Canonical Wnt suppressor, Axin2, promotes coloncarcinoma oncogenic activityZhao-Qiu Wua, Thomas Brabletzb, Eric Fearona, Amanda L. Willisa, Casey Yuexian Hua, Xiao-Yan Lia,and Stephen J. Weissa,1

aDivision of Molecular Medicine and Genetics, Department of Internal Medicine, and the Life Sciences Institute, University of Michigan, Ann Arbor, MI 48109;and bDepartment of Visceral Surgery and Comprehensive Cancer Center Freiburg, University of Freiburg, 79106 Freiburg, Germany

Edited by Dennis A. Carson, University of California at San Diego, La Jolla, CA, and approved May 30, 2012 (received for review February 22, 2012)

Aberrant activation of canonical Wingless-type MMTV integrationsite family (Wnt) signaling is pathognomonic of colorectal cancers(CRC) harboring functional mutations in either adenomatouspolyposis coli or β-catenin. Coincident with Wnt cascade activa-tion, CRCs also up-regulate the expression of Wnt pathway feed-back inhibitors, particularly the putative tumor suppressor, Axin2.Because Axin2 serves as a negative regulator of canonical Wntsignaling in normal cells, recent attention has focused on the util-ity of increasing Axin2 levels in CRCs as a means to slow tumorprogression. However, rather than functioning as a tumor sup-pressor, we demonstrate that Axin2 acts as a potent promoterof carcinoma behavior by up-regulating the activity of the tran-scriptional repressor, Snail1, inducing a functional epithelial-mesenchymal transition (EMT) program and driving metastatic ac-tivity. Silencing Axin2 expression decreases Snail1 activity,reverses EMT, and inhibits CRC invasive and metastatic activitiesin concert with global effects on the Wnt-regulated cancer celltranscriptome. The further identification of Axin2 and nuclearSnail1 proteins at the invasive front of human CRCs supports a re-vised model wherein Axin2 acts as a potent tumor promoterin vivo.

invasion | E-cadherin | basement membrane | GSK3β | tankyrase

Colorectal cancers (CRCs) are generally characterized by ab-errant Wingless-type MMTV integration site family (Wnt)

signaling (1–3). Neoplastic cells located at the invasive front ofCRCs frequently display increased levels of nuclear β-cateninand adopt a mesenchymal-like phenotype—events precipitatedby mutations predominantly occurring in adenomatous polypo-sis coli (APC) or, more rarely, β-catenin (1–3). After nucleartranslocation, β-catenin complexes with DNA-binding proteinsof the T-cell factor/lymphoid enhancer factor (TCF/LEF) familyand triggers the expression of downstream genes linked to theinduction of an invasion program that allows CRC cells totransmigrate the underlying basement membrane, gain access tothe host interstitial matrix, and infiltrate surrounding tissues (1–4). However, in addition to supporting CRC invasion, the cohortof β-catenin/TCF-regulated target genes also includes feedbackinhibitors that serve to modulate the activity of the canonicalWnt signaling cascade (1, 2).One of the most extensively studied canonical Wnt pathway

feedback inhibitors in CRCs is the scaffold protein, Axin2/con-ductin (1, 2). Under physiologic conditions, free β-catenin istargeted for degradation by a multiprotein complex composed ofAPC, GSK3β, CK1, and either Axin1 or Axin2 (1, 2). Althoughlow levels of Axin2 are found in the normal colon epithelium,constitutive activation of Axin2 expression by the β-catenin/TCFtranscriptional complex is observed in CRCs (1–3). Axin2 effi-ciently supports the GSK3β-dependent phosphorylation ofβ-catenin in normal cells, which subsequently marks the post-translationally modified protein for β-TrCP–dependent ubiq-uitination and proteosomal degradation (1, 2). However, theability of elevated levels of Axin2 to modulate β-catenin levels inAPC- or β-catenin–mutant CRC cells is muted (1, 2). Indeed, inan effort to silence the aberrant transcriptional programs acti-vated in CRCs as a function of the constitutive stabilization of

β-catenin, new therapeutics have been promoted that increaseAxin2 (as well as Axin1) levels by targeting the poly-ADP ribo-sylating enzymes, tankyrase 1 and tankyrase 2 (5–7).Although Axin2 is uniformly categorized as a tumor suppres-

sor in CRCs (1, 2, 5–7), its precise function in carcinomatousstates is unknown, particularly with regard to the molecular“rationale” that underlies the decision of neoplastic cells tomaintain high levels of a putative suppressor in vitro and in vivo.Herein, we demonstrate that endogenous Axin2 promotes,rather than suppresses, a β-catenin/TCF-initiated, EMT programin both β-catenin–mutant and APC-mutant CRCs. Up-regulationof Axin2 levels in CRCs triggers marked increases in Snail1 ac-tivity and induces EMT, whereas silencing endogenous Axin2expression initiates a mesenchymal-epithelium–like switch thatnot only down-regulates Snail1, but also triggers widespreadchanges in the canonical Wnt signaling transcriptome. BecauseAxin2 and Snail1 are further shown to be colocalized at the in-vasive front of CRC tissues, these studies suggest that Axin2serves as an effective tumor promoter—rather than tumor sup-pressor—that not only controls the induction of a Snail1-de-pendent EMT program, but also exerts global control over geneexpression networks critical to invasive and metastatic behaviors.

ResultsCanonical Wnt Signaling Triggers a Colon Cancer Invasion Program. InHCT116 or SW48 CRC cells that harbor heterozygous, stabi-lizing mutations in β-catenin, each of the lines constitutivelyexpresses β-catenin and Axin2 and TOPflash reporter activity(Fig. 1 A and B). In the presence of exogenous Wnt3a, however,the cells respond with a significant increase in β-catenin proteinlevels, Axin2 mRNA levels, and TOPflash reporter activity (Fig.1 A and B). By contrast, in APC-defective SW620 cells, consti-tutive levels of β-catenin, Axin2, or TOPflash activity are maxi-mally activated and unaffected by Wnt3a (Fig. 1 A and B). Bothβ-catenin and APC mutant cell lines trigger canonical Wnt sig-naling via a β-catenin/TCF-dependent pathway because Axin2mRNA levels are significantly suppressed when either HCT116or SW620 cells are stably transfected with a dominant-negative,FLAG-tagged TCF-4 construct (TCF-DN) (Fig. 1C) (8). Fur-ther, although both HCT116 and SW620 populations are dom-inated by E-cadherin–negative or E-cadherin–low-expressingcells and assume mesenchymal cell-like phenotypes in vitro, theTCF-DN stable transfectants up-regulate E-cadherin proteinlevels and E-cadherin promoter activity while adopting an epi-thelial cell-like appearance (Fig. 1 C and D).The most definitive characteristic of cancer cell EMT pro-

grams is the ability of the transformed cell to traverse an intact,

Author contributions: Z.-Q.W., E.F., and S.J.W. designed research; Z.-Q.W., T.B., A.L.W.,C.Y.H., X.-Y.L., and S.J.W. performed research; T.B. contributed new reagents/analytictools; Z.-Q.W., T.B., E.F., X.-Y.L., and S.J.W. analyzed data; and Z.-Q.W. and S.J.W. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: sjweiss@umich.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1203015109/-/DCSupplemental.

11312–11317 | PNAS | July 10, 2012 | vol. 109 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1203015109

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020

type IV collagen-rich BM (4, 9). As such, HCT116 or SW620cells were cultured atop the chorioallantoic membrane (CAM)of live, 11-d-old chicken embryos wherein the upper epithelialcell layer is subtended by an intact BM (9). Under these con-ditions, both CRC cell lines rapidly degrade the underlying BMand invade the subjacent interstitial tissues (Fig. S1). By contrast,TCF-DN-transfectants of each cell line lose invasive potentialand remain confined to the upper CAM surface (Fig. S1).

β-Catenin/TCF Signaling Induces Snail-Regulated EMT and TumorInvasion in Colon Cancer Cells. During development and carcino-genesis, activation of canonical Wnt signaling has been linked toexpression of the zinc-finger transcriptional repressor, Snail1 (3,8, 10). To determine whether the constitutive Wnt signalingactivity associated with CRC cells is sufficient to trigger Snail1protein expression, HCT116, SW48, or SW620 cells were cul-tured in the absence or presence of exogenous Wnt3a. AlthoughWnt3a did not alter Snail1 mRNA expression levels in any of thecell lines tested, Snail1 protein levels increased significantly inWnt3a-treated HCT116 or SW48 cells, whereas APC-mutantSW620, SW480, or DLD1 cells express Snail1 protein in a con-stitutive fashion (Fig. 2A and Fig. S1). Induction of Snail1 pro-tein expression is linked directly to the canonical Wnt pathwaybecause HCT116 or SW620 cells that stably express a TCF-DNconstruct decrease Snail1 protein levels in the absence or pres-ence of Wnt3a without affecting Snail1 mRNA expression (Fig.2B and Fig. S2). Although the expression of Snail1 correlateswith a Wnt-dependent EMT program, multiple transcriptionfactors have been reported to trigger similar EMT-like responsesin CRCs, particularly Snail2/Slug or ZEB1 (11–13). However,down-regulating β-catenin/TCF activity with the TCF-DN con-struct did not affect mRNA expression levels of either of thesetranscription factors (Fig. S2). Further, when Snail1 expressionis silenced in either HCT116 or SW620 cells by either of twoshRNA constructs, E-cadherin is reexpressed, whereas TOP-Flash reporter activity is down-regulated in a manner consistentwith the ability of Snail1 to modulate β-catenin/TCF activity (Fig.S2) (14). Finally, the EMT-associated invasion programs ex-hibited by SW620 or HCT116 cells are suppressed by more than80% after Snail1 knockdown in vivo (Fig. S2). Hence, the con-stitutive Wnt signaling activity that distinguishes the majority ofall CRCs triggers a Snail1-dependent EMT program that ismarked by both E-cadherin repression and BM invasion.

Axin2-Dependent Augmentation of the CRC EMT Program. Theconstitutive activation of the β-catenin/TCF pathway in CRC

tissues is known to trigger robust Axin2 expression in vivo ina presumed effort to down-regulate Wnt signaling (1, 2). Un-expectedly, when Axin2 is overexpressed in HCT116 or SW620cells not only is TOPflash activity maintained (Fig. S3), but in-vasive activity is enhanced (see below). Because these resultsraise the possibility that endogenous Axin2 promotes an EMT-like program in CRC cells, Axin2 levels were stably repressed by∼80% in HCT116 or SW620 cells with either of two specificshRNA constructs (without affecting Axin1 levels) (Fig. 2C).Under these conditions, both cell lines reexpress E-cadherin,increase E-cadherin promoter activity by more than twofold, andassume an epithelial phenotype (Fig. 2 C and D and Fig. S3). Intandem with the increased expression of E-cadherin and theexpected decrease in the pool of free β-catenin, TOPflash re-porter activity is also decreased by 25% (Fig. S3).Recent studies indicate that canonical Wnt activation controls

Snail1 protein levels by regulating GSK3β activity, wherein phos-phorylation of an N-terminal, serine-rich motif in Snail1 triggersits β-TRCP–dependent ubiquitination and proteosomal degra-dation (8, 10). Consistent with an operative GSK3β-Snail1 axis inCRC cells, HCT116 or SW620 cells cultured in the presenceof the GSK3 inhibitor, CHIR 99021 (15), increase Snail1 protein(but not mRNA) levels (Fig. 2E). In tandem with increased levelsof Snail1 protein, levels of p-GSK3β(S9), an inactive form of thekinase (15–17), decrease slightly, whereas protein levels of p-GSK3β(Y216), the catalytically active form of the kinase, fall tobackground (Fig. 2E). Immunohistochemical analysis confirmsthat the CHIR 99021-induced increase in Snail1 protein ex-pression is confined to the nuclear compartment where anexpected decrease in the kinase activity of GSK3β is observed(Fig. S3). Likewise, Snail1 protein accumulates to higher levelswhen CRC cells are transfected with an epitope-tagged Snail1expression vector harboring a stabilizing S→A(96) point muta-tion in its GSK3β phosphorylation motif relative to a wild-typeSnail1 construct (Fig. 2F) (8, 10). Given that Axin family mem-bers can potentially impact GSK3 intracellular localization andsubstrate phosphorylation (10, 18, 19), the ability of Axin2 tomodulate Snail1 protein levels was assessed. Importantly, whenAxin2-silenced colon cancer cells are transfected with the wild-type epitope-tagged Snail1 construct, Snail1 protein levels reacha barely detectable level relative to that observed in controlCRC cells (Fig. 2F). By contrast, when Axin2-silenced cellsare transfected with the stabilized S→A(96) mutant, Snail1 ismaintained at levels comparable to those detected in the controlpopulation (Fig. 2F). Conversely, when endogenous Axin2 levelsare increased by culturing CRCs with the tankyrase inhibitor,

Fig. 1. Canonical Wnt signaling pathway triggersa CRC cell invasion program. (A) CRC cells weretreated with vehicle or Wnt3a (100 ng/mL) for 24 h.β-catenin protein and Axin2 mRNA levels wereassessed by Western blotting (Upper) and qRT-PCR(Lower), respectively. (B) Cells cotransfected withTOPflash reporter and pRL-TK constructs were trea-ted with Wnt3a, and the cell lysates were analyzedfor luciferase reporter activity (mean ± SEM; n = 3).(C) Cells infected with mock or TCF-DN–expressingretrovirus were selected with G418, recovered,and reinfected with mock or an IRES-GFP-Axin2–expressing lentivirus. GFP-positive cells were sortedby FACS, cultured, and subjected to Western blot-ting (Upper) and qRT-PCR (Lower) analyses, respec-tively. (D Left) Mock or TCF-DN stably expressing CRCcells were fixed and stained with anti–E-cadherinantibody. Nuclei were stained with TOTO3 (blue).(Scale bar: 40 μm.) (D Right) CRC were cotransfectedwith an E-cadherin luciferase reporter and pRL-TKconstructs, and luciferase reporter activity was de-termined (mean ± SEM; n = 3). **P < 0.01.

Wu et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11313

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020

XAV939 (5, 6), Snail1 levels increase while E-cadherin expres-sion falls (Fig. S3). Taken together, these results indicate thatincreasing Axin2 levels—either via APC/β-catenin mutations orpharmacologic intervention—promotes CRC EMT. Indeed,consistent with the ability of Axin2 to stabilize Snail1, nuclearβ-catenin, Axin2, and Snail1 are colocalized at the invasive frontof colon carcinoma tissues (Fig. S4).

Axin2-Dependent Control of Nuclear GSK3β Activity. Axin2 has beenreported to stabilize nuclear Snail1 protein levels in Wnt-stim-ulated breast carcinoma cells by inducing the export of GSK3βfrom the nuclear compartment (10). However, silencing Axin2 inSW620 or HCT116 CRCs did not alter the subcellular distribu-tion of GSK3β (Fig. 3A and Fig. S5), indicating the existence ofalternate schemes for regulating the Axin2/Snail1 axis. In Axin2-silenced CRCs, no changes are detected in the levels of p-GSK3β(S9), but nuclear levels of the active, Y216 phosphorylated formof GSK3β increase significantly in parallel with enhanced nuclearGSK3β kinase activity and decreased nuclear Snail1 levels (Fig.3B and Fig. S5) (15–17). Axin2 silencing did not affect cytosoliclevels of p-GSK3β(Y216), cytosolic GSK3β kinase activity, orβ-catenin content (Fig. 3A and Fig. S5). Conversely, when Axin2levels are increased in transfected SW620 or HCT116 CRC cells,nuclear Snail1 protein levels increase in tandem with a decreasein nuclear p-GSK3β(Y216) content and nuclear GSK3β kinaseactivity (Fig. 3 C and D and Fig. S5). Furthermore, consistent

with the ability of Axin2 to increase Snail1 levels, exogenousAxin2 reverses the induction of E-cadherin observed in HCT116or SW620 cells that are transduced with TCF-DN (Fig. 1C).Finally, confirming the important role of Y216 in regulatingGSK3β-mediated Snail1 phosphorylation, whereas Snail1 is ac-tively phosphorylated in vitro by recombinant wild-type GSK3β(but not a catalytically inactive K85A mutant), the GSK3βY216F mutant displays an ∼90% loss in Snail1 phosphorylatingactivity (Fig. 3E). Hence, Axin2 serves as a negative regulator ofnuclear GSK3β activity that allows for the stabilization of nuclearSnail1 levels and activity.

Axin2 and the Regulation of the CRC Transcriptome and InvasionPrograms. To assess the impact of Axin2 expression on globalCRC function, mRNA was isolated from control and Axin2-si-lenced SW620 cells for expression profiling. Using a minimum oftwofold change in both of the two Axin2-silenced SW620 cellpopulations as a cutoff, depressing endogenous Axin2 levelsfour- to fivefold alters the expression of almost 3,000 uniquetranscripts (1,291 transcripts increased and 1,506 transcriptsdecreased; Dataset S1). Gene ontology (GO) analysis furtherhighlights major changes in cell cycle regulation, cell death, andmetabolic processes (Fig. S6 and Dataset S2). Consistent withthe proposition that at least a subset of the affected genes reflecttheir identity as Snail1 and/or EMT-associated target transcripts,Axin2 silencing decreases expression of mesenchymal cell

Fig. 2. Axin2-dependent EMT program in CRCcells. (A) CRC cells were treated with vehicle orWnt3a (100 ng/mL) for 24 h. Snail1 protein andmRNA levels were assessed by Western blotting(Upper) and qRT-PCR (Lower), respectively. (B)Mock or TCF-DN stably expressing cells were in-cubated in the presence or absence of Wnt3a, andSnail1 protein level was determined by Westernblot analysis. (C) Axin2, E-cadherin, and vimentinlevels were determined in cell lysates recoveredfrom CRC cells stably expressing Scr-shRNA, Axin2-shRNA-4, or Axin2-shRNA-5. (D Left) Stable trans-fectants of CRC cells were fixed and stained withanti–E-cadherin antibody. Nuclei were stained withTOTO3 (blue). (Scale bar: 40 μm.) (D Right) Cellswere stained with anti–E-cadherin and subjected toFACS analysis. (E) CRC cells were treated withCHIR99021 (100 ng/mL) for 24 h, and protein andmRNA levels were assessed by Western blottingand qRT-PCR, respectively. (F) CRC cells stablyexpressing Scr-shRNA or Axin2-shRNA were trans-fected with a mock, Flag-Snail1-WT, or Flag-Snail1-S96A construct, and Snail1 and β-actin levels in celllysates were determined by Western blotting.*P < 0.05, **P < 0.01.

11314 | www.pnas.org/cgi/doi/10.1073/pnas.1203015109 Wu et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020

markers, vimentin, and fibronectin, in tandem with significantreductions in stem cell-associated markers (e.g., BMI1, CD44,and CTGF), tumor-promoting matrix metalloproteinases (e.g.,MT1-MMP and MMP-7), and Snail1 regulatory factors (e.g.,LOXL2) (all targets confirmed by real-time PCR; Fig. 4A andFig. S6) (18–20). In turn, Axin2 silencing led to up-regulatedexpression of multiple epithelial-associated transcripts, includingE-cadherin, H-cadherin, and claudin 3 (Fig. 4A). Interestingly,a series of the affected transcripts have been identified as targetsof the β-catenin/TCF transcriptional network, including L1CAM,IL8, CYR61, and MMP7 (Fig. 4B and Fig. S6) (www.stanford.edu/group/nusselab/cgi-bin/wnt). Indeed, although Axin2 silenc-ing would be predicted to increase the expression of β-catenin/TCF targets (i.e., by increasing the steady-state concentration ofβ-catenin/TCF complexes), the majority of canonical Wnt sig-naling targets identified in the Axin2 dataset—many of whichrepresent Snail1 and/or EMT targets (e.g., FZD7, MET, MT1-

MMP, IL8, PLAUR, CD44s; refs. 9 and 21–25)—are all down-regulated after Axin2 silencing in both SW620 and SW480 cells(Fig. 4A and Fig. S6), suggesting that the ability of Axin2 to driveSnail1 activity and trigger EMT-like programs overrides itsability to act as a feedback inhibitor of β-catenin/TCF signalingin CRC cells.Given the apparent mesenchymal-epithelial shift in CRC be-

havior after Axin2 knockdown, the ability of Axin2-silencedcancer cells to breach the CAMbasementmembrane was assessedin vivo. As shown, Axin2-silenced SW620 cells, HCT116, DLD1,or SW480 are no longer able to invade the CAM interstitium asassessed by H&E staining, CRC immunolocalization with anti-human cytokeratin antibodies or immunofluorescence (Fig. 5 Aand B and Figs. S1 and S7). By contrast, Axin2 overexpressionaccelerates the invasive properties of HCT116 or SW620 cells,most readily observed when the incubation time in the chickenembryo system is shortened from 4 to 2 d (Fig. 5C and Fig. S7).

Fig. 3. Axin2-dependent control of nuclear GSK3β activity in CRC cells. (A) CRC cells stably expressing Scr-shRNA or Axin2-shRNA were fractioned into cy-toplasmic and nuclear pools and subjected to Western blot analysis to determine Axin2, Snail1, GSK3β, p-GSK3β (Tyr216 and Ser9), or β-catenin levels. HDAC1and β-tubulin are used as markers for nuclear and cytoplasmic fractions, respectively. (B) Cytoplasmic and nuclear fractions from the above cell lines weresubjected to anti-GSK3β antibody immunoprecipitation, followed by GSK3β in vitro kinase assay with phosphoglycogen synthase peptide as a substrate. Insetdisplays the equal quantities of immunoprecipitated GSK3β in the nuclear fractions (mean ± SEM; n = 3). (C and D) CRC cells were transfected with Flag orFlag-Axin2 vectors, and cytoplasmic or nuclear fractions were subjected to Western blot analysis (C) or to anti-GSK3β antibody immunoprecipitation, followedby GSK3β in vitro kinase assay (D). (E) HA-tagged GSK3β protein was immunoprecipitated from HCT116 cells that were transfected with HA-GSK3β-WT, -K85Aor -Y216F constructs, and the immunoprecipitated complex were subjected to GSK3β in vitro kinase assay. **P < 0.01.

Fig. 4. Axin2 is a master regulator of Wnt/β-catenin/TCF signaling gene expression programs. (A) GO terms identifying cellular processes that are differentiallyexpressed in polyclonal populations of SW620 cells that stably express scramble- or Axin2-shRNAs. GO terms generated from up-regulated and down-regulatedgenes are colored red and green, respectively (P ≤ 0.05; Dataset S2). (B) Quantitative RT-PCR analysis for the indicated transcripts. (Mean ± SEM; n = 3; **P < 0.01.)(A) Heat map of predicted Wnt/β-catenin/TCF pathway targets relative to Axin2-silenced SW620 cells using either of two distinct shRNA constructs.

Wu et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11315

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020

The ability of Axin2-silenced CRCs to express invasive activitycan, however, be reconstituted, but only when Snail1 proteinlevels are reestablished by expressing the stabilized Snail1-S96Amutant (Fig. 5D and Fig. S7). Because Snail1 has been shown toplay a critical role in driving metastatic events in mouse modelsin vivo (26, 27), the ability of Axin2-silenced SW620 cells tosupport metastatic activity after tail vein injection in nude micewas assessed. As predicted, Axin2-suppressed CRCs displaya profound inhibition of macrometastatic and micrometastaticactivity over an 8-wk time course (Fig. S8), reinforcing theprooncogenic activity of sustained Axin2 expression.

DiscussionVirtually all CRCs are characterized by an inappropriate activa-tion of the canonical Wnt pathway that triggers the induction ofthe β-catenin/TCF target, Axin2 (1–3). In an effort to squelchβ-catenin/TCF activity in carcinomatous states, recent efforts havefocused on developing pharmacologic interventions that increaseAxin1 or Axin2 expression in CRC cells to levels capable ofsuppressing Wnt signaling (1–3). Although the possibility thatCRCs might preserve Axin2 expression for more ominous pur-poses has seldom been considered (28), we demonstrate that CRCcells use the elevated endogenous levels of Axin2 to promote

a Snail1-dependent EMT program that directs tissue-invasivepotential (Fig. S9).During CRC progression, cancer cells situated at the invasion

front have been reported to display an EMT-like phenotype(3, 11). Indeed, invasive CRC cells frequently display the fin-gerprint of canonical Wnt signaling in tandem with the expres-sion of EMT-inducing transcription factors, including Snail1(11, 13, 29). Given the required role for β-catenin/TCF signalingin the Snail1-dependent EMT program activated in CRCs,coupled with the feedback inhibitor role normally assigned toAxin2, overexpressing Axin2 would be predicted to suppresscanonical Wnt signaling pathways and decrease tissue-invasiveactivity. However, Axin2 overexpression not only failed to inhibitβ-catenin/TCF reporter activity, but also stimulated invasion.More importantly, silencing endogenous Axin2 expression re-versed the CRC EMT program, while also blocking BM trans-migration and metastatic activity in vivo. Unexpectedly, Axin2exerted far more complex effects on the CRC transcriptome thananticipated. The elevated levels of Axin2 commonly found inCRC cells increased—rather than decreased—the expression ofmultiple β-catenin/TCF targets in tandem with the silencing ofa large cohort of tumor suppressor-associated transcripts, in-cluding CYLD, MeOX2, and CADM1 (Dataset S1) (30–32).These findings complement a recent study describing an un-expected, positive regulatory function for Axin1 and Axin2 inWnt signaling (33). Nevertheless, it is unlikely that all Axin2-induced changes can be attributed to complementary changes inSnail1 activity or EMT, e.g., silencing L1CAM (a β-catenin/TCFtarget) has been reported to decrease CRC metastasis withoutinducing EMT (34). Nevertheless, our findings demonstrate thatSnail1 remains a key Axin2 target as silencing Snail1 expressionabrogated invasive and metastatic activities.The functional impact of Axin2 on Snail1 activity was linked

directly to its ability to regulate nuclear GSK3β activity. Becausethe transcriptional repressor activity of Snail1 localizes to thenuclear compartment (35), we considered the possibility thatAxin2 might alter GSK3β nuclear localization in a process sim-ilar to that described for Wnt-stimulated breast cancer cells (10).Instead, nuclear Axin2 directly inhibited nuclear GSK3β activitywithout affecting its cellular distribution. Currently, the mecha-nisms by which Axin2 alters GSK3β activity and phosphorylationremain the subject of conjecture, but recent studies have linkedchanges in GSK3β Y216 phosphorylation and/or activity to avariety of factors, including PDGF-DD signaling or p38 MAPKactivity (17, 36). Independent of changes in nuclear GSK3β ac-tivity, Axin2–GSK3β binding interactions can potentially alterthe ability of kinase to recognize and phosphorylate target sub-strates (37). Whereas GSK3β catalyzes β-catenin phosphoryla-tion when both components are incorporated into an APC/Axindocking complex, alternate targets, such as Snail1, are phos-phorylated directly by GSK3β (10, 37). In this regard, when theability of GSK3β to phosphorylate Snail1 was assessed undercell-free conditions in either the absence or presence ofrecombinant Axin2, GSK3β-dependent phosphorylation of Snail1is inhibited as the Axin2:GSK3β ratio is increased (Fig. S10).Hence, Axin2 not only decreases nuclear GSK3β kinase activitydirectly, but also competitively inhibits the active kinase fromtargeting Snail1. Preliminary efforts designed to map Axin2regions critical to Snail1 regulation failed to identify sites requiredfor controlling β-catenin/TCF reporter activity or Snail1 proteinlevels (Fig. S10), but exogenously introduced mutants would havethe opportunity to form heterodimers with the wild-type Axin2monomers expressed by the host cell. Further studies in Axin2-silenced CRCs will be required to map the essential domains re-sponsible for regulating GSK3β activity (38).Consistent with our in vitro findings, and complementary

reports documenting increased levels of nuclear β-catenin,Axin2, or Snail1 in CRCs in vivo (1–3, 29), these critical playerswere identified at the invasive front of a limited set of patientsamples. Caution should, however, be exercised in attempting todevelop simple correlations between nuclear Axin2 or Snail1levels and disease status. First, the effects of nuclear Axin2 on

Fig. 5. Axin2-dependent CRC cell invasion. (A and B) SW620 cells stablyexpressing Scr-shRNA, Axin2-shRNA-4, or Axin2-shRNA-5 were cultured onthe chick CAM for 4 d. CAMs were sectioned and assessed for invasion byH&E staining (Left), by immunohistochemical analysis with anti–cytokeratin-18 antibody (Center), and by fluorescence microscopy (Right; labeled cells,green; type IV collagen, red; cell nuclei, blue). Invasion was assessed as de-scribed (mean ± SEM; n = 3). (Scale bar: 400 μm.) Relative invasive activity isquantified in B. (C) HCT116 and SW620 cells transfected with Flag or Flag-Axin2 construct were cultured on the chick CAM for 2 d, and relative invasiveactivity of the cells is quantified. (D) SW620 cells stably expressing Axin2-shRNA were transfected with mock or Flag-Snail-S96A vectors and cultured onthe chick CAM for 4 d and relative invasive activity quantified. **, ##, P < 0.01.

11316 | www.pnas.org/cgi/doi/10.1073/pnas.1203015109 Wu et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020

Snail1 activity are confined to posttranslational events impactingGSK3β regulation of Snail1 protein activity rather than Snail1transcript levels. Consequently, Axin2-dependent EMT will onlybe realized under conditions known to up-regulate Snail1 tran-scription (e.g., FGF, TNF or TGFβ) or translation (39). Second,while the presence of nuclear Snail1 protein is potentially in-dicative of its EMT-inducing potential, it should be stressed thatits transcriptional repressor activity is further regulated bycomplex phosphorylation events mediated by as yet uncharac-terized kinases (35). Likewise, with the realization that Snail1silences target genes by forming complexes with a multiplicity ofrepressors, including HDAC1/2, Sin3a, LSD1, AJUBA/PRMT5,and EZH2 (39), the terminal impact of Snail1 on cell functionwill also be modulated by the coexpression of these and otheraccessory factors. Finally, when fully engaged, the ability of theβ-catenin/TCF–Axin2–Snail1 axis to engage an EMT-like pro-gram accompanied by tissue-invasive activity may not correlatewith effects on other tumorigenic events such as proliferation ormetastatic potential (40). For example, in CRC, tissue-invasivecells migrating away from primary tumor sites have been shownto down-regulate proliferative activity (3, 11), although we donot detect changes in CRC proliferation in vitro after Axin2 si-lencing (Fig. S10). Similarly, cancer cell EMT programs canpromote local invasion and intravasation, but the adoption ofmesenchymal cell-like properties may not necessarily provepermissive for postextravasation growth at metastatic sites, al-though this contention also remains controversial (27, 40).Nevertheless, the ability of Snail1 to induce EMT, trigger BMinvasion, exert antiapoptotic effects, and imbue expressing cellswith stem cell-like properties is noteworthy (3, 25, 39).

The ability of Axin2 to promote a protumorigenic phenotypein CRC challenges current dogma, but we note that aberrantWnt/β-catenin signaling has also been reported to induce chro-mosomal instability in CRC via an Axin2-dependent process (28,38). Taken together, these findings suggest that the long-heldsupposition that therapeutic efforts aimed at increasing Axin2levels in CRC patients should prove beneficial in the clinicalsetting requires reevaluation as does the contention that tank-yrase inhibitors exert their effects by targeting Axin familymembers alone (5–7). Indeed, we alternatively conclude that thetumor suppressor function of Axin2 in normal cells is most likelycoopted by cancer cells to promote, rather than suppress, keyaspects of the cancer progression program.

Materials and MethodsDetailed protocols regarding cell culture, plasmid and shRNA construction,pharmacologic treatment of cells, cell extract preparation, antibodies, immu-noprecipitation, in vitro kinase assays, luciferase reporter assay, recombinantprotein purification, metastasis assays, transcriptional profiling, immuno-histochemistry, Chick CAM invasion assay, and statistical analysis are de-scribed in SI Materials and Methods. All cell lines were obtained from ATCC.Animal protocols were approved by the Unit for Laboratory Animal Medi-cine at the University of Michigan.

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth Grants CA116516 (to S.J.W. and E.F.) and R01 CA071699 (to S.J.W.),the Breast Cancer Research Foundation, and Michigan Diabetes Researchand Training Center Cell and Molecular Biology Core National Institutes ofHealth Grant P60 DK020572.

1. Fearon ER (2011) Molecular genetics of colorectal cancer. Annu Rev Pathol 6:479–507.2. Klaus A, Birchmeier W (2008) Wnt signalling and its impact on development and

cancer. Nat Rev Cancer 8:387–398.3. Fodde R, Brabletz T (2007) Wnt/beta-catenin signaling in cancer stemness and ma-

lignant behavior. Curr Opin Cell Biol 19:150–158.4. Rowe RG, Weiss SJ (2009) Navigating ECM barriers at the invasive front: The cancer

cell-stroma interface. Annu Rev Cell Dev Biol 25:567–595.5. Huang SM, et al. (2009) Tankyrase inhibition stabilizes axin and antagonizes Wnt

signalling. Nature 461:614–620.6. Chen B, et al. (2009) Small molecule-mediated disruption of Wnt-dependent signaling

in tissue regeneration and cancer. Nat Chem Biol 5:100–107.7. Waaler J, et al. (2012) A novel tankyrase inhibitor decreases canonical Wnt signaling

in colon carcinoma cells and reduces tumor growth in conditional APC mutant mice.Cancer Res 72:2822–2832.

8. Yook JI, Li XY, Ota I, Fearon ER, Weiss SJ (2005) Wnt-dependent regulation of theE-cadherin repressor Snail. J Biol Chem 280:11740–11748.

9. Ota I, Li XY, Hu Y, Weiss SJ (2009) Induction of a MT1-MMP andMT2-MMP-dependentbasement membrane transmigration program in cancer cells by Snail1. Proc Natl AcadSci USA 106:20318–20323.

10. Yook JI, et al. (2006) AWnt-Axin2-GSK3beta cascade regulates Snail1 activity in breastcancer cells. Nat Cell Biol 8:1398–1406.

11. Schmalhofer O, Brabletz S, Brabletz T (2009) E-cadherin, beta-catenin, and ZEB1 inmalignant progression of cancer. Cancer Metastasis Rev 28:151–166.

12. Larriba MJ, et al. (2009) Snail2 cooperates with Snail1 in the repression of vitamin Dreceptor in colon cancer. Carcinogenesis 30:1459–1468.

13. Sánchez-Tilló E, et al. (2010) ZEB1 represses E-cadherin and induces an EMT by re-cruiting the SWI/SNF chromatin-remodeling protein BRG1. Oncogene 29:3490–3500.

14. Stemmer V, de Craene B, Berx G, Behrens J (2008) Snail promotes Wnt target geneexpression and interacts with beta-catenin. Oncogene 27:5075–5080.

15. Cohen P, Goedert M (2004) GSK3 inhibitors: Development and therapeutic potential.Nat Rev Drug Discov 3:479–487.

16. Cole A, Frame S, Cohen P (2004) Further evidence that the tyrosine phosphorylationof glycogen synthase kinase-3 (GSK3) in mammalian cells is an autophosphorylationevent. Biochem J 377:249–255.

17. Kumar A, et al. (2010) Platelet-derived growth factor-DD targeting arrests patho-logical angiogenesis by modulating glycogen synthase kinase-3beta phosphorylation.J Biol Chem 285:15500–15510.

18. Hlubek F, Spaderna S, Jung A, Kirchner T, Brabletz T (2004) Beta-catenin activatesa coordinated expression of the proinvasive factors laminin-5 gamma2 chain andMT1-MMP in colorectal carcinomas. Int J Cancer 108:321–326.

19. Wilson CL, Heppner KJ, Labosky PA, Hogan BL, Matrisian LM (1997) Intestinal tu-morigenesis is suppressed in mice lacking the metalloproteinase matrilysin. Proc NatlAcad Sci USA 94:1402–1407.

20. Peinado H, et al. (2005) A molecular role for lysyl oxidase-like 2 enzyme in Snailregulation and tumor progression. EMBO J 24:3446–3458.

21. Lester RD, Jo M, Montel V, Takimoto S, Gonias SL (2007) uPAR induces epithelial-mesenchymal transition in hypoxic breast cancer cells. J Cell Biol 178:425–436.

22. Toiyama Y, et al. (2012) Co-expression of hepatocyte growth factor and c-Met pre-dicts peritoneal dissemination established by autocrine hepatocyte growth factor/c-Met signaling in gastric cancer. Int J Cancer 130:2912–2921.

23. Ueno K, et al. (2008) Frizzled-7 as a potential therapeutic target in colorectal cancer.Neoplasia 10:697–705.

24. Brown RL, et al. (2011) CD44 splice isoform switching in human and mouse epitheliumis essential for epithelial-mesenchymal transition and breast cancer progression. J ClinInvest 121:1064–1074.

25. Hwang WL, et al. (2011) SNAIL regulates interleukin-8 expression, stem cell-like ac-tivity, and tumorigenicity of human colorectal carcinoma cells. Gastroenterology 141:279–291, 291, e1–e5.

26. Olmeda D, et al. (2007) SNAI1 is required for tumor growth and lymph node me-tastasis of human breast carcinoma MDA-MB-231 cells. Cancer Res 67:11721–11731.

27. von Burstin J, et al. (2009) E-cadherin regulates metastasis of pancreatic cancer in vivoand is suppressed by a SNAIL/HDAC1/HDAC2 repressor complex. Gastroenterology137:361–371.

28. Hadjihannas MV, et al. (2006) Aberrant Wnt/beta-catenin signaling can inducechromosomal instability in colon cancer. Proc Natl Acad Sci USA 103:10747–10752.

29. Francí C, et al. (2009) Snail1 protein in the stroma as a new putative prognosis markerfor colon tumours. PLoS ONE 4:e5595.

30. Iliopoulos D, Jaeger SA, Hirsch HA, Bulyk ML, Struhl K (2010) STAT3 activation ofmiR-21 and miR-181b-1 via PTEN and CYLD are part of the epigenetic switch linkinginflammation to cancer. Mol Cell 39:493–506.

31. Raveh S, Gavert N, Spiegel I, Ben-Ze’ev A (2009) The cell adhesion nectin-like mole-cules (Necl) 1 and 4 suppress the growth and tumorigenic ability of colon cancer cells.J Cell Biochem 108:326–336.

32. Valcourt U, Thuault S, Pardali K, Heldin CH, Moustakas A (2007) Functional role ofMeox2 during the epithelial cytostatic response to TGF-beta. Mol Oncol 1:55–71.

33. Lui TT, et al. (2011) The ubiquitin-specific protease USP34 regulates axin stability andWnt/β-catenin signaling. Mol Cell Biol 31:2053–2065.

34. Gavert N, Vivanti A, Hazin J, Brabletz T, Ben-Ze’ev A (2011) L1-mediated colon cancercell metastasis does not require changes in EMT and cancer stem cell markers. MolCancer Res 9:14–24.

35. MacPherson MR, et al. (2010) Phosphorylation of serine 11 and serine 92 as newpositive regulators of human Snail1 function: Potential involvement of casein kinase-2 and the cAMP-activated kinase protein kinase A. Mol Biol Cell 21:244–253.

36. Thornton TM, et al. (2008) Phosphorylation by p38 MAPK as an alternative pathwayfor GSK3beta inactivation. Science 320:667–670.

37. Stoothoff WH, Bailey CD, Mi K, Lin SC, Johnson GV (2002) Axin negatively affects tauphosphorylation by glycogen synthase kinase 3beta. J Neurochem 83:904–913.

38. Hadjihannas MV, Brückner M, Behrens J (2010) Conductin/axin2 and Wnt signallingregulates centrosome cohesion. EMBO Rep 11:317–324.

39. de Herreros AG, Peiró S, Nassour M, Savagner P (2010) Snail family regulation andepithelial mesenchymal transitions in breast cancer progression. J Mammary GlandBiol Neoplasia 15:135–147.

40. Celià-Terrassa T, et al. (2012) Epithelial-mesenchymal transition can suppress majorattributes of human epithelial tumor-initiating cells. J Clin Invest 122:1849–1868.

Wu et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11317

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Janu

ary

6, 2

020