β-Catenin induces T-cell transformation by promotinggenomic instabilityMarei Dosea, Akinola Olumide Emmanuela, Julie Chaumeilb, Jiangwen Zhangc, Tianjiao Suna, Kristine Germara,Katayoun Aghajania, Elizabeth M. Davisd, Shilpa Keerthivasana, Andrea L. Bredemeyere, Barry P. Sleckmane,Steven T. Rosenf, Jane A. Skokb, Michelle M. Le Beaud, Katia Georgopoulosg, and Fotini Gounaria,1

aSection of Rheumatology and Knapp Center for Lupus and Immunology Research, University of Chicago, Chicago, IL 60637; bDepartment of Pathology, NewYork University School of Medicine, New York, NY 10016; cFaculty of Arts and Sciences (FAS) Center for Systems Biology, Harvard University, Cambridge,MA 02138; dSection of Hematology/Oncology and the Comprehensive Cancer Center, University of Chicago, Chicago, IL 60637; eDepartment of Pathologyand Immunology, Washington University School of Medicine, St. Louis, MO 63110; fRobert H. Lurie Comprehensive Cancer Center, Feinberg School ofMedicine, Northwestern University, Chicago, IL 60611; and gCutaneous Biology Research Center, Massachusetts General Hospital, Charlestown, MA 02129

Edited* by Harvey Cantor, Dana-Farber Cancer Institute, Boston, MA, and approved December 4, 2013 (received for review August 20, 2013)

Deregulated activation of β-catenin in cancer has been correlatedwith genomic instability. During thymocyte development, β-cate-nin activates transcription in partnership with T-cell–specific tran-scription factor 1 (Tcf-1). We previously reported that targetedactivation of β-catenin in thymocytes (CAT mice) induces lympho-mas that depend on recombination activating gene (RAG) andmyelocytomatosis oncogene (Myc) activities. Here we show thatthese lymphomas have recurring Tcra/Myc translocations thatresulted from illegitimate RAG recombination events and resem-bled oncogenic translocations previously described in human T-ALL. We therefore used the CAT animal model to obtain mecha-nistic insights into the transformation process. ChIP-seq analysisuncovered a link between Tcf-1 and RAG2 showing that the twoproteins shared binding sites marked by trimethylated histone-3lysine-4 (H3K4me3) throughout the genome, including near thetranslocation sites. Pretransformed CAT thymocytes had increasedDNA damage at the translocating loci and showed altered repairof RAG-induced DNA double strand breaks. These cells were ableto survive despite DNA damage because activated β-catenin pro-moted an antiapoptosis gene expression profile. Thus, activatedβ-catenin promotes genomic instability that leads to T-cell lympho-mas as a consequence of altered double strand break repair andincreased survival of thymocytes with damaged DNA.

beta-catenin/Tcf-1 | DNA recombination Tcf7 | Ctnnb1

Development of lymphocytes involves recombination of theirgenomic DNA to allow for expression of antigen receptor

genes. Thymocytes first rearrange the T-cell receptor (Tcr) β, γ,and δ loci at the CD4−CD8− double-negative-3 (DN3) stage ofdevelopment and then the Tcrα (Tcra) locus at the CD4+CD8+

double-positive (DP) stage. DNA double strand breaks (DSBs)generated during these processes are catalyzed by the recom-bination activating gene (RAG) recombinase complex. Thus,differentiating T cells sustain programmed RAG-mediated DNADSBs, in addition to random DNA damage that results fromtranscription initiation, DNA replication, and spatial reconfigu-ration of the chromatin architecture. An essential component ofthe RAG complex is the RAG2 protein, which binds H3K4me3and colocalizes with this histone mark throughout the genome (1–3). This widespread binding of RAG2 to DNA is puzzling, and it isthought to contribute to off-target generation of DSBs (i.e., DSBsoutside the immune receptor loci) (4). DNA ends generated bythe RAG complex recruit nonhomologous end joining (NHEJ)proteins, including Xrcc4, Ligase IV, DNA-PKcs, Artemis, andXLF/Cernunnos, that mediate rapid repair (5). The precise mech-anisms in place to maintain genome integrity in the face of thesebreaks remain under intense investigation (6).Failure to repair illegitimate DSBs and/or purge the cells that

have damaged DNA induces oncogenic translocations and is a se-vere impediment to successful therapeutic eradication of cancercells. Oncogenic events often target conserved developmental

pathways and mutations in the Wnt signaling pathway are acommon example (7). In normal development Wnt activity is short-lived and expression of Wnt target genes oscillates, suggestingthat Wnt activity is controlled by inherent negative feedbackloops (8). Uncontrolled activation of the central effector of Wntsignaling, β-catenin, has been causatively linked to genome instabilityin multiple cancers, including hematopoietic malignancies (9–12).However, how uncontrolled β-catenin promotes genomic instabilityin these malignancies is unknown.β-Catenin exerts Wnt-mediated transcription functions by

interacting with members of the TCF/LEF family of HMG do-main DNA binding proteins. The T-cell–specific TCF/LEF factorTcf-1 (product of the Tcf7 gene), one of the earliest transcriptionalregulators induced in thymus seeding T-cell progenitors, is essentialfor T-cell commitment (13, 14). Tcf-1 constitutively interacts withDNA and is thought to mediate activation of transcription whenbound by β-catenin and repression when bound by Groucho. Earlierstudies showed that Tcf-1 is recurrently required during T-cell de-velopment (15–17). Tcf-1 is most abundant in DP thymocytes, andits absence compromises DP thymocyte survival (17, 18).Here we use a mouse model of β-catenin–induced T-cell

malignancy to address how constitutive activation of β-catenin in-duces genomic instability. Previously, we reported that stabilization

Significance

Understanding molecular mechanisms that underlie genomicinstability will remove a major obstacle to effective treatmentof cancer. Here we characterize a unique animal model thatallows insight into mechanisms of genomic instability leadingto oncogenic translocations. We show that thymocyte-specificactivation of β-catenin induces genomically unstable lympho-mas with Tcra/Myc translocations, reminiscent of human leu-kemia. Tcf-1, the partner of β-catenin, colocalized throughoutthe genome with the RAG2 recombinase at DNA sites thoughtto be vulnerable to illegitimate recombination. Pretransformedthymocytes showed increased DNA damage at the trans-locating loci and altered DNA repair. These cells survived de-spite DNA damage. These surprising observations show thatactivated β-catenin promotes genomic instability and cancer bycompromising DNA repair and enhancing cell survival.

Author contributions: M.D. and F.G. designed research; M.D., A.O.E., J.C., T.S., K. Germar,K.A., E.M.D., S.K., and F.G. performed research; A.L.B., B.P.S., S.T.R., J.A.S., M.M.L.B., andK. Georgopoulos contributed new reagents/analytic tools; M.D., A.O.E., J.Z., M.M.L.B., andF.G. analyzed data; and M.D. and F.G. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: The sequences have been deposited in NCBI Gene Expression Omnibus(accession no. GSE46662).1To whom correspondence should be addressed. E-mail: fgounari@uchicago.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1315752111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1315752111 PNAS | January 7, 2014 | vol. 111 | no. 1 | 391–396

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1

of β-catenin in thymocytes through Cre-mediated deletion of itsproteolytic degradation signal induces RAG-dependent T-celllymphomas (19). This finding supported the notion that geno-mic instability may be the underlying cause of transformation.Our current findings confirmed this notion by showing that

lymphomas with activated β-catenin had recurring translo-cations that resembled translocations observed in human T-ALL.The translocations resulted from illegitimate RAG recom-bination events as they involved the Tcra locus. At the sametime, we linked Tcf-1 and RAG2 by showing that they boundtogether at chromatin sites marked by H3K4me3 genome-wide.Premalignant CAT DP thymocytes had increased DNA damageat the translocation sites and showed altered processing ofRAG-mediated DSBs. These cells survived even when theyhad damaged DNA because β-catenin activation induced anantiapoptosis expression profile. Thus, our findings show thatβ-catenin activation prompts genomic instability that is associatedwith the survival of cells with damaged DNA and with DNArepair errors.

ResultsT-Cell Lymphomas Induced by β-Catenin Activation Have RecurrentTranslocations. Constitutive activation of β-catenin as seen inhuman cancer can be modeled in mice by Cre-mediated deletionof its proteolytic degradation domain (20). We have previouslyreported that CD4Cre-mediated activation of β-catenin in thismanner (CAT mice) leads to DP T-cell lymphomas, targeting∼90% of CAT mice with a latency of 99 d (19). Transformationrequired RAG-mediated DNA damage, and RAG activity wasnot simply needed for progression to the DP stage (19). Basedon this finding we speculated that tumorigenesis resulted fromβ-catenin–induced genomic instability. Indeed, spectral karyo-type (SKY) analyses of eight independent lymphomas identifieda wide range of chromosomal translocations and amplifications(Table S1) indicating genomic instability. Six samples had a re-curring translocation of Tcra to the myelocytomatosis oncogene(Myc) locus, t (13, 14) (C2;D1), and one had a translocation ofTcrb to the Myc locus (Fig. 1A and Fig. S1A) suggesting error-prone repair of RAG-induced DSBs. Fluorescence in situ hy-bridization (FISH) showed that in the recurrent translocation thejoining (J), constant (C), and Eα enhancer regions of the Tcralocus were placed downstream of Myc into the Plasmacytomavariant translocation 1 (Pvt1) locus (Fig. 1 B and C). Chimeric Pvt1/Tcra transcripts were also detectable in these lymphomas (Fig.S1B). The translocation places the Myc locus under the control ofthe Tcra Eα enhancer, explaining the overexpression of Myc inCAT lymphomas (19). A similar translocation, targeting TCRA tothe MYC/PVT1 locus, marks ∼2% of human T-ALL (21). Se-quencing translocation breakpoints from such T-ALL samplesidentified cryptic RAG recombination signal sequences at thePVT1 side of the breakpoint and lead to suggestions that theyrepresented illegitimate RAG-mediated events (22). The Pvt1locus is also frequently targeted by proviral integrations in murineleukemia virus-induced T-cell lymphomas in mice (23) and in rats(24). It is also the site for breakpoints in ∼20% of human Burkitt’s

lymphomas (25) and in murine plasmacytomas (26). In conclu-sion, CAT thymocytes are a useful model to gain insights into thecontribution of β-catenin to genomic instability because theysustain frequent illegitimate RAG recombination events leadingto recurrent oncogenic translocations.

Widespread Overlap of Tcf-1 and RAG2 Binding at H3K4me3 Sites.RAG2 binds H3K4me3-modified histones throughout the ge-nome through its PHD domain, although the physiological sig-nificance of this widespread distribution is not clear (1–3). Toaddress whether β-catenin may influence RAG activity, wemapped the DNA binding pattern of its partner, Tcf-1, in thy-mocytes by ChIP-seq. This experiment revealed that mostH3K4me3 sites (69%) overlapped with Tcf-1 binding sites. Wethus asked if Tcf-1 and RAG2 also had similar genome-widedistribution patterns and compared our Tcf-1 ChIP-seq data tothe public data set for RAG2 (1–3). We observed overlappingTcf-1 at RAG2 sites throughout the genome (Fig. 2 A and B).Moreover, Tcf-1 and RAG2 peaks had a similar shape andsummit location (Fig. 2C), suggesting that they bind in closeproximity. Approximately 80% of all RAG2 binding sites wereshared with both Tcf-1 and H3K4me3 (“trimethylated Tcf-1sites” in the following) (Fig. 2D). Conversely, 84% of all trime-thylated Tcf-1 sites were bound by RAG2 (compared with only65% of all H3K4me3 sites). These data identify trimethylatedTcf-1 sites as bona fide RAG2 binding sites. They further suggestthat Tcf-1 is present at most DNA sites that are vulnerable to off-target RAG-mediated DSBs.We next examined whether the cobinding of Tcf-1 and RAG2

could predict expression levels of the target genes. To this endwe assessed the average expression of genes in WT, CAT, andlymphoma-derived DP cells, stratified by Tcf-1 and RAG2binding and by the presence or absence of H3K4me3 marks. Onaverage, cooccupied genes did not change expression betweenWT, CAT, or lymphoma DP cells. Genes bound by Tcf-1 showedhigher expression than the average gene expression of all genesin DP thymocytes. Cobinding of RAG2 and Tcf-1 marked geneswith significantly higher expression compared with genes onlybound by Tcf-1 (Fig. 2E). Thus, Tcf-1 and RAG2 bind activelytranscribing genes that have an increased probability to sustainDNA damage as a result of transcription initiation.

Trimethylated Tcf-1 Sites Are Preserved upon Activation of β-Catenin.The dependence of CAT lymphomas on RAG-mediated DNAdamage and the overlapping binding of Tcf-1 and RAG2 suggestthat Tcf-1 may have a role in DNA recombination and/or repairthat is influenced by β-catenin, leading to genomic instability. Wetherefore determined the effect of β-catenin activation on tri-methylated Tcf-1 binding sites. By extending our ChIP-seqanalyses to CAT thymocytes, we observed that Tcf-1 peak shapeand summit location at shared Tcf-1 sites were similar in CATand WT DP thymocytes (Fig. 3A). Ward’s clustering of tagdensity distributions filtered on WT Tcf-1 sites showed thatsimilarities between Tcf-1 and H3K4me3 patterns persist, al-though some Tcf-1 sites are lost or have reduced enrichment in

Fig. 1. β-Catenin activation causes genomic in-stability. Representative images of (A) SKY meta-phase from a CAT lymphoma with Tcra/Myctranslocations (see Table S1 for full karyotypes) and(B) FISH analysis. (C) Cartoon of the translocation aspredicted by FISH and the presence of a fusiontranscript (Fig. S1B). BACs used for FISH analysis areindicated. Not drawn to scale.

392 | www.pnas.org/cgi/doi/10.1073/pnas.1315752111 Dose et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1

CAT thymocytes (Fig. 3B). This trend excluded trimethylatedTcf-1 sites, which predict RAG2 binding (Fig. 2). The number oftrimethylated Tcf-1 sites remained unaltered (Fig. 3C), with∼70% completely identical sites in WT and CAT thymocytes.These findings indicated that activation of β-catenin did notmarkedly influence the deposition of H3K4me3 in the proximityof Tcf-1 sites. We thus predicted that RAG2 binding would alsobe unaltered by stabilization of β-catenin. Indeed, quantitativeChIP (ChIP followed by qPCR) of RAG2 at select binding sitesshowed that RAG2 binding did not significantly change in CATDP thymocytes compared with WT (Fig. S2). We conclude thatactivation of β-catenin does not significantly alter trimethylatedTcf-1 sites or RAG2 binding. Thus, any illegitimate RAG activitywould target similar Tcf-1 proximal sites in CAT cells, but high levelsof β-catenin may interfere with the protection of these sites.

Increased DNA Damage at Paired Tcra/Myc Loci in CAT Thymocytes.To pursue this idea further, we inspected the translocation sitesin pretransformed thymocytes. The Pvt1 locus lies within an earlyreplicating chromosome fragile region (27). Such sites have beenproposed to sustain frequent DSBs and become acceptor sites fortranslocations when NHEJ repair is compromised (27). We foundseveral overlapping Tcf-1 and RAG2 binding sites near the trans-location breakpoint at the Pvt1 locus (Fig. 4A), supporting thenotion that Pvt1 may sustain off-target RAG-mediated DSBs.A prerequisite for translocations to occur is that the trans-

locating loci have DSBs when they are in close proximity. DSBsnormally trigger the DNA damage response that involves phos-phorylation of histone H2AX at serine 139 (γ-H2AX) aroundthe damaged region (28). To examine whether proximal Tcra andMyc loci had increased damage, we compared the presence ofγ-H2AX on these loci in WT and CAT thymocytes. Three-dimensional immuno-DNA FISH analyses confirmed that in25% of DP thymocytes the Tcra and Myc loci exist in closeproximity (Fig. 4B). However, in WT cells, damage is preventedfrom spreading to the Myc locus because γ-H2AX is rarely foundon Myc. Only 0.7% of WT cells with damaged Tcra also haddamaged Myc (Fig. 4 C and D, i, and Table S2). This differed inCAT thymocytes where damaged Tcra and Myc loci cooccurred

in the same cell eight times more frequently (P = 1.1 × 10−3; Fig.4D, i). Moreover, when Tcra and Myc were in close proximity(“paired”), only CAT but not WT thymocytes had DNA damageon both loci (P = 0.02; Fig. 4D, ii). These data suggest thatβ-catenin activation causes increased DNA damage and/or ab-errant repair at off-target DSBs (in this case the Myc locus)resulting in translocations.

β-Catenin Activation Affects Repair of RAG-Mediated DSBs. Havingshown a genome-wide overlap of Tcf-1 and RAG2 binding andincreased DNA damage, we asked whether CAT cells hadcompromised DNA repair. Defects in NHEJ DNA repair arereflected in the sequence composition of coding joints (CJs),generated during repair of RAG-mediated DSBs. We thereforesequenced Tcra CJs from thymocytes of three WT and threeCAT mice using an established assay (29). Two hundred elevenunique WT and 243 CAT Va2-Ja56 Tcra CJ sequences wereannotated with respect to the presence of N- or P-nucleotides(Table S3). N-nucleotides are nontemplated nucleotides addedduring NHEJ repair by TdT, whereas P-nucleotides result fromasymmetric opening of a coding end hairpin (Fig. 4E). Tabula-tion of the data showed that TdT failed to add N-nucleotides in asignificantly higher fraction of CAT CJ compared with WT (P =0.04, t test; Fig. 4F). This defect was not due to reduced TdTlevels as components of the NHEJ machinery, including TdT,are expressed at comparable levels in WT and CAT thymocytes(Fig. S3). The reduced frequency of N-nucleotide addition sug-gests impaired NHEJ and is reminiscent, albeit milder, of aphenotype that has been described in Ku80-deficient cells (30).Like Ku80-deficient cells, CAT cells had no significant differ-ences in the number of P-nucleotides (P = 0.7, t test) or theoverall number of deletions or insertions at the Tcra locus (Fig.4F and Table S3). This finding suggests that activation ofβ-catenin in thymocytes may affect repair of DNA DSBs.

Activation of β-Catenin Provides a “Survival License” to CAT Thymocytes.Studies in mice suggest that genomically unstable leukemia ariseswhen the balance between DNA repair, cell cycle progression, andsurvival is disturbed. Unresolved DSBs normally trigger cell cyclearrest to allow time for repair or cell death in case repair is notpromptly accomplished (31). However, failure to coordinate DNA

Fig. 2. Trimethylated Tcf-1 sites predict RAG2 binding. ChIP-seq analyses ofWT thymocytes. (A) Similar genome-wide binding patterns of RAG2, Tcf-1,and H3K4me3. Tag density is plotted for a representative area on Chr6. (B)Similarity analysis using Ward’s clustering of chromatin occupancy by Tcf-1and RAG2 (described in ref. 53). (C) Normalized tag density distributioncentered on shared trimethylated Tcf-1 sites. Solid lines, ChIP; dotted lines,control. Lower right shows overlay of ChIP signal for Tcf-1, RAG2, andH3k4me3. (D) Percentage of peaks that RAG2 shares (■) or that are sharedwith RAG2 (□) for the indicated subsets. (E) Log2 expression of all genesexpressed in DP thymocytes or stratified by the presence of Tcf-1, RAG2, andH3K4me3 (me3) at the gene promoter.

Fig. 3. Trimethylated Tcf-1 sites are preserved in CAT thymocytes. ChIP-seq analyses of WT and CAT thymocytes. (A) Tcf-1 signal at sites thatoverlap between WT and CAT plotted as a percentage of maximum. (B)Ward’s hierarchical clustering of tag density for similarity analysis as in Fig.2. me3, H3K4me3. (C) Absolute number of WT and CAT Tcf-1 sites thatoverlap with H3K4me3 and are shared between WT and CAT (black) andoverlap with H3K4me3 but are not shared between WT and CAT (gray) anddo not overlap with H3K4me3 (white).

Dose et al. PNAS | January 7, 2014 | vol. 111 | no. 1 | 393

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1

repair and apoptosis—for example, in mice deficient for NHEJcomponents as well as p53—induces B-cell lymphomas withIgh/Myc translocations analogous to the Tcra/Myc translocationsdescribed here. We therefore reasoned that high levels of β-cateninmay promote transformation by altering the balance between cellcycle progression and survival. To examine whether CAT thy-mocytes harbored cell cycle defects, we performed BrdU pulselabeling paired with DNA content analyses. We observed nodifferences between CAT and WT DP thymocytes with respect tothe fraction of cells in the different phases of the cell cycle (Fig. 5A and B). Moreover, we had previously shown that CAT thymo-cytes do not have p53 loss-of-function mutations (19). To furtherrule out cell cycle defects, we determined RAG2 protein levels inG1 versus the S/G2/M phases of the cell cycle. RAG2 is normallypresent at high levels in G1 but is degraded during S/G2/M (32).To evaluate RAG2 levels along the cell cycle, we sorted DPthymocytes from neonatal WT and CAT mice stained with thecell permeable Hoechst 33342 dye according to their DNAcontent. RAG2 protein levels in G1 (2N DNA content) versusS/G2/M (>2N DNA content) were similar in CAT and WT DPthymocytes, indicating normal cell cycle control of RAG2 degrada-tion (Fig. 5C). Furthermore, WT and CAT thymocytes had com-parable levels of Tcf-1 protein both in G1 and S/G2/M phases of thecell cycle, indicating that there was no imbalance between Tcf-1 andRAG2 (Fig. 5C). In summary, therefore, CAT DP thymocytes haveno detectable cell cycle defects.We next asked whether CAT thymocytes showed enhanced

survival. Gene ontology analysis of our transcriptome data (19)revealed that pretransformed CAT cells had an antiapoptosisexpression signature with significant down-regulation of apoptotic

and up-regulation of survival pathways (Figs. S4 and 6A). Fur-thermore, the antiapoptotic protein Bcl-xL was robustly up-regulated in CAT thymocytes, consistent with earlier suggestionsthat it is directly targeted by β-catenin (33) (Fig. 6B). Despiteup-regulation of Bcl-xL, CAT and WT thymocytes showed sim-ilar spontaneous cell death (Fig. 6C). To assess if Bcl-xL provideda survival advantage in response to challenge, we treated WTand CAT thymocytes with genotoxic γ-irradiation (1.25 Gy) andcompared their survival. Indeed, CAT thymocytes were moreresistant to DNA damage than their WT counterparts (Fig. 6D).This resistance was dependent on increased Bcl-xL levels becausetreatment of irradiated cells with ABT263, a pharmacologi-cal inhibitor of the Bcl2 protein family with highest affinity forBcl-xL (34), eliminated the survival advantage of CAT thymo-cytes (Fig. 6D). Moreover, the resistance of CAT thymocytes waslimited to treatment with genotoxic agents, including the top-oisomerase II inhibitor etoposide, and γ-irradiation (Fig. 6E). Bycontrast, a comparable fraction of WT and CAT thymocytes sur-vived treatment with Brefeldin A, an inhibitor of vesicular traffic(Fig. 6E). β-Catenin thus enhances cellular survival in the presenceof DNA damage.

DiscussionWe previously reported that activation of β-catenin at the DPstage of thymocyte development leads to T-cell lymphomas thatdepend on RAG-induced DNA breaks and c-Myc expression(19). These earlier observations had prompted us to speculatethat the resulting lymphomas were genomically unstable. Herewe confirm this hypothesis by demonstrating that the lymphomashave widespread genomic changes including recurrent trans-locations of TCRα to the Myc/Pvt1 locus, and we providemechanistic insights into the transformation process.Activation of β-catenin provides CAT thymocytes with a sur-

vival license as it represses apoptosis and activates survival path-ways. This pattern also includes BclXL that is critical for survivalof DP thymocytes. Indeed, CAT thymocytes survive better whenchallenged with DSB inducing agents, a survival advantage thatis lost when BclXL is pharmacologically inhibited. Survival ofcells with damaged DNA is known to facilitate oncogenic trans-locations. However, increased survival is not sufficient for trans-formation because mice that overexpress the antiapoptotic Bcl-2or Bcl-xL do not develop T-cell malignancies comparable to ourmodel (35). Uncontrolled β-catenin kills two birds with one stoneas it simultaneously promotes survival and DNA damage. Thisseems to depend on the levels of β-catenin because models withmoderate up-regulation of β-catenin develop lymphomas onlywhen p53 is also ablated (36). Our model, based on constitutiveactivation of endogenous β-catenin, expresses high levels ofβ-catenin that evidently suffice to tip the balance toward trans-formation without the need for additional genetic manipulation.Our studies suggest that high levels of β-catenin may influence

the repair of DSBs. In this study, we mapped the genome-wideDNA binding pattern of Tcf-1 in DP thymocytes. This will bea valuable resource for future investigations regarding the tran-scriptional control of T-cell differentiation. We unexpectedlyrevealed a genome-wide overlap between Tcf-1 and the DNA

Fig. 4. CAT thymocytes have increased DNA damage at translocating lociand altered repair of DSBs. (A) Cooccupancy of Tcf-1 and RAG2 at thetranslocating Myc/Pvt1 locus. (B–D) Three-dimensional FISH of DP thymo-cytes. (B) Tcra and Myc loci at <1 μm (NWT = 304, NCAT = 317). (C) γ-H2AXassociation on Tcra/Myc pairs either exclusively on Tcra (Upper) or on bothalleles (Lower). (D) γ-H2AX association on both loci in the same cell, (D, i)irrespective of pairing status or (D, ii) in Tcra/Myc pairs, as percentage oftotal cells (D, i) or of cells with ≥1 Tcra/Myc pair (D, ii). Asterisks indicatestatistical significance. P values: (D, i) 1.1 × 10−3 and (D, ii) 1.9 × 10−2. nd,none detected. (Scale bars, 1 μm.) (E) Exemplary CJ sequences along germ-line sequence (top) to illustrate P (blue) and N nucleotides (orange), dele-tions, and insertions identified in the data set (see Table S3 for all sequen-ces). (F) CJ sequencing results from three mice per genotype were pooledand subjected to statistical analysis (Materials and Methods).

Fig. 5. β-catenin activation does not impair cell cycle checkpoints. (A) Cellcycle profiles of WT and CAT DP thymocytes from mice injected with BrdU3 h before analysis. (B) Histograms summarize data from two mice per ge-notype. (C) Western blot analysis of sorted DP cells in G1 or S/G2/M states ofcell cycle. Thymi from newborn mice were pooled for this experiment.

394 | www.pnas.org/cgi/doi/10.1073/pnas.1315752111 Dose et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1

recombinase RAG2 at trimethylated Tcf-1 binding sites (Tcf-1sites that are marked by H3K4me3 histone marks). We showedthat such sites are predictive of RAG2 binding in thymocytes,and although it has been suggested that RAG2 and Tcf-1 pro-teins do not directly interact (37), their physical proximity on theDNA may imply a functional relationship. CAT thymocytes hadsigns of altered repair as TCR CJ frequently lacked N-nucleo-tides. This is also a feature of Ku80-deficient cells, which has ledto suggestions that Ku80 facilitates access to DSBs for the N-nucleotide adding enzyme TdT (38). β-Catenin has been pro-posed to outcompete an interaction of KU70 with TCF-4 (39),the epithelial cell counterpart of Tcf-1. In further support of thenotion that β-catenin activation may affect DNA repair, we ob-served increased DNA damage in the Myc/Pvt1 locus of pre-malignant CAT DP thymocytes. Taken together, these observationsprovide a mechanistic framework for how high levels of β-cateninmight induce RAG-dependent, genomically unstable lymphomas(19) with recurrent translocations.Wnt signaling relies on β-catenin accumulation for its activity,

and if this is an impediment for genome stability, then strict controlof the amplitude and duration of Wnt activity becomes essential.Apparently, hematopoietic development requires precise lineage-specific dosage of Wnt signaling (40). It has also been suggestedthat Tcf-1 restrains Lef-1 from uncontrollably activating Wnt sig-naling and promoting Lef-1–dependent transformation of Tcf-1–deficient early thymocytes (41, 42). Furthermore, oscillations of

Wnt target gene expression during somitogenesis lead to sug-gestions that a negative feedback loop controls the length of Wntsignaling (8). Deregulated β-catenin activation eliminates theseoscillations and predisposes to cancer. Therefore, multiple lines ofevidence indicate that short-lived bursts and tightly controlled levelsof Wnt activity are required for normal development.Our data provide a fresh perspective on β-catenin and Tcf-1

functions and offer unique paradigms for the etiology of humanleukemia with translocations involving the TCR loci (43, 44).About 2% of human T-ALL have similar translocations to theones detected in CAT mice (21). Sequencing identified crypticRAG recombination signal sequences in the MYC/PVT1 sites ofthe translocations, leading to suggestions that the translocationsresulted from illegitimate RAG recombination events. Murinelymphomas resulting from Pten ablation (45, 46) or Akt activa-tion (47) also show recurrent Tcra/Myc translocations with similararchitecture. Notably, β-catenin is required for the generation ofPten-deficient lymphomas (45). Both Pten ablation and Akt ac-tivation stabilize β-catenin. Detection of similar translocations inthe three animal models prompts us to hypothesize an “axis ofgenome integrity” with β-catenin/Tcf-1 as its most downstream com-ponent. Finally, the organization of the Tcra/Myc translocations inT cells is reminiscent of Igh/Myc translocations in human Burkitt’slymphomas and B-cell lymphomas arising in mice that are doubledeficient in NHEJ and p53 (48, 49). This raises the possibility thatsuch translocations share similar etiologies.Altogether, our findings demonstrate that in thymocytes,

constitutive activation of β-catenin confers genomic instability.They also provide a unique perspective as to why the levels ofβ-catenin are so tightly controlled during normal development.Future studies will be necessary to gain further insights into Tcf-1and β-catenin functions in genome protection and to addresshow they promote genomic instability in human cancer.

Materials and MethodsAnimals. BALB/c CD4Cre;Ctnnb1Δex3 (CAT) mice (19) were used for SKY and FISHanalysis, and C57BL/6 mice (Jackson) were used for all other experiments. Micewere kept in the animal facilities of the University of Chicago according to pro-tocol #71880, approved by the local Institutional Animal Care and Use Committee.

SKY Analyses. Sick CAT mice were euthanized, and SKY was performed onthymic lymphoma cells as described (50). Karyotyping results are in Table S1.

FISH. Labeled BAC probes for Tcra (RP23-105B7; Spectrum OrangeTM) orMyc(RP23-442F1; Spectrum GreenTM) were labeled with nucleotides (AbbottMolecular Diagnostics). FISH was performed as described previously (50).

Three-Dimensional Immuno-DNA FISH. DNA FISH-immunofluorescence forγ-H2AX was carried out on interphase nuclei of sorted DPs as in ref. 51. LabeledBAC probes RP23-255N13 (3′ end of the Tcra locus) and RP24-307D14 (Myc) wereused. Distances between alleles were measured between the center of mass ofeach BAC signal. Alleles were defined as associated with γ-H2AX if BAC signalsand immunofluorescence foci overlapped by at least one pixel. Statistical sig-nificance was calculated by a two-tail Fisher exact test in a pair-wise analysis.Sample sizes were 100 cells minimum per experiment. Data from individualexperiments were pooled for the figure but are shown separately in Table S2.

ChIP-seq. A total of 108 thymocytes from 4-wk-old mice were formaldehyde-fixed and sonicated to a size of ∼300 bp. Antibodies to Tcf-1 (kind gift of HiroshiKawamoto, Kyoto University, Kyoto, Japan) coupled to Protein G Dynabeadswere incubated overnight with sheared chromatin. ChIP-seq libraries were pre-pared from 10 ng of IPed material as before (52) and sequenced on an IlluminaGenome Analyzer 2. Data were analyzed as described (52). Peaks were calledwith model-based analysis of ChIP-Seq (MACS) at a P value cutoff of 10−5. Wehave previously described ChIP-seq for H3K4me3 in WT thymocytes (52). Thesedata are publicly available under accession GSE32311.

RAG2 ChIP. RAG2ChIP followedby quantitative PCRwasperformedasdescribed(1). PCR data were analyzed as in ref. 53. Primer sequences are in Table S4.

Sequencing of Coding Joints (CJ Sequencing). Genomic DNA (50 ng) fromsorted DPs of three WT and three CAT mice was amplified as in ref. 29 (primer

Fig. 6. β-Catenin confers a survival profile upon thymocytes. (A) Heat mapdepicting average fold change of expression between CAT and WT for theindicated genes in WT and CAT DP thymocytes (biological replicates in col-umns). (B) β-catenin stabilization increases Bcl-XL protein levels. Densitom-etry was performed on Western blot images, and density is expressedrelative to β-Actin (Top). Statistical significance was assessed using an unpaired,two-tailed t test (P = 10−5, n = 6 per group). (C–E) Freshly isolated WT and CATthymocytes were cultured as follows, and cell viability was assessed byFACS. (C) Spontaneous cell death with DMSO. (D) Specific cell death (i.e.,relative to the survival of mock treated cells) upon γ-irradiation and 20 h ofculture with ABT263 or DMSO (mock). (E ) Specific cell death upon treatmentwith etoposide, γ-irradiation, or Brefeldin A.

Dose et al. PNAS | January 7, 2014 | vol. 111 | no. 1 | 395

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1

sequences are in Table S4). PCR products were cloned and sequenced at theDNA core facility (University of Chicago). Only unique sequences were an-alyzed to avoid PCR duplicates and N- and P-nucleotides were annotated(Table S3). Replicates were considered repeat measures of the populationmean [p(noN) = 0.042, p(P) = 0.703, t test]. Alternatively, a two-sided Fisher’sexact test was applied to pooled data [p(noN) = 0.0194, p(P) = 0.726].

BrdU Uptake Experiment for Cell Cycle State Analysis. Littermates were injectedi.v. with 10mg BrdU in PBS. Three hours later they were killed, and thymocyteswere stained for CD4, CD8, and TCRβ and then intracellularly with 7-AAD andanti-BrdU and were analyzed on a Fortessa machine (BD Bioscience).

ACKNOWLEDGMENTS. We thank D. Schatz, G. Teng, and H. Kawamoto forreagents; M. Clark, K. Khazaie, and D. Schatz for advice; and C. Daly,E. Bartom, and M. Maienshein-Cline for technical assistance and data mining.This work was supported by National Institutes of Health Grants R21AI076720(to F.G.), R01CA158006 (to K. Georgopoulos), F31AI830542 (to K. Germar), andR01GM086852 (to J.A.S.), and R01CA158006 (to K. Georgopoulos). Furthersupport came from the American Cancer Society Grant ACS/RSG, LIB-113428(to F.G.); Chicago Biomedical Consortium (F.G.); The Lady Tata Memorial Trust(M.D.); and a Specialized Center of Research of the Leukemia and LymphomaSociety (LLS) [SCOR R7019-04 (M.M.L.B.)]. J.A.S. is an LLS scholar. J.C. is anIrvington Institute Fellow of the Cancer Research Institute.

1. Ji Y, et al. (2010) The in vivo pattern of binding of RAG1 and RAG2 to antigen re-ceptor loci. Cell 141(3):419–431.

2. Liu Y, Subrahmanyam R, Chakraborty T, Sen R, Desiderio S (2007) A plant homeo-domain in RAG-2 that binds Hypermethylated lysine 4 of histone H3 is necessary forefficient antigen-receptor-gene rearrangement. Immunity 27(4):561–571.

3. Matthews AG, et al. (2007) RAG2 PHD finger couples histone H3 lysine 4 trimethy-lation with V(D)J recombination. Nature 450(7172):1106–1110.

4. Schatz DG, Ji Y (2011) Recombination centres and the orchestration of V(D)J re-combination. Nat Rev Immunol 11(4):251–263.

5. Lieber MR (2008) The mechanism of human nonhomologous DNA end joining. J BiolChem 283(1):1–5.

6. Nussenzweig A, Nussenzweig MC (2010) Origin of chromosomal translocations inlymphoid cancer. Cell 141(1):27–38.

7. Polakis P (2000) Wnt signaling and cancer. Genes Dev 14(15):1837–1851.8. Jensen PB, Pedersen L, Krishna S, Jensen MH (2010) A Wnt oscillator model for so-

mitogenesis. Biophys J 98(6):943–950.9. Gandhirajan RK, Poll-Wolbeck SJ, Gehrke I, Kreuzer KA (2010) Wnt/β-catenin/LEF-1

signaling in chronic lymphocytic leukemia (CLL): A target for current and potentialtherapeutic options. Curr Cancer Drug Targets 10(7):716–727.

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

11. Luis TC, Ichii M, Brugman MH, Kincade P, Staal FJ (2012) Wnt signaling strengthregulates normal hematopoiesis and its deregulation is involved in leukemia de-velopment. Leukemia 26(3):414–421.

12. Román-Gómez J, et al. (2007) Epigenetic regulation of Wnt-signaling pathway inacute lymphoblastic leukemia. Blood 109(8):3462–3469.

13. Germar K, et al. (2011) T-cell factor 1 is a gatekeeper for T-cell specification in re-sponse to Notch signaling. Proc Natl Acad Sci USA 108(50):20060–20065.

14. Weber BN, et al. (2011) A critical role for TCF-1 in T-lineage specification and dif-ferentiation. Nature 476(7358):63–68.

15. Verbeek S, et al. (1995) An HMG-box-containing T-cell factor required for thymocytedifferentiation. Nature 374(6517):70–74.

16. Schilham MW, Clevers H (1998) HMG box containing transcription factors in lym-phocyte differentiation. Semin Immunol 10(2):127–132.

17. Ioannidis V, Beermann F, Clevers H, Held W (2001) The beta-catenin—TCF-1 pathwayensures CD4(+)CD8(+) thymocyte survival. Nat Immunol 2(8):691–697.

18. Wang R, et al. (2011) T cell factor 1 regulates thymocyte survival via a RORγt-dependent pathway. J Immunol 187(11):5964–5973.

19. Guo Z, et al. (2007) Beta-catenin stabilization stalls the transition from double-posi-tive to single-positive stage and predisposes thymocytes to malignant transformation.Blood 109(12):5463–5472.

20. Harada N, et al. (1999) Intestinal polyposis in mice with a dominant stable mutation ofthe beta-catenin gene. EMBO J 18(21):5931–5942.

21. Larmonie NS, et al. (2013) Breakpoint sites disclose the role of the V(D)J re-combination machinery in the formation of T-cell receptor (TCR) and non-TCR asso-ciated aberrations in T-cell acute lymphoblastic leukemia. Haematologica 98(8):1173–1184.

22. Shima-Rich EA, Harden AM, McKeithan TW, Rowley JD, Diaz MO (1997) Molecularanalysis of the t(8;14)(q24;q11) chromosomal breakpoint junctions in the T-cell leu-kemia line MOLT-16. Genes Chromosomes Cancer 20(4):363–371.

23. Graham M, Adams JM, Cory S (1985) Murine T lymphomas with retroviral inserts in thechromosomal 15 locus for plasmacytoma variant translocations. Nature 314(6013):740–743.

24. Villeneuve L, Rassart E, Jolicoeur P, Graham M, Adams JM (1986) Proviral integrationsite Mis-1 in rat thymomas corresponds to the pvt-1 translocation breakpoint inmurine plasmacytomas. Mol Cell Biol 6(5):1834–1837.

25. Graham M, Adams JM (1986) Chromosome 8 breakpoint far 3′ of the c-myc oncogenein a Burkitt’s lymphoma 2;8 variant translocation is equivalent to the murine pvt-1locus. EMBO J 5(11):2845–2851.

26. Cory S, GrahamM, Webb E, Corcoran L, Adams JM (1985) Variant (6;15) translocationsin murine plasmacytomas involve a chromosome 15 locus at least 72 kb from thec-myc oncogene. EMBO J 4(3):675–681.

27. Debacker K, Kooy RF (2007) Fragile sites and human disease. Hum Mol Genet 16(R2):R150–R158.

28. Peterson CL, Côté J (2004) Cellular machineries for chromosomal DNA repair. GenesDev 18(6):602–616.

29. Difilippantonio S, et al. (2008) 53BP1 facilitates long-range DNA end-joining duringV(D)J recombination. Nature 456(7221):529–533.

30. Zhu C, Bogue MA, Lim DS, Hasty P, Roth DB (1996) Ku86-deficient mice exhibit severecombined immunodeficiency and defective processing of V(D)J recombination in-termediates. Cell 86(3):379–389.

31. Jackson SP, Bartek J (2009) The DNA-damage response in human biology and disease.Nature 461(7267):1071–1078.

32. Lin WC, Desiderio S (1994) Cell cycle regulation of V(D)J recombination-activatingprotein RAG-2. Proc Natl Acad Sci USA 91(7):2733–2737.

33. Xie H, Huang Z, Sadim MS, Sun Z (2005) Stabilized beta-catenin extends thymocytesurvival by up-regulating Bcl-xL. J Immunol 175(12):7981–7988.

34. Tse C, et al. (2008) ABT-263: A potent and orally bioavailable Bcl-2 family inhibitor.Cancer Res 68(9):3421–3428.

35. Chao DT, et al. (1995) Bcl-XL and Bcl-2 repress a common pathway of cell death. J ExpMed 182(3):821–828.

36. Xu M, et al. (2008) Beta-catenin expression results in p53-independent DNA damageand oncogene-induced senescence in prelymphomagenic thymocytes in vivo.Mol CellBiol 28(5):1713–1723.

37. Aidinis V, et al. (1999) The RAG1 homeodomain recruits HMG1 and HMG2 to facilitaterecombination signal sequence binding and to enhance the intrinsic DNA-bendingactivity of RAG1-RAG2. Mol Cell Biol 19(10):6532–6542.

38. Boubakour-Azzouz I, Bertrand P, Claes A, Lopez BS, Rougeon F (2012) Terminal de-oxynucleotidyl transferase requires KU80 and XRCC4 to promote N-addition at non-V(D)J chromosomal breaks in non-lymphoid cells. Nucleic Acids Res 40(17):8381–8391.

39. Idogawa M, et al. (2007) Ku70 and poly(ADP-ribose) polymerase-1 competitivelyregulate beta-catenin and T-cell factor-4-mediated gene transactivation: Possiblelinkage of DNA damage recognition and Wnt signaling. Cancer Res 67(3):911–918.

40. Luis TC, et al. (2011) Canonical wnt signaling regulates hematopoiesis in a dosage-dependent fashion. Cell Stem Cell 9(4):345–356.

41. Yu S, et al. (2012) The TCF-1 and LEF-1 transcription factors have cooperative andopposing roles in T cell development and malignancy. Immunity 37(5):813–826.

42. Tiemessen MM, et al. (2012) The nuclear effector of Wnt-signaling, Tcf1, functions as a T-cell-specific tumor suppressor for development of lymphomas. PLoS Biol 10(11):e1001430.

43. Mathieu-Mahul D, et al. (1985) Molecular cloning of a DNA fragment from humanchromosome 14(14q11) involved in T-cell malignancies. EMBO J 4(13A):3427–3433.

44. Shima EA, et al. (1986) Gene encoding the alpha chain of the T-cell receptor is movedimmediately downstream of c-myc in a chromosomal 8;14 translocation in a cell linefrom a human T-cell leukemia. Proc Natl Acad Sci USA 83(10):3439–3443.

45. Guo W, et al. (2008) Multi-genetic events collaboratively contribute to Pten-nullleukaemia stem-cell formation. Nature 453(7194):529–533.

46. Liu X, et al. (2010) Distinct roles for PTEN in prevention of T cell lymphoma and au-toimmunity in mice. J Clin Invest 120(7):2497–2507.

47. Timakhov RA, et al. (2009) Recurrent chromosomal rearrangements implicate onco-genes contributing to T-cell lymphomagenesis in Lck-MyrAkt2 transgenic mice. GenesChromosomes Cancer 48(9):786–794.

48. Difilippantonio MJ, et al. (2002) Evidence for replicative repair of DNA double-strandbreaks leading to oncogenic translocation and gene amplification. J Exp Med 196(4):469–480.

49. Zhu C, et al. (2002) Unrepaired DNA breaks in p53-deficient cells lead to oncogenicgene amplification subsequent to translocations. Cell 109(7):811–821.

50. Le Beau MM, et al. (1996) Cytogenetic and molecular delineation of a region of chro-mosome 7 commonly deleted in malignant myeloid diseases. Blood 88(6):1930–1935.

51. Hewitt SL, et al. (2009) RAG-1 and ATM coordinate monoallelic recombination andnuclear positioning of immunoglobulin loci. Nat Immunol 10(6):655–664.

52. Zhang J, et al. (2012) Harnessing of the nucleosome-remodeling-deacetylase complexcontrols lymphocyte development and prevents leukemogenesis. Nat Immunol 13(1):86–94.

53. Dose M, et al. (2006) c-Myc mediates pre-TCR-induced proliferation but notdevelopmental progression. Blood 108(8):2669–2677.

396 | www.pnas.org/cgi/doi/10.1073/pnas.1315752111 Dose et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

24,

202

1