressed. Moreover, deletion of this
present with mutated RNF43 and concomitant down- variant from CRC cell lines still induced an increase in
regulation of ZNRF3, either by mutations or epigenetic WNT–β-​catenin signalling, suggesting that LOF of this
mechanisms54. However, modelling RNF43 and ZNRF3 RNF43 mutant is incomplete154. Although the reason for
loss in various cells and tissues indicates context-​ these discrepancies remains unknown, the mutational
dependent effects, in which one of the paralogues might background and tissue context might play a role.
be dominant over the other. For instance, individual
deletion of mouse Rnf43 drives inappropriate WNT sig- An oncogenic role for RNF43 truncations in cancer.
nalling and hyperplastic growth in the pancreas145 and Accor­ding to the rules of NMD, nonsense or frameshift
stomach146, whereas Znrf3 loss drives tumour formation muta­tions that hit the centre of the large penultimate
Nonsense-​mediated mRNA in the adrenal cortex147. Notably, RNF43 LOF mutation-​ exon of RNF43 have a high probability to escape NMD150.
decay
bearing WNT-​hypersensitive CRC subsets frequently How these truncated RNF43 variants interfere with
(NMD). A molecular
mechanism of mRNA display epigenetic downregulation of additional WNT WNT signalling and tumorigenesis remains poorly
surveillance that recognizes pathway negative feedback regulators, including AXIN2, understood, as current knowledge largely relies on gene
and eliminates mRNAs as well as NOTUM and WNT inhibitory factor 1 (WIF1), knockout strategies. A recent study revealed a class of
harbouring premature that counteract the activity of WNTs148. Thus, such alter- truncated RNF43 variants that promotes tumorigenesis
termination codons, thus
contributing to the protection
ations may further enhance the duration and intensity by a mechanism distinct from LOF39. These RNF43
of eukaryote cells from of the WNT responses of RNF43- and ZNRF3-​mutant variants that have lost the distal half of their cytosolic
putative toxic proteins. cells and promote tumour progression. tail (Fig. 5a) fully retained their ability to downregulate

14 | January 2021 | volume 21 www.nature.com/nrc


Reviews

WNT receptors yet induced WNT-​independent RNF43 tail to facilitate its role in β-​catenin downregu­
β-​catenin-​mediated transcription (Fig. 5e). Loss of the lation (Figs 1b,5a). Upon truncation, the interaction of
C terminus was found to interfere with a second suppres- RNF43 with the destruction complex is misregulated,
sor role of RNF43 that involves binding and regulating leading to trapping of AXIN1 and CK1 at the plasma
the activity of the β-​catenin destruction complex. In par- membrane and, at the same time, β-​catenin stabiliza-
ticular, the kinase CK1 was identified as a dynami­cally tion and target gene transcription (Fig. 5e). Furthermore,
regulated interaction partner that phosphorylates the the introduction of RNF43 truncating mutations in

a
DIR CK1 Onco
RNF43 SP PA TM RING

0 200 400 600 783


amino
DIR acids
ZNRF3 SP PA TM RING

0 200 400 600 800 936


amino
acids
b Loss of functional domains c Dominant negative mislocalization

RNF43 FZD WNT LRP5 WNT


or LRP6

DVL DVL
Ubiquitination APC AXIN1 APC AXIN1
GSK3β Mutant GSK3β
CK1 RNF43 CK1
Wild-type RNF43
or ZNRF3
Hypersensitive Hypersensitive
WNT signalling WNT signalling

ER

d RNF43 loss of function e Oncogenic RNF43


WNT Oncogenic
RNF43

P P
DVL CK1 CK1 Ub
APC AXIN1 AXIN1AXIN1 Endosome Ub
NMD GSK3β Ub
CK1

Nonsense
RNF43 mRNA AXIN1/2
WNT- CK1 β-Catenin
Hypersensitive independent Lysosomal
WNT signalling signalling degradation APC GSK3β

Fig. 5 | RNF43 and ZNRF43 structural organization and tumorigenic mechanisms of RNF43 mutations. a | WNT target
genes ring finger 43 (RNF43) and zinc and ring finger 3 (ZNRF3) are essential regulators of stem cell homeostasis. The
encoded homologous transmembrane proteins act in a negative feedback loop by employing their intracellular E3-​ligase
catalytic RING domain to ubiquitylate and remove the receptors Frizzled (FZD) from the cell surface by endocytosis13,14.
The mode of RNF43- and ZNRF3-​mediated substrate recognition remains poorly characterized but might involve an
interaction of the centrally located Dishevelled-​interaction region (DIR) with the Dishevelled (DVL) protein that binds
FZD under WNT-​stimulated conditions198. More distal carboxy-​terminal interactions with casein kinase 1 (CK1) underlie
a second RNF43-​mediated suppressor role that facilitates β-​catenin degradation by the destruction complex. This role
is exploited by truncating cancer mutations in a defined region of the RNF43 C-​terminal tail to drive oncogenic WNT
signalling39. b | RNF43 truncations that remove functional intracellular domains lead to hypersensitive WNT signalling.
c | RNF43 missense mutations that mislocalize and are dominant negative over wild-​type RNF43 and ZNRF3 alleles drive
hypersensitive WNT signalling. d | RNF43 truncations that lead to loss of function by nonsense-​mediated mRNA decay
(NMD) mediate hypersensitive WNT signalling. e | Oncogenic RNF43 truncations that trap CK1 and AXIN1 at the plasma
membrane drive WNT-​independent β-​catenin signalling activation. APC, adenomatous polyposis coli; ER, endoplasmic
reticulum; GSK3β, glycogen synthase kinase 3β; LRP5, low-​density lipoprotein receptor-​related protein 5; Onco, region in
RNF43 that mediates oncogenic activity upon truncation (amino acids 504–563); P, phosphorylation; PA, protease-​associated
domain; SP, signal peptide; TM, transmembrane domain; Ub, ubiquitin.

naTure RevIeWs | CANceR volume 21 | January 2021 | 15


Reviews

primary human colon organoids induced activation of XAV939 inhibits the poly(ADP-​ribosyl)ating (PARylating)
a β-​catenin-​mediated growth programme in the absence enzymes Tankyrase 1 (TNKS1) and TNKS2, which prevents
of ZNRF3 mutations, thus confirming their dominant AXIN1 and AXIN2 PARylation as well as subsequent
oncogenic activity39. RNF146-​mediated ubiquitylation and proteasomal
Unlike APC or RNF43 LOF mutations67,156,157, onco- degradation161. Besides promoting AXIN stabilization,
genic RNF43 mutations induced a state of senescence-​ XAV939 treatment enhances levels of TNKS1 and
like growth arrest, and their expression was tolerated TNKS2, which might further contribute to WNT path-
only in TP53-​null human colon organoids39. These find- way suppression by a non-​catalytic scaffolding role, via
ings indicate that, like many conventional oncogenes158, co-​polymerization with AXIN1 and AXIN2162. Despite
these truncated RNF43 variants induce oncogenic these promising findings, follow-​up studies investi-
stress and might represent a late event during tumori­ gating the relationship between truncated APC length
genesis. Indeed, patient-​derived tumours carrying and sensitivity to TNKS inhibitors revealed conflicting
oncogenic RNF43 mutations harboured mutations in results100,163,164. In particular, cell lines carrying APC
senescence-​related pathways as well39. Another impor- MCR truncations (retaining one or two 20Rs) displayed
tant distinction from LOF mutations is that expression resistance to TNKS inhibition in vitro, whereas tumours
of oncogenic RNF43 variants confers resistance to carrying such APC mutations displayed greater sensi-
anti-​WNT-​based therapy39. Thus, improved knowledge tivity to treatment in vivo in mouse models, in com-
of tumour-​promoting mechanisms of patient-​derived parison with shorter APC truncations163,164. Potentially,
RNF43 mutations will help to classify and predict this discrepancy might be explained by differences in
therapeutic efficacy, as discussed below. expression levels of the negative regulators of the WNT
pathway AXIN2 and APC2, both of which are stabilized
Opportunities for precision medicine upon TNKS inhibitor treatment139,165. Alternatively,
The pivotal role of WNT–β-​catenin signalling alterations resistance to TNKS inhibition might be promoted by
in human cancer growth has attracted great interest in the co-​o ccurring oncogenic alterations in other cancer
development of strategies for therapeutic targeting of this pathways as shown for mutant KRAS166, although here
pathway. These ambitions were further enforced by the the underlying mechanism remains to be resolved.
observation that, even at advanced stages of APC-​mutant A further concern is the observation that elevated doses
intestinal cancer in mice, restoration of APC expression of single-​agent TNKS inhibitor treatment necessary for
drives differentiation of tumour cells and restoration of tumour inhibition elicits intestinal toxicity, weight loss
normal tissue homeostasis159. Despite the identification and death in rodents163,167. Nevertheless, the application
of numerous WNT pathway antagonists as well as inhibi­ of reduced doses of TNKS inhibitors in combination
tory small molecules, none of these drugs has achieved treatment regimens revealed promising antitumour
clinical approval. A major challenge for clinical targeting effects163,167. Thus, elucidation of how cellular context
is to inhibit WNT signalling in cancer cells while avoid- impacts on TNKS inhibitor effectiveness as well as their
ing on-target toxicity in healthy tissues owing to altera- value in combination with other targeted drugs will be
tions in adult stem cell homeostasis. For a comprehensive essential to determine optimal strategies for therapeutic
evaluation of WNT pathway inhibitors that are currently application.
being evaluated in clinical trials, we refer the reader to Another potential vulnerability that emerges upon
other recent reviews17,160. Below, we discuss how recent APC truncation was recently revealed. In CRC cells
advances in the understanding of WNT–β-​catenin path- displaying loss of the APC CID domain, increased USP7-
way tumour suppressor alterations in cancer cells revealed mediated de-​ubiquitylation of β-​catenin was identi-
druggable vulnerabilities and facilitated the discovery of fied as a main driver of aberrant WNT signalling98.
drugs with potential for the development of innovative Consequently, treatment with USP7 inhibitors sup-
therapeutic approaches. In addition, we highlight how pressed WNT activation in APC-​mutant CRC cell lines
accumulating insights into the mechanism of action of and intestinal organoids and inhibited tumour growth
patient-​d erived mutations improves stratification in vivo98,168,169. However, these findings were challenged
of patients for applications of precision medicine. by a recent study in which USP7 inhibitors were instead
identified as positive regulators of WNT–β-​catenin
Targeting destruction complex activity signalling, via inhibition of AXIN stabilization170. The
Improving β-​catenin destruction complex activity in molecular basis for discrepancy between these studies
cancer cells imposes a challenge, owing to the frequent remains unclear and will require a careful comparison of
loss of key protein–protein interaction elements of its cellular systems and different classes of USP7 inhibitors
main organizers, APC and AXIN. Yet recent progress in that were used.
understanding the molecular events that drive WNT sig- In a screen for compounds that co-​regulate β-​catenin
nalling following such LOF events has revealed several and RAS protein stability, a small molecule was iden-
targetable vulnerabilities27. tified that enhances destruction complex activity. This
Residual destruction complex activity observed in compound, called KYA1797, binds the AXIN1 RGS
tumours carrying APC truncations offers an opportu­nity domain in a surface located close to the APC binding
to enhance β-​catenin turnover in cancer cells. The small site171. By an incompletely understood mechanism,
molecule XAV939 was identified as a potent enhancer of KYA1797–AXIN1 binding increases GSK3β activity
destruction complex activity by promoting AXIN stabili- and promotes both β-​catenin and RAS degradation171.
zation in APC-​mutant CRC cell lines161. Mechanistically, In line with these findings, follow-up work indicates that

16 | January 2021 | volume 21 www.nature.com/nrc


Reviews

Dysgeusia
β-catenin and RAS interact directly and, furthermore, Finally, selective blockade of WNT3 binding to the LRP6
Distortion of the sense of taste. degradation of β-​catenin precedes GSK3β-​dependent receptor by single-​chain antibodies inhibited hyper-
RAS degradation 172. KYA1797 inhibits the growth proliferation and promoted terminal differentiation of
of CRCs that carry APC and KRAS mutations in Rnf43- and Znrf3-​mutant mouse intestinal organoids186.
mice via the degradation of both RAS and β-catenin Thus, more selective targeting of WNT signalling by
oncogenes171,172. antagonizing specific receptors or WNTs holds pro­mise
to overcome some of the limitations associated with
Targeting WNT-​dependent cancers PORCN inhibitor treatment.
The discovery that the RNF43 and ZNRF3 muta- Lastly, the spectrum of targetable WNT-​dependent
tions confer a state of WNT-​dependent growth has tumours might be broader then currently anticipated.
opened up the possibility to identify cancer subsets Unexpectedly, the growth of APC-​mutant human gas-
that are sensitive to treatment with WNT inhibitors tric cancer tumours was inhibited in vitro and in vivo
or receptor-​blocking antibodies13,14,145. A highly potent by treatment with Vantictumab (OMP-18R5), an
approach to block autocrine and paracrine WNT signal- inhibi­tor of several FZDs that has been evaluated in
ling is inhibition of the production of all active WNTs phase Ib clinical trials for advanced pancreatic, lung
by targeting Porcupine (PORCN), an ER-​resident and breast cancer 187,188. The extent to which other
O-​acyltransferase that mediates WNT palmitoylation, an human cancer (sub)types carrying downstream WNT
essential step in WNT biosynthesis173. Indeed, preclini- pathway mutations may still depend on WNT recep-
cal data showed that PORCN inhibitors effectively sup- tor activity for their survival has yet to be investigated.
pressed WNT-​driven growth of engineered Rnf43- and Additionally, mutations outside the WNT pathway may
Znrf3-​mutant mouse intestinal tumour models as well also yield WNT-​hypersensitive tumours, as indicated
as patient-​derived RNF43-​mutant CRC and pancreatic for combined mutations in TP53 and CDH1 (encoding
ductal adenocarcinoma (PDAC) organoids69,155,174. In an E-​cadherin) in gastric cancer49. Thus, there might be an
effort to exploit this vulnerability, ongoing clinical trials unexplored subset of cancers that could benefit from
employ RNF43 mutations as a predictive biomarker for WNT inhibition.
treatment of patients with CRC and PDAC with PORCN
inhibitors175–180. Despite promising preclinical results, Conclusions and perspectives
on-​target adverse effects by PORCN inhibitors including Knowledge of the molecular mechanisms underlying
bone loss and dysgeusia were reported, likely limiting the tumorigenesis is a necessary step for the implementa-
dosages that can be applied in the clinic181–183. However, tion of precision cancer medicine. Although model-
recent work indicates that PORCN inhibitors might be ling WNT tumour suppressor inactivation using gene
used successfully at lower doses in drug combination deletion approaches in cells and organisms has gener-
strategies, offering improved antitumour efficacy183. ated a wealth of information, these models only repre-
In particular, dual treatment with PORCN and PI3K sent a limited part of the mutational spectra identified
inhibi­tors potently inhibited the growth of PDAC xeno­ in human cancer genomes (Fig. 2c). Moreover, current
grafts, owing to suppression of cell proliferation and prediction algorithms for driver mutations in WNT
glucose metabolism184. Thus, evaluation of synergistic pathway tumour suppressors are suboptimal owing to
effects of PORCN inhibitors with other targeted drugs the scattered distribution of mutations as well as the
appears to be a valuable strategy that deserves explora- presence of large, intrinsically disordered regions in
tion for the treatment of WNT-​hypersensitive cancers. the encoded proteins189,190 (Box 1). As illustrated by the
Furthermore, based on recent insights, the efficacy of examples given in this Review, a more comprehensive
PORCN inhibitors may be enhanced by improved strati­ experimental evaluation of the functional effects of
fication of patients with RNF43 mutations. To achieve different classes of patient-​derived tumour suppressor
this, discrepancies in the classification of RNF43 muta- mutations is needed to distinguish driver mutations from
tions as passengers or drivers will need to be resolved. passenger mutations, define common mechanistic rules,
Moreover, patients who carry RNF43 mutations that understand tissue specificity and improve guidance for
promote WNT-​independent growth39, or other coexis­ therapeutic decisions. Recent advances in genome edit-
ting mutations that predict resistance to PORCN ing and applications of organoids as faithful models of
inhibition128,154, would need to be excluded. Finally, in tumour growth and metastasis are timely developments
tissues where RNF43 and ZNRF3 display redundancy, that will be of great value to address these issues191.
the status of ZNRF3 expression will help to confirm An immediate challenge for successful application
WNT-​addicted tumours. of WNT-​blocking or receptor-​blocking agents is the
As PORCN inhibitors affect all WNT-​mediated path- robust identification of WNT-​addicted cancer subsets.
ways, more selective strategies are being developed to Eligible patients may be stratified based on the presence
mitigate on-​target toxicity. For example, a genome-​wide of confirmed RNF43 LOF mutations and, in specific
CRISPR screen identified FZD5 as the primary WNT tissues, concomitant LOF of ZNRF3. In addition, a more
receptor essential for viability in RNF43-​mutant PDAC comprehensive search for additional predictive markers
cell lines185. In agreement, FZD5 inhibitory antibodies that signify WNT-​d ependent growth, as reported
blocked the growth of RNF43-​mutant PDAC cell lines for combined TP53 and CDH1 mutations in gastric
in vitro and in vivo185. In addition, FZD5 targeting cancer49, deserves further exploration and might help
inhibited survival of RNF43-​mutant CRC organoids, to expand targetable patient subsets. Furthermore, an
whereas APC-​mutant organoids remained unaffected185. in-​depth analysis and understanding of how epigenetic

naTure RevIeWs | CANceR volume 21 | January 2021 | 17


Reviews

silencing and transcriptional loss of negative feedback by in vitro drug screening using patient-​derived tumour
regulators contribute to the development and progres- organoids.
sion of WNT-​driven cancers will be needed to iden- Finally, although we focused on the molecular con-
tify and expand clinically actionable patient subsets. sequences of individual WNT tumour suppressor muta-
Of equal importance is the need to explore mutations tions, clearly the roles of mutational context, tumour
and pathways that provide resistance to WNT-​blocking heterogeneity and interactions with other major cancer
or receptor-​blocking therapy. Besides a specific set of pathways deserve further investigation. Hopefully, by
truncating RNF43 mutations, downstream pathway combining molecular insights with information coming
mutations in APC and AXIN1 were identified as poten- from early-​phase clinical trials, the efforts to target WNT
tial escape routes, although such effects might depend signalling will translate soon into effective therapies.
on mutation positioning and the tissue context39,128,154,188.
The identification of reliable markers may be accelerated Published online 23 October 2020

1. Nusse, R. & Varmus, H. E. Many tumors induced by 20. Rubinfeld, B. et al. Association of the APC gene product PORCN inhibitor treatment. This work illustrates
the mouse mammary tumor virus contain a provirus with β-​catenin. Science 262, 1731–1734 (1993). the relevance of studying patient-​derived mutations
integrated in the same region of the host genome. 21. Munemitsu, S., Albert, I., Souza, B., Rubinfeld, D. & for understanding disease mechanisms and
Cell 31, 99–109 (1982). Polakis, P. Regulation of intracellular β-​catenin levels applications of precision medicine.
2. Nusse, R. & Varmus, H. Three decades of Wnts: by the adenomatous polyposis coli (APC) tumor-​ 40. Zebisch, M. et al. Structural and molecular basis
a personal perspective on how a scientific field suppressor protein. Proc. Natl Acad. Sci. USA 92, of ZNRF3/RNF43 transmembrane ubiquitin ligase
developed. EMBO J. 31, 2670–2684 (2012). 3046–3050 (1995). inhibition by the Wnt agonist R-​spondin. Nat. Commun.
3. Korinek, V. et al. Depletion of epithelial stem-​cell 22. Rubinfeld, B. et al. Binding of GSK3β to the 4, 2787 (2013).
compartments in the small intestine of mice lacking APC–β-​catenin complex and regulation of 41. Chen, P. et al. The structural basis of R-​spondin
Tcf-4. Nat. Genet. 19, 379–383 (1998). complex assembly. Science 272, 1023–1026 (1996). recognition by LGR5 and RNF43. Genes Dev. 27,
4. Huelsken, J., Vogel, R., Erdmann, B., Cotsarelis, G. 23. Hart, M. J., De Los Santos, R., Albert, I. N., Rubinfeld, B. 1345–1350 (2013).
& Birchmeier, W. β-​Catenin controls hair follicle & Polakis, P. Downregulation of β-​catenin by human 42. Szenker-​Ravi, E. et al. RSPO2 inhibition of RNF43
morphogenesis and stem cell differentiation in Axin and its association with the APC tumor and ZNRF3 governs limb development independently
the skin. Cell 105, 533–545 (2001). suppressor, β-​catenin and GSK3β. Curr. Biol. 8, of LGR4/5/6. Nature 557, 564–569 (2018).
5. Pinto, D., Gregorieff, A., Begthel, H. & Clevers, H. 573–581 (1998). 43. Lebensohn, A. M. & Rohatgi, R. R-​spondins can
Canonical Wnt signals are essential for homeostasis 24. Hart, M. et al. The F-​box protein β-​TrCP associates potentiate WNT signaling without LGRs. eLife 7,
of the intestinal epithelium. Genes Dev. 17, with phosphorylated β-​catenin and regulates its e33126 (2018).
1709–1713 (2003). activity in the cell. Curr. Biol. 9, 207–211 (1999). 44. Park, S. et al. Differential activities and mechanisms
6. Tan, X., Behari, J., Cieply, B., Michalopoulos, G. K. 25. Amit, S. et al. Axin-​mediated CKI phosphorylation of the four R-​spondins in potentiating Wnt/β-​catenin
& Monga, S. P. S. Conditional deletion of β-​catenin of β-​catenin at Ser 45: a molecular switch for the signaling. J. Biol. Chem. 293, 9759–9769 (2018).
reveals its role in liver growth and regeneration. Wnt pathway. Genes Dev. 16, 1066–1076 (2002). 45. Dubey, R. et al. R-​Spondins engage heparan sulfate
Gastroenterology 131, 1561–1572 (2006). 26. Liu, C. et al. Control of β-​catenin phosphorylation/ proteoglycans to potentiate Wnt signaling. eLife 9,
7. Yang, W. et al. Wnt/β-​catenin signaling contributes to degradation by a dual-​kinase mechanism. Cell 108, 1–25 (2020).
activation of normal and tumorigenic liver progenitor 837–847 (2002). 46. Kinzler, K. W. et al. Identification of FAP locus genes
cells. Cancer Res. 68, 4287–4295 (2008). 27. Van Kappel, E. C. & Maurice, M. M. Molecular from chromosome 5q21. Science 253, 661–665
8. Morin, P. J. et al. Activation of β-​catenin–Tcf signaling regulation and pharmacological targeting of the (1991).
in colon cancer by mutations in β-​catenin or APC. β-​catenin destruction complex. Br. J. Pharmacol. 47. Liu, W. et al. Mutations in AXIN2 cause colorectal
Science 275, 1787–1790 (1997). 174, 4575–4588 (2017). cancer with defective mismatch repair by activating
9. Zeng, L. et al. The mouse fused locus encodes Axin, 28. Tamai, K. et al. LDL-​receptor-related proteins in Wnt β-​catenin/TCF signalling. Nat. Genet. 26, 146–147
an inhibitor of the Wnt signaling pathway that signal transduction. Nature 407, 530–535 (2000). (2000).
regulates embryonic axis formation. Cell 90, 29. Mao, J. et al. Low-​density lipoprotein receptor-​related 48. Satoh, S. et al. AXIN1 mutations in hepatocellular
181–192 (1997). protein-5 binds to axin and regulates the canonical carcinomas, and growth suppression in cancer cells
10. Ikeda, S. et al. Axin, a negative regulator of the Wnt Wnt signaling pathway. Mol. Cell 7, 801–809 (2001). by virus-​mediated transfer of AXIN1. Nat. Genet. 24,
signaling pathway, forms a complex with GSK-3β 30. Cliffe, A., Hamada, F. & Bienz, M. A role of Dishevelled 245–250 (2000).
and β-​catenin and promotes GSK-3β-​dependent in relocating Axin to the plasma membrane during 49. Nanki, K. et al. Divergent routes toward Wnt and
phosphorylation of β-​catenin. EMBO J. 17, wingless signaling. Curr. Biol. 13, 960–966 (2003). R-spondin niche independency during human gastric
1371–1384 (1998). 31. Tamai, K. et al. A mechanism for Wnt coreceptor carcinogenesis. Cell 174, 856–869.e17 (2018).
11. Behrens, J. et al. Functional interaction of an axin activation. Mol. Cell 13, 149–156 (2004). This study reveals that combined CDH1 and TP53
homolog, conductin, with β-​catenin, APC, and GSK3β. 32. Zeng, X. et al. A dual-​kinase mechanism for Wnt mutations in gastric cancer generate a state of
Science 280, 596–599 (1998). co-​receptor phosphorylation and activation. Nature RSPO-​independent tumour growth. These tumours
12. Lustig, B. et al. Negative feedback loop of Wnt 438, 873–877 (2005). are sensitive to PORCN inhibitor treatment in vivo,
signaling through upregulation of conductin/Axin2 33. Chen, S. et al. Structural and functional studies of indicating the existence of alternative mutational
in colorectal and liver tumors. Mol. Cell. Biol. 22, LRP6 ectodomain reveal a platform for Wnt signaling. routes that lead to WNT-​dependent growth states.
1184–1193 (2002). Dev. Cell 21, 848–861 (2011). 50. Seshagiri, S. et al. Recurrent R-​spondin fusions in
13. Hao, H. X. et al. ZNRF3 promotes Wnt receptor 34. Janda, C. Y., Waghray, D., Levin, A. M., Thomas, C. & colon cancer. Nature 488, 660–664 (2012).
turnover in an R-​spondin-sensitive manner. Nature Garcia, K. C. Structural basis of Wnt recognition by 51. Han, T. et al. R-​Spondin chromosome rearrangements
485, 195–202 (2012). Frizzled. Science 336, 59–64 (2012). drive Wnt-​dependent tumour initiation and
14. Koo, B. K. et al. Tumour suppressor RNF43 is a 35. Tauriello, D. V. F. et al. Wnt/β-​catenin signaling maintenance in the intestine. Nat. Commun. 8,
stem-​cell E3 ligase that induces endocytosis of requires interaction of the Dishevelled DEP domain 1–12 (2017).
Wnt receptors. Nature 488, 665–669 (2012). and C terminus with a discontinuous motif in Frizzled. 52. Giannakis, M. et al. RNF43 is frequently mutated in
Together with Hao et al. (2012), this study is the Proc. Natl Acad. Sci. USA 109, E812–E820 (2012). colorectal and endometrial cancers. Nat. Genet. 46,
first to uncover negative feedback roles of the WNT 36. Behrens, J. et al. Functional interaction of β-​catenin 1264–1266 (2014).
target genes RNF43 and ZNRF3. The discovery that with the transcription factor LEF-1. Nature 382, 53. Yang, L.-H. et al. Abnormal hypermethylation and
both proteins mediate WNT receptor endocytosis 638–642 (1996). clinicopathological significance of Axin gene in lung
and that RNF43 is mutated in cancer has founded 37. Molenaar, M. et al. XTcf-3 transcription factor mediates cancer. Tumor Biol. 34, 749–757 (2013).
the concept that WNT hypersensitivity drives tumour β-​catenin-induced axis formation in Xenopus embryos. 54. Lannagan, T. R. M. et al. Genetic editing of colonic
growth. In addition, this study demonstrates that Cell 86, 391–399 (1996). organoids provides a molecularly distinct and
RNF43- and ZNRF3-mutant cancers are sensitive 38. Jho, E.-h et al. Wnt/β-​catenin/Tcf signaling induces orthotopic preclinical model of serrated carcinogenesis.
to PORCN inhibitor treatment. the transcription of Axin2, a negative regulator of Gut 68, 684–692 (2019).
15. Polakis, P. Wnt signaling in cancer. Cold Spring Harb. the signaling pathway. Mol. Cell Biol. 22, 1172–1183 55. Nishisho, I. et al. Mutations of chromosome 5q21
Perspect. Biol. 4, 9 (2012). (2002). genes in FAP and colorectal cancer patients. Science
16. Reya, T. & Clevers, H. Wnt signalling in stem cells and 39. Spit, M. et al. RNF43 truncations trap CK1 to drive 253, 665–669 (1991).
cancer. Nature 434, 843–850 (2005). niche‐independent self‐renewal in cancer. EMBO J. 56. Groden, J. et al. Identification and characterization of
17. Zhong, Z. & Virshup, D. M. Wnt signaling and drug 39, e103932 (2020). the familial adenomatous polyposis coli gene. Cell 66,
resistance in cancer. Mol. Pharmacol. 97, 72–89 This study uncovers a class of RNF43 truncating 589–600 (1991).
(2020). cancer mutations that drive oncogenic β-catenin- 57. Joslyn, G. et al. Identification of deletion mutations
18. Martínez-​Jiménez, F. et al. A compendium of mutational mediated transcription despite exhibiting WNT and three new genes at the familial polyposis locus.
cancer driver genes. Nat. Rev. Cancer 20, 555–572 receptor downregulation. Mechanistically, these Cell 66, 601–613 (1991).
(2020). RNF43 mutants interfere with β-catenin turnover by 58. Rowan, A. J. et al. APC mutations in sporadic
19. Nusse, R. & Clevers, H. Wnt/β-​catenin signaling, selectively trapping AXIN1 and CK1 at the plasma colorectal tumors: a mutational ‘hotspot’ and
disease, and emerging therapeutic modalities. Cell membrane. Contrary to RNF43 LOF mutations, interdependence of the ‘two hits’. Proc. Natl Acad.
169, 985–999 (2017). these RNF43 mutations confer resistance to Sci. USA 97, 3352–3357 (2000).

18 | January 2021 | volume 21 www.nature.com/nrc


Reviews

59. Segditsas, S. & Tomlinson, I. Colorectal cancer and 86. Park, J. Y. et al. Mutations of β-​catenin and AXIN I 104. Lewis, A. et al. Severe polyposis in Apc1322T mice
genetic alterations in the Wnt pathway. Oncogene genes are a late event in human hepatocellular is associated with submaximal Wnt signalling and
25, 7531–7537 (2006). carcinogenesis. Liver Int. 25, 70–76 (2005). increased expression of the stem cell marker Lgr5.
60. Fearon, E. F. & Vogelstein, B. A genetic model for 87. Zhu, M. et al. Somatic mutations increase hepatic Gut 59, 1680–1686 (2010).
colorectal tumorigenesis. Cell 61, 759–767 (1989). clonal fitness and regeneration in chronic liver disease. 105. Fodde, R. et al. A targeted chain-​termination mutation
61. Nicholson, A. M. et al. Fixation and spread of somatic Cell 177, 608–621.e12 (2019). in the mouse Apc gene results in multiple intestinal
mutations in adult human colonic epithelium. Cell Stem 88. Marquardt, J. U. et al. Sequential transcriptome tumors. Proc. Natl Acad. Sci. USA 91, 8969–8973
Cell 22, 909–918.e8 (2018). analysis of human liver cancer indicates late stage (1994).
62. Kinzler, K. W. & Vogelstein, B. Lessons from hereditary acquisition of malignant traits. J. Hepatol. 60, 106. Gaspar, C. et al. A targeted constitutive mutation in
colorectal cancer. Cell 87, 159–170 (1996). 346–353 (2014). the Apc tumor suppressor gene underlies mammary
63. Polakis, P., Porfiri, E., Albert, I., Robbins, P. & 89. Rebouissou, S. & Nault, J. C. Advances in molecular but not intestinal tumorigenesis. PLoS Genet. 5,
Rubinfeld, B. The adenomatous polyposis coli (APC) classification and precision oncology in hepatocellular e1000547 (2009).
tumor suppressor, catenins and cancer. FASEB J. 12, carcinoma. J. Hepatol. 72, 215–229 (2020). This study describes a mouse model in which
A1302 (1998). This review describes the molecular subtypes of APC truncations distal to the CID (position 1572T)
64. Muzny, D. M. et al. Comprehensive molecular HCC and their link to clinical prognosis. Based on underlie mammary, but not intestinal, tumorigenesis.
characterization of human colon and rectal cancer. transcriptional profiles and clinical presentation, Furthermore, these hypomorphic APC mutants
Nature 487, 330–337 (2012). AXIN1-mutant and CTNNB1-mutant cases are drive lower levels of WNT signalling than shorter
65. Yaeger, R. et al. Clinical sequencing defines the shown to belong to distinct HCC subclasses, APC forms, supporting the hypothesis that specific
genomic landscape of metastatic colorectal cancer. emphasizing differences in the underlying dosages of WNT signalling initiate tumorigenesis in
Cancer Cell 33, 125–136.e3 (2018). pathogenic mechanisms. a tissue-​specific manner.
66. Guinney, J. et al. The consensus molecular subtypes of 90. Abitbol, S. et al. AXIN deficiency in human and mouse 107. Smits, R. et al. Apc1638T: a mouse model delineating
colorectal cancer. Nat. Med. 21, 1350–1356 (2015). hepatocytes induces hepatocellular carcinoma in the critical domains of the adenomatous polyposis coli
This study classifies CRCs into four CMSs based on absence of β-​catenin activation. J. Hepatol. 68, protein involved in tumorigenesis and development.
gene expression data. 1203–1213 (2018). Genes Dev. 13, 1309–1321 (1999).
67. Yan, H. H. N. et al. RNF43 germline and somatic This study reveals that AXIN1 mutations in HCC 108. Li, V. S. W. et al. Wnt signaling through inhibition of
mutation in serrated neoplasia pathway and its do not induce a conventional β-​catenin-mediated β-​catenin degradation in an intact Axin1 complex. Cell
association with BRAF mutation. Gut 66, transcriptional signature but instead are linked 149, 1245–1256 (2012).
1645–1656 (2017). with the activation of Notch and YAP signalling 109. Voloshanenko, O. et al. Wnt secretion is required to
68. Dienstmann, R. et al. Consensus molecular subtypes pathways. maintain high levels of Wnt activity in colon cancer
and the evolution of precision medicine in colorectal 91. Hanahan, D. & Weinberg, R. A. The hallmarks of cells. Nat. Commun. 4, 1–13 (2013).
cancer. Nat. Rev. Cancer 17, 79–92 (2017). cancer. Cell 100, 57–70 (2000). 110. Pronobis, M. I., Rusan, N. M. & Peifer, M. A novel
69. Bond, C. E. et al. RNF43 and ZNRF3 are commonly 92. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: GSK3-regulated APC:Axin interaction regulates
altered in serrated pathway colorectal tumorigenesis. the next generation. Cell 144, 646–674 (2011). Wnt signaling by driving a catalytic cycle of efficient
Oncotarget 7, 70589–70600 (2016). 93. Lamlum, H. et al. The type of somatic mutation at β-​catenin destruction. eLife 4, 1–31 (2015).
70. Sekine, S. et al. Frequent PTPRK–RSPO3 fusions and APC in familial adenomatous polyposis is determined 111. Choi, S. H., Estarás, C., Moresco, J. J., Yates, J. R. &
RNF43 mutations in colorectal traditional serrated by the site of the germline mutation: a new facet to Jones, K. A. α-​Catenin interacts with APC to regulate
adenoma. J. Pathol. 239, 133–138 (2016). Knudson’s ‘two-​hit’ hypothesis. Nat. Med. 5, β-​catenin proteolysis and transcriptional repression
71. McCarthy, A. J., Serra, S. & Chetty, R. Traditional 1071–1075 (1999). of Wnt target genes. Genes Dev. 27, 2473–2488
serrated adenoma: an overview of pathology and 94. Albuquerque, C. The ‘just-​right’ signaling model: APC (2013).
emphasis on molecular pathogenesis. BMJ Open somatic mutations are selected based on a specific 112. Su, Y. Y. et al. APC is essential for targeting
Gastroenterol. 6, e000317 (2019). level of activation of the β-​catenin signaling cascade. phosphorylated β-​catenin to the SCFβ-​TrCP
72. Panarelli, N. C., Vaughn, C. P., Samowitz, W. S. & Hum. Mol. Genet. 11, 1549–1560 (2002). ubiquitin ligase. Mol. Cell 32, 652–661 (2008).
Yantiss, R. K. Sporadic microsatellite instability-​high This study puts forward the ‘just-​right’ signalling 113. Toki, H. et al. Novel mouse model for gardner
colon cancers rarely display immunohistochemical model, in which biallelic APC mutations are selected syndrome generated by a large-​scale N-​ethyl-
evidence of Wnt signaling activation. Am. J. Surg. to retain some β-​catenin downregulating activity to N-​nitrosourea mutagenesis program. Cancer Sci.
Pathol. 39, 313–317 (2015). achieve a specific degree of β-​catenin signalling 104, 937–944 (2013).
73. Li, Y. et al. Frequent RNF43 mutation contributes that is optimal for tumour formation within the 114. Cai, J., Maitra, A., Anders, R. A., Taketo, M. M. &
to moderate activation of Wnt signaling in colorectal colon. Pan, D. β-​Catenin destruction complex-​independent
signet-​ring cell carcinoma. Protein Cell 11, 292–298 95. Kohler, E. M., Chandra, S. H. V., Behrens, J. & regulation of Hippo–YAP signaling by APC in intestinal
(2020). Schneikert, J. β-​Catenin degradation mediated by the tumorigenesis. Genes Dev. 29, 1493–1506 (2015).
74. Lai, C. et al. RNF43 frameshift mutations contribute CID domain of APC provides a model for the selection 115. Kaplan, K. B. et al. A role for the adenomatous
to tumourigenesis in right-​sided colon cancer. Pathol. of APC mutations in colorectal, desmoid and duodenal polyposis coli protein in chromosome segregation.
Res. Pract. 215, 152453 (2019). tumours. Hum. Mol. Genet. 18, 213–226 (2009). Nat. Cell Biol. 3, 429–432 (2001).
75. Salem, M. E. et al. Molecular analyses of left‐ and 96. Gaspar, C. & Fodde, R. APC dosage effects in 116. Nakagawa, H. et al. Identification of a brain-​specific
right‐sided tumors in adolescents and young adults tumorigenesis and stem cell differentiation. APC homologue, APCL, and its interaction with
with colorectal cancer. Oncologist 25, 404–413 Int. J. Dev. Biol. 48, 377–386 (2004). β-​catenin. Cancer Res. 58, 5176–5181 (1998).
(2020). 97. Roberts, D. M. et al. Deconstructing the β-​catenin 117. Van, Es, J. H. et al. Identification of APC2, a homologue
76. Christie, M. et al. Different APC genotypes in proximal destruction complex: mechanistic roles for the tumor of the adenomatous polyposis coli tumour suppressor.
and distal sporadic colorectal cancers suggest distinct suppressor APC in regulating Wnt signaling. Mol. Biol. Curr. Biol. 9, 105–108 (1999).
WNT/β-​catenin signalling thresholds for tumourigenesis. Cell 22, 1845–1863 (2011). 118. Schneikert, J. et al. Functional comparison of human
Oncogene 32, 4675–4682 (2013). This study discovers that the APC 20R2 region adenomatous polyposis coli (APC) and APC-​like in
77. Leedham, S. J. et al. A basal gradient of Wnt and and CID are essential for a late step in destruction targeting β-​catenin for degradation. PLoS ONE 8,
stem-​cell number influences regional tumour complex-​mediated β-​catenin turnover. e68072 (2013).
distribution in human and mouse intestinal tracts. 98. Novellasdemunt, L. et al. USP7 is a tumor-​specific 119. Mokarram, P. et al. Distinct high-​profile methylated
Gut 62, 83–93 (2013). WNT activator for APC-​mutated colorectal cancer by genes in colorectal cancer. PLoS ONE 4, e7012
78. Iacopetta, B. Are there two sides to colorectal cancer? mediating β-​catenin deubiquitination. Cell Rep. 21, (2009).
Int. J. Cancer 101, 403–408 (2002). 612–627 (2017). 120. He, Y. et al. Hypermethylation of APC2 is a predictive
79. LaPointe, L. C. et al. Map of differential transcript This study employs CRISPR–Cas9-mediated epigenetic biomarker for Chinese colorectal cancer.
expression in the normal human large intestine. gene editing to systematically investigate the Dis. Markers 2018, 8619462 (2018).
Physiol. Genomics 33, 50–64 (2008). consequences of incremental APC truncations 121. Daly, C. S. et al. Functional redundancy between Apc
80. Franko, J. et al. Prognosis of patients with peritoneal in mouse intestinal organoids. The results reveal and Apc2 regulates tissue homeostasis and prevents
metastatic colorectal cancer given systemic therapy: that loss of the CID in APC is essential for driving tumorigenesis in murine mammary epithelium.
an analysis of individual patient data from prospective WNT-​independent activation of β-​catenin. Oncogene 36, 1793–1803 (2017).
randomised trials from the Analysis and Research in 99. Miyoshi, Y. et al. Somatic mutations of the APC gene 122. Hamada, F. et al. Negative regulation of wingless
Cancers of the Digestive System (ARCAD) database. in colorectal tumors: mutation cluster region in the signaling by d-​Axin, a Drosophila homolog of Axin.
Lancet Oncol. 17, 1709–1719 (2016). APC gene. Hum. Mol. Genet. 1, 229–233 (1992). Science 283, 1739–1742 (1999).
81. Hugen, N. & Nagtegaal, I. D. Distinct metastatic 100. Schatoff, E. M. et al. Distinct colorectal cancer-​ 123. Korswagen, H. C. et al. The Axin-​like protein PRY-1
patterns in colorectal cancer patients based on associated APC mutations dictate response to is a negative regulator of a canonical Wnt pathway
primary tumour location. Eur. J. Cancer 75, 3–4 tankyrase inhibition. Cancer Discov. 9, 1358–1371 in C. elegans. Genes Dev. 16, 1291–1302 (2002).
(2017). (2019). 124. Chia, I. V. & Costantini, F. Mouse Axin and Axin2/
82. Forner, A., Reig, M. & Bruix, J. Hepatocellular 101. Moser, A. R., Pitot, H. C. & Dove, W. F. A dominant Conductin proteins are functionally equivalent
carcinoma. Lancet 391, 1301–1314 (2018). mutation that predisposes to multiple intestinal in vivo. Mol. Cell Biol. 25, 4371–4376 (2005).
83. Boyault, S. et al. Transcriptome classification of HCC neoplasia in the mouse. Science 247, 322–324 125. Feng, G. J. et al. Conditional disruption of Axin1
is related to gene alterations and to new therapeutic (1990). leads to development of liver tumors in mice.
targets. Hepatology 45, 42–52 (2007). 102. Moser, A. R. et al. ApcMin, a mutation in the murine Gastroenterology 143, 1650–1659 (2012).
84. Zucman-​Rossi, J. et al. Differential effects of inactivated APC gene, predisposes to mammary carcinomas and 126. Dong, B. et al. Activating CAR and β-​catenin induces
Axin1 and activated β-​catenin mutations in human focal alveolar hyperplasias. Proc. Natl Acad. Sci. USA uncontrolled liver growth and tumorigenesis.
hepatocellular carcinomas. Oncogene 26, 774–780 90, 8977–8981 (1993). Nat. Commun. 6, 5944 (2015).
(2007). 103. Pollard, P. et al. The Apc1322T mouse develops severe 127. Ringel, T. et al. Genome-​scale CRISPR screening
85. Zhang, S. et al. The polyploid state plays a tumor-​ polyposis associated with submaximal nuclear in human intestinal organoids identifies drivers of
suppressive role in the liver. Dev. Cell 44, 447–459. β-​catenin expression. Gastroenterology 136, TGF-β resistance. Cell Stem Cell 26, 431–440.e8
e5 (2018). 2204–2213.e13 (2009). (2020).

naTure RevIeWs | CANceR volume 21 | January 2021 | 19


Reviews

128. Picco, G. et al. Loss of AXIN1 drives acquired 148. Kleeman, S. O. et al. Exploiting differential Wnt target 171. Cha, P. H. et al. Small-​molecule binding of the axin
resistance to WNT pathway blockade in colorectal gene expression to generate a molecular biomarker RGS domain promotes β-​catenin and Ras degradation.
cancer cells carrying RSPO3 fusions. EMBO Mol. Med. for colorectal cancer stratification. Gut 69, Nat. Chem. Biol. 12, 593–600 (2016).
9, 293–303 (2017). 1092–1103 (2020). 172. Lee, S.-K. et al. β-​Catenin–RAS interaction serves
This study reveals that AXIN1 loss is a mutational 149. Tsukiyama, T. et al. Molecular role of RNF43 as a molecular switch for RAS degradation via GSK3β.
route for resistance to PORCN inhibitor treatment. in canonical and noncanonical Wnt signaling. EMBO Rep. 19, e406060 (2018).
129. Shen, J., Yu, Z. & Li, N. The E3 ubiquitin ligase Mol. Cell Biol. 35, 2007–2023 (2015). 173. Chen, B. et al. Small molecule–mediated disruption
RNF146 promotes colorectal cancer by activating 150. Lindeboom, R. G. H., Supek, F. & Lehner, B. of Wnt-​dependent signaling in tissue regeneration
the Wnt/β-​catenin pathway via ubiquitination of Axin1. The rules and impact of nonsense-​mediated mRNA and cancer. Nat. Chem. Biol. 5, 100–107 (2009).
Biochem. Biophys. Res. Commun. 503, 991–997 decay in human cancers. Nat. Genet. 48, 1112–1118 174. Koo, B. K., Van Es, J. H., Van Den Born, M. &
(2018). (2016). Clevers, H. Porcupine inhibitor suppresses paracrine
130. Gao, Y. et al. Overexpression of RNF146 in non-​small 151. Ivanov, I., Lo, K. C., Hawthorn, L., Cowell, J. K. & Wnt-​driven growth of Rnf43;Znrf3-mutant neoplasia.
cell lung cancer enhances proliferation and invasion of Ionov, Y. Identifying candidate colon cancer tumor Proc. Natl Acad. Sci. USA 112, 7548–7550 (2015).
tumors through the wnt/β-​catenin signaling pathway. suppressor genes using inhibition of nonsense-​ 175. US National Library of Medicine. ClinicalTrials.gov
PLoS ONE 9, e85377 (2014). mediated mRNA decay in colon cancer cells. https://clinicaltrials.gov/ct2/show/study/NCT01351103
131. Liu, D., Li, L., Yang, Y., Liu, W. & Wu, J. The Axin2 Oncogene 26, 2873–2884 (2007). (2011).
rs2240308 polymorphism and susceptibility to 152. Wang, K. et al. Whole-​genome sequencing and 176. US National Library of Medicine. ClinicalTrials.gov
lung cancer in a Chinese population. Tumor Biol. comprehensive molecular profiling identify new https://clinicaltrials.gov/ct2/show/NCT02278133
35, 10987–10991 (2014). driver mutations in gastric cancer. Nat. Genet. 46, (2014).
132. Bahl, C., Sharma, S., Singh, N. & Behera, D. Association 573–582 (2014). 177. US National Library of Medicine. ClinicalTrials.gov
study between genetic variations in Axin2 gene and 153. Tu, J. et al. The most common RNF43 mutant https://clinicaltrials.gov/ct2/show/NCT02521844
lung cancer risk in North Indian population: a multiple G659Vfs*41 is fully functional in inhibiting Wnt (2015).
interaction analysis. Tumor Biol. https://doi.org/ signaling and unlikely to play a role in tumorigenesis. 178. US National Library of Medicine. ClinicalTrials.gov
10.1177/1010428317695533 (2017). Sci. Rep. 9, 18557 (2019). https://clinicaltrials.gov/ct2/show/NCT03447470
133. Marvin, M. L. et al. AXIN2-associated autosomal 154. Li, S. et al. Commonly observed RNF43 mutations (2018).
dominant ectodermal dysplasia and neoplastic retain functionality in attenuating Wnt/β-​catenin 179. US National Library of Medicine. ClinicalTrials.gov
syndrome. Am. J. Med. Genet. A 155, 898–902 signaling and unlikely confer Wnt-​dependency https://clinicaltrials.gov/ct2/show/NCT02675946
(2011). onto colorectal cancers. Oncogene 39, 3458–3472 (2016).
134. Aristizabal-​Pachon, A. F., Carvalho, T. I., Carrara, H. H., (2020). 180. US National Library of Medicine. ClinicalTrials.gov
Andrade, J. & Takahashi, C. S. AXIN2 polymorphisms, 155. Van De Wetering, M. et al. Prospective derivation of https://clinicaltrials.gov/ct2/show/NCT03507998
the β-​catenin destruction complex expression profile a living organoid biobank of colorectal cancer patients. (2018).
and breast cancer susceptibility. Asian Pac. J. Cancer Cell 161, 933–945 (2015). 181. Janku, F. et al. Abstract C45: phase I study
Prev. 16, 7277–7284 (2015). 156. Schwank, G. et al. Functional repair of CFTR by of WNT974, a first-​in-class Porcupine inhibitor,
135. Lammi, L. et al. Mutations in AXIN2 cause familial CRISPR/Cas9 in intestinal stem cell organoids of in advanced solid tumors. Mol. Cancer Ther. 14,
tooth agenesis and predispose to colorectal cancer. cystic fibrosis patients. Cell Stem Cell 13, 653–658 C45–C45 (2015).
Am. J. Hum. Genet. 74, 1043–1050 (2004). (2013). 182. Ng, M. et al. First-​in-human phase 1 study of
136. Rivera, B. et al. A novel AXIN2 germline variant 157. Drost, J. et al. Sequential cancer mutations in cultured ETC-159 an oral PORCN inhibitor in patients with
associated with attenuated FAP without signs of human intestinal stem cells. Nature 521, 43–47 advanced solid tumours. J. Clin. Oncol. 35, 2584
oligondontia or ectodermal dysplasia. Eur. J. Hum. (2015). (2017).
Genet. 22, 423–426 (2014). 158. Liu, Xling, Ding, J. & Meng, Lhua Oncogene-​induced 183. Madan, B. et al. Bone loss from Wnt inhibition
137. Rosales-​Reynoso, M. A. et al. AXIN2 polymorphisms senescence: a double edged sword in cancer. Acta mitigated by concurrent alendronate therapy.
and their association with colorectal cancer in Mexican Pharmacol. Sin. 39, 1553–1558 (2018). Bone Res. 6, 17 (2018).
patients. Genet. Test. Mol. Biomarkers 20, 438–444 159. Dow, L. E. et al. Apc restoration promotes cellular 184. Zhong, Z. et al. PORCN inhibition synergizes with
(2016). differentiation and reestablishes crypt homeostasis PI3K/mTOR inhibition in Wnt-​addicted cancers.
138. Mazzoni, S. M., Petty, E. M., Stoffel, E. M. & in colorectal cancer. Cell 161, 1539–1552 (2015). Oncogene 38, 6662–6677 (2019).
Fearon, E. R. An AXIN2 mutant allele associated This study reveals that restoration of APC expression 185. Steinhart, Z. et al. Genome-​wide CRISPR screens
with predisposition to colorectal neoplasia has in advanced invasive intestinal carcinoma harbouring reveal a Wnt–FZD5 signaling circuit as a druggable
context-​dependent effects on AXIN2 protein function. TP53 and KRAS mutations drives tumour cell vulnerability of RNF43-mutant pancreatic tumors.
Neoplasia 17, 463–472 (2015). differentiation and sustained regression. This work Nat. Med. 23, 60–68 (2017).
139. Thorvaldsen, T. E., Pedersen, N. M., Wenzel, E. M. discovers that WNT pathway alterations remain 186. Fenderico, N. et al. Anti-​LRP5/6 VHHs promote
& Stenmark, H. Differential roles of AXIN1 and important for tumour maintenance at late stages differentiation of Wnt-​hypersensitive intestinal
AXIN2 in tankyrase inhibitor-​induced formation of of cancer development and provides compelling stem cells. Nat. Commun. 10, 365 (2019).
degradasomes and β-​catenin degradation. PLoS ONE evidence that CRC cells can revert to normal 187. Smith, D. C. et al. First-​in-human evaluation of the
12, 1–12 (2017). functioning cells by therapeutic targeting of human monoclonal antibody vantictumab (OMP-
140. Otero, L. et al. Variations in AXIN2 predict risk and WNT signalling. 18R5; anti-​Frizzled) targeting the WNT pathway in a
prognosis of colorectal cancer. BDJ Open 5, 13 (2019). 160. Jung, Y.-S. & Park, J.-I. Wnt signaling in cancer: phase I study for patients with advanced solid tumors.
141. Anvarian, Z. et al. Axin cancer mutants form therapeutic targeting of Wnt signaling beyond J. Clin. Oncol. 31, 2540 (2013).
nanoaggregates to rewire the Wnt signaling β-​catenin and the destruction complex. Exp. Mol. 188. Flanagan, D. J. et al. Frizzled-7 is required for
network. Nat. Struct. Mol. Biol. 23, 324–332 Med. 52, 183–191 (2020). wnt signaling in gastric tumors with and without
(2016). 161. Huang, S. M. A. et al. Tankyrase inhibition stabilizes Apc mutations. Cancer Res. 79, 970–981 (2019).
This study reveals that missense mutations in the axin and antagonizes Wnt signalling. Nature 461, 189. Noutsou, M. et al. Critical scaffolding regions of
AXIN1 RGS domain drive basal β-​catenin-mediated 614–620 (2009). the tumor suppressor Axin1 are natively unfolded.
transcription and tumour growth in vivo by a gain-​ 162. Mariotti, L. et al. Tankyrase requires SAM domain-​ J. Mol. Biol. 405, 773–786 (2011).
of-function mechanism. By mediating structural dependent polymerization to support Wnt–β-​catenin 190. Minde, D. P., Radli, M., Forneris, F., Maurice, M. M.
destabilization of the RGS domain, these mutations signaling. Mol. Cell 63, 498–513 (2016). & Rüdiger, S. G. D. Large extent of disorder in
promote the formation of small-​scale aggregates 163. Lau, T. et al. A novel tankyrase small-​molecule adenomatous polyposis coli offers a strategy to
that rewire the AXIN1 interactome. Prevention of inhibitor suppresses APC mutation-​driven colorectal guard Wnt signalling against point mutations.
aggregation rescues AXIN1 tumour suppressor tumor growth. Cancer Res. 73, 3132–3144 (2013). PLoS ONE 8, 1–9 (2013).
activity, underlining the role of aggregation for 164. Tanaka, N. et al. APC mutations as a potential 191. Merenda, A., Fenderico, N. & Maurice, M. M. Wnt
cancer-​promoting activity. biomarker for sensitivity to tankyrase inhibitors in signaling in 3D: recent advances in the applications
142. Schwarz-​Romond, T. et al. The DIX domain of colorectal cancer. Mol. Cancer Ther. 16, 752–762 of intestinal organoids. Trends Cell Biol. 30, 60–73
Dishevelled confers Wnt signaling by dynamic (2017). (2020).
polymerization. Nat. Struct. Mol. Biol. 14, 165. Croy, H. E. et al. The poly(ADP-​ribose) polymerase 192. Schell, M. J. et al. A multigene mutation classification
484–492 (2007). enzyme Tankyrase antagonizes activity of the β-​catenin of 468 colorectal cancers reveals a prognostic role for
143. Schaefer, K. N. et al. Supramolecular assembly of the destruction complex through ADP-​ribosylation of APC. Nat. Commun. 7, 11743 (2016).
β-​catenin destruction complex and the effect of Wnt Axin and APC2. J. Biol. Chem. 291, 12747–12760 193. Hashimoto, T. et al. Acquisition of WNT pathway
signaling on its localization, molecular size, and (2016). gene alterations coincides with the transition from
activity in vivo. PLoS Genet. 14, e1007339 (2018). 166. Menon, M. et al. A novel tankyrase inhibitor, precursor polyps to traditional serrated adenomas.
144. Eto, T. et al. Impact of loss-​of-function mutations at MSC2504877, enhances the effects of clinical Am. J. Surg. Pathol. 43, 132–139 (2019).
the RNF43 locus on colorectal cancer development CDK4/6 inhibitors. Sci. Rep. 9, 1–16 (2019). 194. Taupin, D. et al. A deleterious RNF43 germline
and progression. J. Pathol. 245, 445–455 (2018). 167. Kahn, M. Can we safely target the WNT pathway? mutation in a severely affected serrated polyposis
145. Jiang, X. et al. Inactivating mutations of RNF43 Nat. Rev. Drug Discov. 13, 513–532 (2014). kindred. Hum. Genome Var. 2, 15013 (2015).
confer Wnt dependency in pancreatic ductal 168. An, T. et al. USP7 inhibitor P5091 inhibits Wnt 195. Quintana, I. et al. Evidence suggests that germline
adenocarcinoma. Proc. Natl Acad. Sci. USA 110, signaling and colorectal tumor growth. Biochem. RNF43 mutations are a rare cause of serrated
12649–12654 (2013). Pharmacol. 131, 29–39 (2017). polyposis. Gut 67, 2230–2232 (2018).
146. Neumeyer, V. et al. Loss of endogenous RNF43 169. Ma, P. et al. The ubiquitin ligase RNF220 enhances 196. Cerami, E. et al. The cBio cancer genomics portal:
function enhances proliferation and tumour growth canonical Wnt signaling through USP7-mediated an open platform for exploring multidimensional
of intestinal and gastric cells. Carcinogenesis 40, deubiquitination of β-​catenin. Mol. Cell Biol. 34, cancer genomics data. Cancer Discov. 2, 401–404
551–559 (2018). 4355–4366 (2014). (2012).
147. Basham, K. J. et al. A ZNRF3-dependent Wnt/ 170. Ji, L. et al. USP7 inhibits Wnt/β-​catenin signaling 197. Weinstein, J. N. et al. The Cancer Genome Atlas
β-​catenin signaling gradient is required for adrenal through promoting stabilization of Axin. Nat. Commun. Pan-​Cancer analysis project. Nat. Genet. 45,
homeostasis. Genes Dev. 33, 209–220 (2019). 10, 4184 (2019). 1113–1120 (2013).

20 | January 2021 | volume 21 www.nature.com/nrc


Reviews

198. Jiang, X., Charlat, O., Zamponi, R., Yang, Y. & Cong, F. binding of APC to β-​catenin and its role in β-​catenin Organi­zation for Scientific Research NWO VICI Grant
Dishevelled promotes Wnt receptor degradation degradation. Mol. Cell 15, 511–521 (2004). 91815604 and ZonMW TOP Grant 91218050 (to M.M.M.).
through recruitment of ZNRF3/RNF43 E3 ubiquitin 205. Spink, K. E., Polakis, P. & Weis, W. I. Structural basis
ligases. Mol. Cell 58, 522–533 (2015). of the Axin–adenomatous polyposis coli interaction. Author contributions
199. Kishida, S. et al. Axin, a negative regulator of the EMBO J. 19, 2270–2279 (2000). J.M.B., N.F. and M.M.M all researched data for the article,
Wnt signaling pathway, directly interacts with 206. Joslyn, G., Richardson, D. S., White, R. & Alber, T. provided substantial contributions to discussions of the con-
adenomatous polyposis coli and regulates the Dimer formation by an N-​terminal coiled coil in tent and contributed equally to writing the article and to the
stabilization of β-​catenin. J. Biol. Chem. 273, the APC protein. Proc. Natl Acad. Sci. USA 90, review and/or editing of the manuscript before submission.
10823–10826 (1998). 11109–11113 (1993).
200. Spink, K. E., Fridman, S. G. & Weis, W. I. Molecular 207. Li, Z., Kroboth, K., Newton, I. P. & Näthke, I. S. Competing interests
mechanisms of β-​catenin recognition by adenomatous Novel self-​association of the APC molecule affects The authors declare no competing interests.
polyposis coli revealed by the structure of an APC– APC clusters and cell migration. J. Cell Sci. 121,
β-​catenin complex. EMBO J. 20, 6203–6212 1916–1925 (2008). Peer review information
(2001). 208. Schaefer, K. N. & Peifer, M. Wnt/β-​catenin signaling Nature Reviews Cancer thanks B. O. Williams, M. Waterman
201. Nakamura, T. et al. Axin, an inhibitor of the Wnt regulation and a role for biomolecular condensates. and the other, anonymous, reviewer(s) for their contribution
signalling pathway, interacts with β-​catenin, GSK-3β Dev. Cell 48, 429–444 (2019). to the peer review of this work.
and APC and reduces the β-​catenin level. Genes Cells
3, 395–403 (1998). Acknowledgements Publisher’s note
202. Schwarz-​Romond, T., Metcalfe, C. & Bienz, M. The authors thank members of the Maurice laboratory for Springer Nature remains neutral with regard to jurisdictional
Dynamic recruitment of axin by Dishevelled protein helpful discussions and suggestions, specifically I. Jordens claims in published maps and institutional affiliations.
assemblies. J. Cell Sci. 120, 2402–2412 (2007). and A. Venhuizen for critically reading the manuscript and
203. Faux, M. C. et al. Recruitment of adenomatous A. Cristobal Gonzales de Durana for providing AXIN1 Related links
polyposis coli and β-​catenin to axin-​puncta. immuno­fluorescence images. This work is part of the Oncode cBioPortal: https://www.cbioportal.org/
Oncogene 27, 5808–5820 (2008). Institute, which is partly financed by the Dutch Cancer
204. Ha, N. C., Tonozuka, T., Stamos, J. L., Choi, H. J. & Society. This work was supported by European Research
Weis, W. I. Mechanism of phosphorylation-​dependent Council Starting Grant 242958, the Netherlands © Springer Nature Limited 2020

naTure RevIeWs | CANceR volume 21 | January 2021 | 21

You might also like