Chromosome instability induced by Mps1 and p53mutation generates aggressive lymphomasexhibiting aneuploidy-induced stressFloris Foijera,b,c,1,2, Stephanie Z. Xieb,1,3, Judith E. Simona, Petra L. Bakkera, Nathalie Contec, Stephanie H. Davisb,Eva Kregeld, Jos Jonkersd, Allan Bradleyc, and Peter K. Sorgerb,2

aEuropean Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, NL-9713 AV, Groningen, TheNetherlands; bDepartment of Systems Biology, Harvard Medical School, Boston, MA 02115; cMouse Genomics, Wellcome Trust Sanger Institute, Hinxton CB101SA, United Kingdom; and dDivision of Molecular Pathology, The Netherlands Cancer Institute, NL-1066 CX, Amsterdam, The Netherlands

Edited by Don W. Cleveland, University of California, San Diego, La Jolla, CA, and approved August 5, 2014 (received for review January 17, 2014)

Aneuploidy is a hallmark of human solid cancers that arises fromerrors in mitosis and results in gain and loss of oncogenes and tumorsuppressors. Aneuploidy poses a growth disadvantage for cellsgrown in vitro, suggesting that cancer cells adapt to this burden.To understand better the consequences of aneuploidy in a rapidlyproliferating adult tissue, we engineered a mouse in which chromo-some instability was selectively induced in T cells. A flanked byLox mutation was introduced into the monopolar spindle 1 (Mps1)spindle-assembly checkpoint gene so that Cre-mediated recombina-tion would create a truncated protein (Mps1DK) that retained thekinase domain but lacked the kinetochore-binding domain andthereby weakened the checkpoint. In a sensitized p53+/− backgroundwe observed that Mps1DK/DK mice suffered from rapid-onset acutelymphoblastic lymphoma. The tumors were highly aneuploid andexhibited a metabolic burden similar to that previously character-ized in aneuploid yeast and cultured cells. The tumors nonethelessgrew rapidly and were lethal within 3–4 mo after birth.

chromosomal instability | mouse models | CIN | tumor metabolism |T-cell acute lymphoblastic lymphoma

Aneuploidy is a hallmark of oncogenesis, affecting two out ofthree cancers (1). Aneuploidy arises during mitosis as a re-

sult of chromosomal instability (CIN) (2–4). The frequent oc-currence of CIN in solid human tumors suggests a fundamentallink between aneuploidy and cancer (5). However, primary mouseembryonic fibroblasts (MEFs) carrying a supernumerary chro-mosome have decreased proliferative potential, as do cells isolatedfrom Down syndrome patients (6, 7) implying that chromosomeimbalance imposes a physiological burden that lowers fitness, atleast in untransformed cells (6, 8–10). In somemousemodels, CINappears to have a significant impact on lifespan at the organismallevel, with increased aneuploidy decreasing life expectancy andvice versa (11–13). However, how the fitness cost imposed by CINis balanced by its potential to promote oncogenic transformationremains poorly understood.Mouse models of CIN involving conditional or hypomorphic

mutations in spindle-assembly checkpoint (SAC) genes providea means to study aneuploidy and assess its impact on cell fitnessand oncogenesis. The SAC detects the presence of maloriented ordetached kinetochores during mitosis and arrests cells in meta-phase until all pairs of sister chromatids achieve the biorientedgeometry that is uniquely compatible with normal disjunction(14–16). The SAC constitutes a signaling cascade [comprisingmonopolar spindle 1 (Mps1), Bub, Mad, CenpE, and RZZ pro-teins] that blocks activation of the anaphase-promoting complex,and thus mitotic progression, until all chromosomes are properlyaligned (17). In the mouse, germ-line deletion of SAC genesresults in early embryonic lethality, whereas heterozygousknockout of Mad2 and other SAC genes generates relatively weaktumor phenotypes late in life (2–4). Paradoxically, some SACmutations (e.g., CenpE heterozygosity) appear to be both tumor

predisposing and tumor suppressing, depending on the context(18). Hypomorphic BubR1 mutations also have the unexpectedproperty of promoting progeria (11).Conditional mutations typically yield tumor phenotypes more

representative of human disease than germ-line mutations (19),but conditional alleles have been little studied in the case of thespindle checkpoint. We therefore engineered a conditional flankedby Lox (FLOX) mutation into Mps1, a gene thought to functionupstream in the SAC pathway (20) and then selectively truncatedthe protein by expressing Cre recombinase in T cells. The Mps1truncation (deletion in the kinetochore domain; Mps1DK)removes the kinase-targeting domain but leaves the rest of theprotein intact. We show that expression of this truncated proteincauses chromosomal instability in MEFs and aneuploidy in twodifferent Cre-expressing mouse lines. Mps1 truncation in combina-tion with heterozygous p53 deletion leads to early-onset lympho-blastic lymphoma and consequent death. In lymphoma cells,changes in the expression of metabolic, splicing, andDNA-synthesis

Significance

Normal cells rarelymissegregate chromosomes, but themajorityof cancer cells have a chromosomal instability (CIN) phenotypethat makes errors more common and results in abnormal chro-mosomal content (aneuploidy). Although aneuploidy promotestransformation via gain of oncogenes and loss of tumor sup-pressors, it also slows cell proliferation and disrupts metabolichomeostasis. Aneuploidy therefore represents a liability as wellas a source of selective advantage for cancer cells. We provokedCIN in murine T cells by weakening the spindle-assembly check-point and then studied the consequences. We found that CINdramatically accelerates cancer in a genetically predisposedbackground and that the resulting aneuploid cancers are meta-bolically deranged, a vulnerability that may open new avenuesto treating aneuploid cancers.

Author contributions: F.F., S.Z.X., and P.K.S. designed research; F.F., S.Z.X., J.E.S., P.L.B.,S.H.D., and E.K. performed research; N.C. and A.B. contributed new reagents/analytictools; F.F., S.Z.X., J.E.S., N.C., J.J., A.B., and P.K.S. analyzed data; and F.F. and P.K.S. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The data reported in this paper have been deposited in the Gene Ex-pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE57334).1F.F. and S.Z.X. contributed equally to this work.2To whom correspondence may be addressed. Email: f.foijer@umcg.nl or peter_sorger@hms.harvard.edu.

3Present address: Princess Margaret and Toronto General Hospitals, University HealthNetwork, Toronto, Canada M5G 2C1.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1400892111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1400892111 PNAS | September 16, 2014 | vol. 111 | no. 37 | 13427–13432

GEN

ETICS

genes are very similar to changes previously identified in aneu-ploid yeast and cultured murine cells (6, 8) and appear to con-stitute a hallmark of chromosomal imbalance.

ResultsTo provoke CIN in a tissue-restricted fashion, we engineereda conditional Mps1f truncation allele by flanking exons 3 and 4 ofthe Mps1 locus with lox-sites; correct targeting of Mps1 in mouseES cells was confirmed by Southern blotting and RT-PCR (Fig.S1 A–C and SI Materials and Methods). Upon expression of Crerecombinase (21), the Mps1f allele generates a truncated Mps1kinase lacking residues 47–154, a domain involved in kinetochorebinding (Fig. 1A) (22). When expressed in MEFs as a GFP fusion,the Mps1DK protein had the anticipated molecular weight but,unlike wild-type Mps1-GFP, did not accumulate to the same levelson prometaphase kinetochores [Fig. 1B (compare upper and lower

panels), Fig. S1 D and E, and Movies S1 and S2]. We concludethat the Mps1DK mutation impairs but does not prevent kineto-chore binding, a conclusion supported by overexpression studies inhuman MCF 10A cells (Fig. S1 F and G and Movies S3 and S4).

The Mps1DK Truncation Weakens the SAC and Causes CIN. To de-termine the consequences of Mps1 mutation for chromosomesegregation at a cellular level, we isolated Mps1f/f embryos, gener-ated MEFs, and transduced them with retroviruses expressingdoxycycline (Dox)-inducible Cre (GFP-T2A-Cre). We found thatexposure of these cells to Dox resulted in highly efficient switching,yielding Mps1DK MEFs within 24–48 h (Fig. 1C, compare lanes1 and 2 with lane 4). Some recombination also was observed in theabsence of Dox (Fig. 1C, lane 3). When treated with the spindlepoison nocodazole, both control Mps1f/f and Mps1DK MEFs arres-ted in mitosis, showing that both cell types can respond to spindledisassembly (Fig. 1D). However, when time-lapse imaging was usedto assay the duration of arrest, Mps1DK/DK cells were observed toexit mitosis 150 ± 16 min after DNA condensation, in contrast to264 ± 15 min in control cells (congenic Mps1f/f cells not exposedto Cre), a significant difference (P < 0.0001) (Fig. 1E). The ob-servation that cells expressingMps1DK are unable to sustain mitoticarrest in the presence of spindle damage suggests that the SACis impaired but not inactivated by the Mps1DK mutation (23) andconfirms our goal in creating the allele.Time-lapse imaging in the absence of nocodazole showed that

H2B-Cherry–transduced Mps1DK MEFs spent ∼40% longer inmitosis than control cells (Fig. 1F and Fig. S1H). In addition, theyfrequently contained lagging chromosomes (Fig. 1 F and G andMovies S5, S6, and S7; control cells are shown in Movies S8 andS9), and half of all the cells failed to form a proper metaphase plate(Movie 10), resulting in polyploidy (Fig. 1G and Movie S11). Al-though this extended mitosis might appear to be a paradoxicalphenotype for a SAC hypomorph, it has been shown that otherSAC proteins both promote and sense chromosome–microtubuleattachment and that partial inactivation of these proteins actuallylengthens mitosis because the residual SAC function is able tosense incomplete attachment (22). We conclude that the Mps1DK

mutation causes a partial loss of checkpoint function and alsoimpairs kinetochore–microtubule attachment, preventing Mps1DK

cells from stably arresting in the presence of spindle poisons andmissegregating chromosomes under normal growth conditions (22).

Mps1DK Provokes Aggressive T-Cell Acute Lymphoblastic Lymphomain a p53 Heterozygous Background. The Mps1DK truncation wasintroduced into a highly mitotic, nonessential adult tissue bycrossing Mps1f/f mice with animals bearing a T-cell–specific Lck-Cre transgene (24). PCR revealed efficient switching of Mps1f toMps1DK in T cells from 8- to 10-wk-old animals (Fig. S2A) con-comitant with changes in the DNA content of G1 thymocytes(compare peak width and coefficient of variation for G1 peaksin Mps1DK and wild-type animals in Fig. 2 A and B), but life spanwas unaffected (Fig. 2C, red line). When Mps1f/f Lck-Cre+ andwild-type control mice were injected with paclitaxel, a microtu-bule-stabilizing drug that interferes with spindle assembly, ele-vated levels of mitotic cells were observed 5 h later in bothgenotypes (Fig. 2D). This result is consistent with data fromMEFs showing that Mps1DK-expressing cells arrest in the pres-ence of spindle damage.To assay tumor formation in a sensitized environment, we

generated Mps1f/f Lck-Cre+ animals heterozygous for a FLOX-p53 allele (25). Loss of p53 suppresses aneuploidy-associatedapoptosis in multiple cell types and is oncogenic in thymocytes(26–28). Mps1f/fp53f/+ Lck-Cre+ mice rapidly developed T-cellacute lymphoblastic lymphomas (T-ALL); ∼50% of animals haddied of the disease by age 3.5 mo, and 100% had died by age5 mo (Fig. 2C, dark green line). In contrast, heterozygosity atMps1had little effect on the survival of p53-null mice: Mps1f/+p53f/f

A

50 aa

1 830200 520

Kinetochore-binding domain

Kinasedomain

Deleted in Mps1DK

C

E

D

CreDox-ind +-

++

-+

--Dox

Mps1fMps1DK

B

++ + +-CreDox-ind -+- - +--Dox+- + -+Noco

Num

ber o

f mito

tic c

ells

(% p

H3po

s )

-

0

2

4

6

8

10

Mps1WT-GFP Mps1WT-GFP

Mps1DK-GFP Mps1DK-GFP6:00

6:00

12:00

12:00

p<0.0001

***

0 20 40 60 80 100

Time frommetaphase

Time inmitosis

Mps1f/f + CreDox-ind

Mps1f/f

***

p<0.0001

N=36N=90

N=82N=99

Time (min)

F

0 50 100 150 200 250 300

N=34N=64

Time before exiting from mitosis in nocodazole (min)

***

p<0.0001

G

Mps1f/f MEFsCreDox-ind

+-

Normal mitosisLagging chr.TetraploidyzationBinucleationPolyploidyzationFailed mitosisCell deathUnclear

0

20

40

60

80

100

Per

cent

age

of c

ells

CreDox-ind +-

Mps1f/f MEFs

Fig. 1. Mps1 truncation leads to mitotic delay, severe abnormalities, anda weakened SAC. (A) Schematic representation of the Mps1 truncation al-lele. (B) Time-lapse image stills showing clear kinetochore localization ofretrovirally transduced wild-type GFP-Mps1 in prometaphase (Upper) andreduced binding of GFP-Mps1DK to kinetochores (Lower) in MEFs. DNA waslabeled with retroviral H2B-Cherry. (C) PCR detecting the truncation/deletion alleles forMps1 in genomic DNA isolated from control or Cre-infectedMps1f/f MEFs. (D) Average mitotic index of Dox-inducible, Cre-transducedMEFs after 6 h of nocodazole treatment. The mitotic index is the percentageof mitotic cells as measured by phospho-histone H3 (pH3) staining. Error barsshow the SEM of at least four biological replicates. (E) Average time ofmitotic exit for nocodazole-arrested control Mps1f/f (blue) or Cre-infectedMps1f/f (red) MEFs. (F) Average duration of mitosis (Upper) and time frommetaphase to cytokinesis (Lower) of Mps1f/f control (blue) and Cre-infected(red) MEFs as determined by time-lapse microcopy. Error bars show the SEMof the number of cells indicated within the bar. (G) Distribution of mitoticphenotypes for control and Cre-infected Mps1f/f MEFs as observed by time-lapse microscopy. Tetraploidization: a seemingly normal cell failed cytoki-nesis, resulting in one large tetraploid cell; binucleation: a seemingly normalcell failed cytokinesis, resulting in one cell with two nuclei; polyploidization:a seemingly tetraploid or polyploid cell failed cytokinesis, resulting ina polyploid cell; failed mitosis: a combination of mitotic errors.

13428 | www.pnas.org/cgi/doi/10.1073/pnas.1400892111 Foijer et al.

Lck-Cre+ animals had survival curves indistinguishable fromthose of p53f/f Lck-Cre+ mice (Fig. S2B). In addition, Mps1 wild-type p53f/+ Lck-Cre+ mice rarely developed disease, and tumor-free survival was indistinguishable from that of Lck-Cre+ controlanimals (Fig. 2C, compare dark blue and black lines) (29). Thus,the Mps1DK mutation is strongly oncogenic in T cells on a p53f/+

background.Mps1DK promoted loss of heterozygosity (LOH) at the p53

locus: the PCR product corresponding to p53Δ was substantiallymore abundant than the product corresponding to wild-type p53in DNA from Mps1f/fp53f/+ Lck-Cre+ tumors (Fig. S2C, comparetumors 9–17 with 18–21). Quantitative PCR (qPCR) data wereconsistent with this finding: p53 mRNA was virtually undetect-able in tumors recovered from Mps1f/f p53f/+ animals (Fig. 2E).To characterize the LOH event, we extracted probe values fromarray-based comparative genomic hybridization (aCGH) for 17

tumors arising in Mps1f/f p53f/+ animals (Fig. S2D). In all but twoanimals (tumors 12 and 39), hybridization to p53 probes was low,similar to that of p53-null tumors from Mps1f/f p53f/f animals(compare Fig. S2 D and E). Hybridization to neighboring probeswas unaffected, suggesting that the wild-type copy of p53 hadbeen replaced by p53Δ through either CIN or mitotic re-combination. We conclude that Mps1 truncation facilitatesp53 LOH, a highly oncogenic event in thymocytes (29–31).The protumorigenic effects of Mps1 truncation do not appear

to involve p53 LOH alone. Tumor induction was significantlyfaster in Mps1f/f p53f/+ Lck-Cre+ (and Mps1f/f p53f/f Lck-Cre+)mice than in p53f/f Lck-Cre+ mice, with death of 50% of theformer animals by age 3.5 mo as opposed to 5 mo for the latter(P < 0.0001) (Fig. 2C). Analysis of mRNA and genomic DNAconfirmed efficient Cre-mediated deletion of p53 in tumorshaving either genotype (Fig. 2E and Fig. S2 C–E). Moreover,accelerated tumorigenesis in Mps1f/f p53f/f double-mutant animalsrelative to p53f/f animals was confirmed with a second Cre driver,mouse mammary tumor virus (MMTV)-Cre, which is transcribedin both T cells and the mammary gland (Fig. S2F) (32). However,no mammary tumors were observed in these animals, presumablybecause T-ALL developed before breast cancers could emerge.To show that tumors comprised cells in which the Mps1f loci

had been excised and thus that lymphomagenesis was not drivenby p53 loss alone (a concern because p53 is such a strong tumorsuppressor in T cells), we measured the efficiency of Cre-medi-ated recombination at the Mps1 genomic locus. We assayedMps1 mRNA levels using PCR and aCGH, and we performedWestern blotting. In tumors isolated from Mps1f/f p53f/f andMps1f/f p53f/+ mice, bands corresponding to Mps1DK were thepredominant amplified products (Fig. S2G and Dataset S1) inboth Lck-Cre+ andMMTV-Cre+ backgrounds. In the aCGH data,we observed nearly complete loss of hybridization to sequencesexcised by Cre-mediated recombination of the Mps1f/f locus (Fig.2F). qPCR of tumor RNA also confirmed the loss of Mps1 ex-pression: Probes selective for the nonmutated 3′ domain (Mps1probe set B; Fig. 2G and Fig. S2H) yielded a strong qPCRproduct, whereas probes corresponding the 5′ region of Mps1deleted in Mps1DK (Mps1 probe set A) were ∼20-fold lessabundant in tumors than in wild-type thymus DNA. Moreover,RT-PCR followed by Sanger sequencing confirmed the presenceof correctly recombined Mps1DK transcript (in tumors 13 and 23)and full-length Mps1 in p53f/f tumors (tumors 36 and 63) (Fig.S2I and Dataset S2). qPCR also showed that Mps1DK is over-expressed in tumors by approximately fourfold relative to wild-type Mps1 in parental cells, presumably because the elevatedexpression of the hypomorphic allele confers a selective advan-tage on cells. Finally, by Western blotting we could detect aprotein band corresponding to the expected length of the Mps1DK

protein inMps1f/f p53f/f Lck-Cre+ tumor samples (tumors 18 and 19;Fig. 2H). We conclude that theMps1DK allele was maintained in thegreat majority of T cells throughout the development of a tumor,and thus the acceleration in tumorigenesis observed in double-mutant animals reflects ongoing synergy between Mps1 andp53 mutations.

Mps1DK-Driven Tumors Exhibit Recurring Chromosomal Abnormalities.To determine the extent of aneuploidy in Mps1DK T-ALL, weused aCGH to quantify chromosome copy number across thegenome and interphase FISH to quantify chromosome number insingle cells. aCGH revealed frequent loss and gain events formultiple chromosomes (four representative plots are shown inFig. 3A, and normalized aCGH data are summarized in DatasetS3). As a simple measure of CIN, we summed the total number ofchromosome gain and loss events in each tumor to create an“aneuploidy index.” The aneuploidy index ranged from 3 to 19in 30 tumor samples examined and was significantly higher in

A B

0

2

4

6

8

Coe

ffici

ent o

f var

iatio

n G

1

Control Mps1f/f

Lck-Cre+D

Mito

tic c

ells

(%pH

3pos )

- + Paclitaxel + -- + - +Lck-Cre

0.0

0.5

1.0

1.5

2.0

2.5

C

Age (months)

0

20

40

60

80

100

Per

cent

sur

vivi

ng

120 4 8

Mps1f/f p53f/f Lck-Cre+ (n=11)

Lck-Cre+ (n=12)

Mps1f/f Lck-Cre+ (n=29)Mps1f/f p53f/+ Lck-Cre+ (n=27)

p53f/f Lck-Cre+ (n=20)p53f/+ Lck-Cre+ (n=23)

E

Wild type

0

1

2p53 Set Ap53 Set B

Mps1f/f

p53f/+

Cre+

Rel

ativ

e fo

ld e

xpre

ssio

n

Mps1f/f

p53f/f

Cre+

F

G

HMps1WT

Mps1DK

Actin*

38 63 18 19 TID

Con

trol

Mps

1f/f L

ck-C

re+

DNA-content2n 4n

Cel

l num

ber

6

Mps1f/f

p53f/f

Cre+

Wildtype

0

1

2

3

4

5

Mps1 Set AMps1 Set B

Mps1f/f

p53f/+

Cre+

Rel

ativ

e fo

ld e

xpre

ssio

n

-5.0 1:1 5.0

Log2 ratio

Mps1

f/f p53f/f C

re+

p53f/f C

re+

TID

181920212223242548495052

3637386364

Mps1 locus Mps1

f/f Cre

+

265355

710111213141516173942434445464754

Mps1 locus

TID

Mps1

f/fp53f/+ C

re+

Mps1 locus

TID

Fig. 2. Mps1 truncation provokes aneuploidy in vivo and decreases T-ALLlatency in a p53-compromised background. (A) Distribution of DNA contentin control and Mps1DK T cells. At least 10,000 T cells were counted. (B) Co-efficient of variation values for the DNA content within G1 peaks in cell-cycleprofiles of Mps1DK/DK and Cre− T cells. Error bars show the SEM of more thanfive biological replicates for experimental animals or more than two repli-cates for control animals. (C) Kaplan–Meier curves showing overall survivalof indicated genotypes. (D) Average mitotic index (% pH3) of thymocytesisolated from paclitaxel- or control-injected mice 4–6 h posttreatment. Errorbars show SEM of more than five biological replicates in experimental ani-mals or more than two replicates in control animals. (E) qPCRs showingcomplete loss of expression of p53 (p53 probes A and B) in lymphomas ofindicated genotypes. Error bars show SEM for at least three tumors pergenotype. (F) aCGH data showing the loss of the kinetochore-binding se-quence in Mps1 in tumors with the indicated genotypes. Each rectanglerepresents a single aCGH probe value. Three probes values are shown: oneprobe recognizing the kinetochore-binding domain (Center) and two probesflanking the 5′ (Left) and 3′ (Right) sides of the deleted region. Log2 ratiosless than −5 indicate complete loss of the indicated probe. Numbers on theleft refer to tumor identification numbers (TID) (Dataset S1). (G) qPCRsshowing full conversion of Mps1WT to Mps1DK in tumors with the indicatedgenotypes. Primer set A recognizes the sequence deleted in Mps1DK, and setB detects a fragment in the kinase domain. (H) Mps1 protein levels in p53f/f

(TIDs 38 and 63 and full-length Mps1) and Mps1DK p53f/f tumors (TIDs 18 and19) showing the conversion of full-length to truncated Mps1 in the lattergenotype. The asterisk indicates a background band recognized in all lysatesand runs just below Mps1DK that is detected only in TIDs 18 and 19.

Foijer et al. PNAS | September 16, 2014 | vol. 111 | no. 37 | 13429

GEN

ETICS

tumors arising in Mps1f/f p53f/f or Mps1f/f p53f/+ animals (averageindices of 8.2 and 7.6, respectively) than in p53f/f animals (averageindex 2.4; P = 0.047 and P = 0.028, respectively) (Fig. 3B).Similarly, unsupervised single-linkage hierarchical clustering ofcumulative aCGH data showed that tumors from Mps1f/f miceheterozygous (Mps1f/f p53f/f) or homozygous (Mps1f/f p53f/+) forp53 deletion clustered together and had more chromosomal ab-normalities (Fig. 3C; green numbers) than tumors from p53f/f

animals that had less severe aneuploidy (Fig. 3C; blue numbers).Probes lying on the same chromosomes coclustered across all tumorsamples, demonstrating that a significant fraction of the aneuploidyin these tumors involved gain and loss of whole chromosomes.Amplification of Chr15 was particularly frequent in aCGH data,

regardless of genotype, and is known to be common feature ofmouse T-ALLs (7, 33, 34). In addition, amplification of Chr4 andChr14 and to a lesser extent Chr9 was observed in many tumors,and Chr13 and Chr19 were commonly deleted, suggesting thatthose chromosomes carry genes important for transformation oraneuploid tumor progression. Interphase FISH confirmed aneu-ploidy in nontransformed Mps1-mutant thymocytes and hetero-geneity in chromosome number within a single tumor. Forexample, Chr15 and Chr17 were aneuploid in a greater numberof Mps1f/f Lck-Cre+ thymocytes than in wild-type thymocytes(an average of 5% vs. 11% of cells for Chr15 and 6% vs. 12%of cells for Chr17) (Fig. S2J). InMps1f/f p53f/+ Lck-Cre+ T-ALLtumors we observed Chr15 trisomy in up to 80% cells, but the

fraction of cells involved differed among animals (Fig. 3D andFig. S2J).To identify common focal loss and gain events, we performed

genome-wide cumulative segmental gain or loss analysis (SGOL)comparing Mps1DK-driven and p53f/f Lck-Cre+ tumors. For bothtumor classes, SGOL revealed strong deletion peaks on Chr6and Chr14, consistent with a unique pattern of recombined T-cellreceptor α/β loci. These results strongly suggest that tumorsarose from a single parental T cell (Fig. S3 A and B). We canreconcile the SGOL and FISH data by hypothesizing thatT-ALLs are clonal early in their development [at the time ofT-cell receptor (TCR) rearrangement] but that ongoing CINresults in subsequent chromosome loss and gain. Selection isexpected to maintain some aneuploidies, e.g., Chr15 amplifica-tion, whereas other chromosomes (e.g., Chr17) might be sub-jected to ongoing loss and gain.Recent studies on budding yeast and MEFs carrying supernu-

merary chromosomes have shown proportional increases in genecopy number and transcription (6, 8). To determine if theseincreases are present in tumors driven by Mps1 truncation, weused Illumina expression arrays to analyze the transcriptomes of22 tumor samples that previously had been studied by aCGH aswell as thymus DNA isolated from 6-wk-old wild-type mice. Astrong correlation between mRNA and gene copy number wasobserved when aCGH values and expression levels were sorted bychromosomal position (Fig. S3 C and D). When we calculated theaverage expression changes per individual chromosome for eachtumor (using healthy thymic samples as a control) and thencompared the value with the aCGH intensity (Fig. 4 A and B andDataset S4), the correlation between expression and copy numberwas R2 = 0.44–0.76 for the most commonly aneuploid chromo-somes (Chr4, Chr14, and Chr15). (Fig. 4C and Fig. S4 show cor-relation for all chromosomes.) We conclude that in tumors, as incultured cells, chromosomal imbalances on average are translatedinto increases and decreases in transcription, and thus there islittle or no dosage compensation.

Mps1DK Tumors Show Evidence of Aneuploidy-Induced Stress. Toidentify genes significantly over- and underexpressed in T-ALLs,we sorted genes by their cumulative expression changes across allsamples (annotated in Fig. S5A and Dataset S4). For Chr4,Chr14, and Chr15, the majority of genes had a positive cumu-lative score; the reverse was true for Chr19, reflecting the cor-relation between changes in transcription and gene copy number.Among the outliers we found several that exhibited an inversecorrelation between copy number and expression including thekeratins Krt, 5, 7, 8, and 18, Epsi1, and Chdr1. These genes areexpressed in the thymic cortex and not in tumor cells, and theirlower expression in mutant animals likely reflects T-ALL–mediated depletion of cortical tissue. A second set of outlier geneswas significantly overexpressed relative to other genes on the samechromosome. This set includes genes involved in cell metabolism(Srm, Gln3, Cox6a, Drospha, and Adk), cellular stress (Serp2 andHsf1), cell cycle (Recql4, Cdkn2a, Skp2, and Tnc), and epigeneticregulation (Prmt5 and Cbx5). Surprisingly,Myc, a known oncogenein T-ALL (7), was not among the strongest positive outliers onChr15. In future work it should be possible to use Mps1DK-drivenaneuploidy to identify new oncogenes or tumor suppressors in-volved in T-ALL as well as genes involved in cell survival inthe presence of CIN.To begin to identify biological pathways altered by aneuploidy,

we compared probes that were up- or down-regulated at least1.5-fold in >15% of tumors analyzed (4 of 22); we identified∼3,300 genes by this analysis. We then used WebGestalt (35) tofind Gene Ontogeny (GO) categories that were significantlyenriched (using a Bonferroni corrected P value < 0.05). Themost commonly deregulated pathways were TCR signaling,mRNA processing, cell cycle, and pathways involved in cellular

0

5

-5

Tumor 15

Mps1f/f p53f/+ Lck-Cre+

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

X

Y

0

5

-5

Tumor 20

Mps1f/f p53f/f Lck-Cre+

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

X

Y

0

5

-5

Tumor 38

p53f/f Lck-Cre+

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

X

Y

0

5

-5

Tumor 36

p53f/f Lck-Cre+

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

X

Y

A

1314

12

1920

1017

24

21

1849

16

23

485250

47

46

74443

1525

37

366338

1 5 2, var. 3 6 107 81913 174 14 159 11 12 16X 18

Mps1 p53 tumorsp53 tumors

Hierarchically clustered CGH probes

Hie

rach

ical

ly c

lust

ered

tum

ors

39

22

11

24

4535

-1.0 1.01:1

log2 CGH ratio of tumor DNA over reference liver DNA

C

B

Ane

uplo

idy

inde

x

p53f/f Mps1f/f

p53f/fMps1f/f

p53f/+

0

5

10

15

20

p=0.0047

p=0.0278

Mps1f/f Lck-Cre+

Mps1f/f Lck-Cre-

Mps1f/f p53f/+ Lck-Cre+

DChr

Fig. 3. Mps1 truncation leads to CIN and clonally stable karyotypes intumors. (A) Representative aCGH profiles for four tumors showing whole-chromosome instability. (B) Aneuploidy index (total number of wholechromosomes gained and lost as assessed by aCGH) for tumors with theindicated genotypes. (C) Single-linkage hierarchical cluster analysis for in-dividual tumors (top to bottom) and CGH probes (left to right). Clusteringwas separated visually into 20 clear groups. (D) Representative interphaseFISH images of control (Top), Mps1DK T cells (Middle), andMps1DK T-ALL cells(Bottom) showing copy numbers for Chr15 (green) and Chr17 (red).

13430 | www.pnas.org/cgi/doi/10.1073/pnas.1400892111 Foijer et al.

metabolism (Fig. S3E). When we performed hierarchical clus-tering of deregulated genes (Fig. 5), dividing clusters into fivegroups based on whether genes were strongly or weakly up- ordown-regulated, T-cell differentiation and signaling factors weredown-regulated; this result is consistent with histological datashowing that tumors are poorly differentiated. In contrast, path-ways involved in cellular metabolisms (GO terms for “metabolicpathways,” “RNA metabolic pathways,” “spliceosome,” “trans-lation factors,” “nucleotide synthesis,” and others) were up-regu-lated. Misregulation of these processes previously has beenassociated with aneuploid stress in cultured mammalian cells andbudding yeast (6, 8, 36, 37). We conclude that dysregulation ofmetabolic pathways is a common feature of CIN in multipleorganisms and cell types, including rapidly growing tumors.

DiscussionIn this article we report the development and analysis of mice inwhich CIN and consequent aneuploidy are induced in a rapidlyproliferating but nonessential adult tissue by conditionally trun-cating the SAC kinase Mps1. We show that the Mps1 truncation,which deletes a kinetochore-targeting domain but leaves kinaseactivity and other functions intact, can provoke but not sustaina checkpoint arrest in the presence of spindle poison; it also

causes chromosome misalignment and generates lagging chro-mosomes, consistent with the dual role of SAC proteins in pro-moting and sensing kinetochore attachment. When mutated inmurine T cells, Mps1 truncation causes aneuploidy, but this aneu-ploidy is insufficient for efficient oncogenic transformation. How-ever, in a predisposed p53 heterozygous-deletion background, Mps1truncation results in rapidly growing T-ALL. We observe fre-quent p53 LOH in Mps1f/f p53f/+ Lck-Cre+ animals, consistentwith the known role of p53 as a potent tumor suppressor in Tcells. However, p53 LOH cannot be the only oncogenic event inT-ALLs, because tumor latency is significantly shorter in Mps1f/f

p53f/+ Lck-Cre+ animals than in p53f/f Lck-Cre+ animals. aCGHalso reveals higher levels of aneuploidy in compound-mutantanimals than in single mutants. Taken together, these data sug-gest that Mps1 mutation causes p53 LOH and that the combinedMps1-p53 mutant genotype results in more efficient gain and lossof cancer genes than p53 mutation alone.We interpret aCGH and interphase FISH data to show that

T-ALLs are initially clonal (based on TCR rearrangement) butthat ongoing CIN results in tumors with recurrent aneuploidies inChr4, Chr14, Chr15, and Chr19 as well as sporadic changes inother chromosomes. The net result is changes in the expression ofoncogenes and tumor suppressors driving T-ALL as well aschanges in as yet unidentified pathways involved in tolerizing cellsto aneuploidy. A significant body of literature has emerged overthe past few years examining the consequences of aneuploidy forthe physiology of yeast and cultured murine fibroblasts. Thesestudies have revealed recurrent up-regulation of pathways involvedin mRNA processing and other metabolic processes (6, 8, 37). Ourdata show that T-ALLs generated by Mps1 truncation and p53loss also exhibit transcriptional signatures similar to those ofaneuploid MEFs (6) and untransformed tissues (36). Thus,T-ALLs cells can grow rapidly in the face of frequent chromo-some loss and gain, but CIN nonetheless imposes a burden ontumor cells similar to that observed in MEFs and budding yeastcells. This finding is potentially significant, because proteotoxicstress and metabolic dysregulation have become important tar-gets for cancer therapy and may represent a means to kill highlyaneuploid cancers selectively.

Materials and MethodsAnalysis of Mice. Mice used in this study had a mixed C57BL/6 and 129/Sv D3genetic background.Mps1 conditional mice harboring a deletion of residues47–154 were generated as described in SI Materials and Methods. p53 con-ditional-knockout mice were obtained from Anton Berns (Department of

Average aCGH signal per chromosome in individual tumorsA

B

0

1.0

0.5

-0.5

-1.0

p53f/f tumors

Ave

rage

log2

sig

nal

0

12

34

56

78

910

1112

1314

1516

1718

19

1.0

0.5

-0.5

-1.0

Mps1f/f p53f/f / Mps1f/f p53f/+ tumors

Ave

rage

log2

sig

nal

Average RNA expression array signal per chromosome in individual tumors

0

1.0

0.5

-0.5

-1.0

Mps1f/f p53f/f / Mps1f/f p53f/+ tumors

Ave

rage

log2

sig

nal

0

1.0

0.5

-0.5

-1.0

Ave

rage

log2

sig

nal

Chr. 15R2= 0.44

Chr. 4R2= 0.77

Chr. 14R2= 0.63

Ave

rage

log2

sig

nal

expr

essi

on a

rray

Average log2 signal aCGH array

Correlation copy number alterations and expression changesC

-0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8 -0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8 -0.2

0.0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8

p53f/f tumors

12

34

56

78

910

1112

1314

1516

1718

19

12

34

56

78

910

1112

1314

1516

1718

19 12

34

56

78

910

1112

1314

1516

1718

19

Fig. 4. Gene copy number results in proportional transcription changes inaneuploid tumors. (A and B) Average aCGH (A) or RNA expression values (B)were calculated per chromosome for each tumor and plotted for each tumorgroup. Each individual symbol represents the average value of that chro-mosome in one tumor; black crosses show the mean and SEM across alltumors per chromosome. (C) Linear regression plots showing the correlationstrength (coefficient of correlation, R2) between copy number changes(aCGH) and expression changes (expression arrays) for frequently gainedChr4, Chr14, and Chr15.

45 44 42 13 11 10 7 38 63 46 47 18 19 20 21 22 23 48 49 50 52 36I:Macrophage markers, antigen processing and presentation,T-cell proliferation, T-cell differentiation

II: T-cell receptor signaling, chemokinesignaling, interferon signaling, MAPK signaling, T-cell differentiation

III: mRNA processing, Amico acid metabolism,Metabolic pathways, Purine/ pyrimidine biosynthesis,Ribosome biogenesis, RNA transport

IV: Spliceosome, Proteasome,mRNAprocessing, Metabolic pathways

V: RNA transport, Translation factors, Metabolic processes

TID

-1.0 1.01:1 log2 ratio of tumor RNA expression over RNA expression from non-transformed proliferating T-cells

Fig. 5. Mps1DK-driven CIN activates genes regulating various aspects ofcellular metabolism. Single-linkage hierarchical clustering of transcriptomedata shows the most significant deregulated pathways per group. GroupsI–V were grouped by visual inspection. Full gene lists and enrichment analysesare in Dataset S4.

Foijer et al. PNAS | September 16, 2014 | vol. 111 | no. 37 | 13431

GEN

ETICS

Genetics, The Netherlands Cancer Institute, Amsterdam, The Netherlands)(25), Lck-Cre transgenic mice from Taconic (38), and MMTV-Cre mice (39)from the MIT mouse repository. Mice were intercrossed to obtain the de-scribed strains and genotyped as described previously (25, 39). (GenotypingPCR primers are given in Dataset S5.) For survival studies, mice were moni-tored for tumor development weekly starting at 2.5 mo of age by lookingfor difficulty in breathing (a consequence of thymic hypertrophy). Tissueswere fixed in 10% formalin and then paraffin-embedded for histology.Animal protocols were approved by the MIT, Harvard Medical School, andUniversity of Groningen Medical Center Committees on Animal Care and bythe U.K. Home Office.

MEF Isolation, Transduction, Flow Cytometry, and Time-Lapse Analysis. MEFswere isolated as described previously (40) at E13.5 from Mps1f/f embryos andwere cultured under low-oxygen (3%) conditions. Retroviral particles (pRetroxsystem; Clontech) were produced in Lipofectamine 2000-transfected (Life Tech-nologies) 293T cells.MEFswere first transducedwith pRetrox-rtTAvirus andnextwith pRetrox-GFP-T2A-Cre virus, allowing us to assess Cre expression withoutlinking Cre toGFPdirectly (41), orwith pRetroxGFP-Mps1WT orGFP-Mps1DK virus(constructs are described in detail in SI Materials and Methods). For time-lapseimaging, cells subsequently were transduced with pRetrox-H2B-Cherry or pRe-trox-H2B-GFP to visualize theDNAandwere treatedwith 250ng/mL nocodazole(Sigma) where indicated. For flow cytometry, transduced cells were treatedwith1 μg/mL Dox (Sigma) for 48–72 h to induce the retroviral inserts, then weretreatedwith 250ng/mL nocodazole for 6 h, andwere fixed in 70%ethanol. Cellsthen were stained with phycoerythrin-labeled pHistoneH3 antibody (Cell Sig-naling) and FxCycle Brilliant blue (Life Technologies) and analyzed on an LSRIIanalyzer (BD Biosciences). Data were analyzed using FlowJo software. For time-lapse analysis, cells were seeded on LabTek imaging chambers (Thermo Fisher),induced with 1 μg/mL Dox for 48–72 h, and imaged for up to 36 h on a Delta-Vision Elite imaging station (Applied Precision, GE Healthcare) in a low-oxygen

imaging chamber (Oko Laboratories). Movies were deconvolved and analyzedusing SoftWorx suite (GE Healthcare).

DNA/RNA/Protein Isolation, Array, and Blotting Experiments. Genomic DNAfrom tumor samples and control liver was isolated using genomic DNA tissuekits (Qiagen). Genotyping primers are listed inDataset S5. Proteinwas isolatedusing a total protein extraction kit (Millipore). For protein detection, proteinswere blotted using standardWestern blotting protocols. The antibodies usedwere Mps1 (TTK-C19, Santa Cruz) and HRP-conjugated goat anti-rabbit (CellSignaling). For aCGH experiments, labeled DNA (SI Materials and Methods)was hybridized to 244Kmouse genome CGH arrays (Agilent) according to themanufacturer’s protocol and was analyzed as described in SI Materials andMethods. RNA was isolated from tumor samples and from the thymuses from6-wk-old control mice using the RNeasy kit (Qiagen). Labeled RNA (SIMaterials and Methods) was hybridized to Illumina v6.2 BeadChip arrays andwas scanned and analyzed as described in SI Materials and Methods. qPCRprimers are listed in Dataset S5. All array data have been deposited in theNational Center for Biotechnology Information Gene Expression Omnibusdatabase under accession no. GSE57334.

ACKNOWLEDGMENTS. We thank members of the P.K.S. laboratory, S. W. M.Bruggeman, L. Albacker, and L. Kleiman for critically reading the manuscriptand fruitful discussions; B. Bakker for assistance with cloning; RoderickBronson for help with pathology; and A. Burds for sharing Lck-Cre andMMTV-Cre mice. This work was supported by Dutch Cancer Society GrantRUG 2012-5549; by grants from Stichting Kinder Oncologie Groningen andEuropean Molecular Biology Organization (to F.F.); by National Institutes ofHealth (NIH) Grants CA084179 and CA139980 (to P.K.S.); by NIH Grant P30-CA14051 to the MIT mouse transgenic facility; and by the Wellcome Trust(F.F. and A.B.).

1. Duijf PH, Schultz N, Benezra R (2013) Cancer cells preferentially lose small chromo-somes. Int J Cancer 132(10):2316–2326.

2. Schvartzman JM, Sotillo R, Benezra R (2010) Mitotic chromosomal instability andcancer: Mouse modelling of the human disease. Nat Rev Cancer 10(2):102–115.

3. Holland AJ, Cleveland DW (2009) Boveri revisited: Chromosomal instability, aneu-ploidy and tumorigenesis. Nat Rev Mol Cell Biol 10(7):478–487.

4. Foijer F, Draviam VM, Sorger PK (2008) Studying chromosome instability in the mouse.Biochim Biophys Acta 1786(1):73–82.

5. Baker DJ, Jin F, Jeganathan KB, van Deursen JM (2009) Whole chromosome instabilitycaused by Bub1 insufficiency drives tumorigenesis through tumor suppressor geneloss of heterozygosity. Cancer Cell 16(6):475–486.

6. Williams BR, et al. (2008) Aneuploidy affects proliferation and spontaneous immor-talization in mammalian cells. Science 322(5902):703–709.

7. Jones L, et al. (2010) Gain of MYC underlies recurrent trisomy of the MYC chromo-some in acute promyelocytic leukemia. J Exp Med 207(12):2581–2594.

8. Torres EM, et al. (2007) Effects of aneuploidy on cellular physiology and cell division inhaploid yeast. Science 317(5840):916–924.

9. Kops GJ, Foltz DR, Cleveland DW (2004) Lethality to human cancer cells throughmassive chromosome loss by inhibition of the mitotic checkpoint. Proc Natl Acad SciUSA 101(23):8699–8704.

10. Mao R, Zielke CL, Zielke HR, Pevsner J (2003) Global up-regulation of chromosome 21gene expression in the developing Down syndrome brain. Genomics 81(5):457–467.

11. Baker DJ, et al. (2004) BubR1 insufficiency causes early onset of aging-associatedphenotypes and infertility in mice. Nat Genet 36(7):744–749.

12. Baker DJ, et al. (2006) Early aging-associated phenotypes in Bub3/Rae1 hap-loinsufficient mice. J Cell Biol 172(4):529–540.

13. Baker DJ, et al. (2013) Increased expression of BubR1 protects against aneuploidy andcancer and extends healthy lifespan. Nat Cell Biol 15(1):96–102.

14. Jallepalli PV, Lengauer C (2001) Chromosome segregation and cancer: Cuttingthrough the mystery. Nat Rev Cancer 1(2):109–117.

15. Taylor SS, Scott MI, Holland AJ (2004) The spindle checkpoint: A quality controlmechanism which ensures accurate chromosome segregation. Chromosome Res 12(6):599–616.

16. Rieder CL, Cole RW, Khodjakov A, Sluder G (1995) The checkpoint delaying anaphasein response to chromosome monoorientation is mediated by an inhibitory signalproduced by unattached kinetochores. J Cell Biol 130(4):941–948.

17. Musacchio A, Salmon ED (2007) The spindle-assembly checkpoint in space and time.Nat Rev Mol Cell Biol 8(5):379–393.

18. Weaver BA, Silk AD, Montagna C, Verdier-Pinard P, Cleveland DW (2007) Aneuploidyacts both oncogenically and as a tumor suppressor. Cancer Cell 11(1):25–36.

19. Jonkers J, Berns A (2002) Conditional mouse models of sporadic cancer. Nat RevCancer 2(4):251–265.

20. Janssen A, Kops GJ, Medema RH (2009) Elevating the frequency of chromosome mis-segregation as a strategy to kill tumor cells. Proc Natl Acad Sci USA 106(45):19108–19113.

21. Hached K, et al. (2011) Mps1 at kinetochores is essential for female mouse meiosis I.Development 138(11):2261–2271.

22. Nijenhuis W, et al. (2013) A TPR domain-containing N-terminal module of MPS1 isrequired for its kinetochore localization by Aurora B. J Cell Biol 201(2):217–231.

23. Rudner AD, Murray AW (1996) The spindle assembly checkpoint. Curr Opin Cell Biol8(6):773–780.

24. Hennet T, Hagen FK, Tabak LA, Marth JD (1995) T-cell-specific deletion of a poly-peptide N-acetylgalactosaminyl-transferase gene by site-directed recombination. ProcNatl Acad Sci USA 92(26):12070–12074.

25. Jonkers J, et al. (2001) Synergistic tumor suppressor activity of BRCA2 and p53 ina conditional mouse model for breast cancer. Nat Genet 29(4):418–425.

26. Burds AA, Lutum AS, Sorger PK (2005) Generating chromosome instability through thesimultaneous deletion of Mad2 and p53. Proc Natl Acad Sci USA 102(32):11296–11301.

27. Thompson SL, Compton DA (2010) Proliferation of aneuploid human cells is limited bya p53-dependent mechanism. J Cell Biol 188(3):369–381.

28. Fujiwara T, et al. (2005) Cytokinesis failure generating tetraploids promotes tumori-genesis in p53-null cells. Nature 437(7061):1043–1047.

29. Donehower LA, et al. (1992) Mice deficient for p53 are developmentally normal butsusceptible to spontaneous tumours. Nature 356(6366):215–221.

30. Purdie CA, et al. (1994) Tumour incidence, spectrum and ploidy in mice with a largedeletion in the p53 gene. Oncogene 9(2):603–609.

31. Jacks T, et al. (1994) Tumor spectrum analysis in p53-mutant mice. Curr Biol 4(1):1–7.32. Wagner KU, et al. (2001) Spatial and temporal expression of the Cre gene under the

control of the MMTV-LTR in different lines of transgenic mice. Transgenic Res 10(6):545–553.

33. Zha S, et al. (2010) ATM-deficient thymic lymphoma is associated with aberrant tcrdrearrangement and gene amplification. J Exp Med 207(7):1369–1380.

34. Takabatake T, et al. (2008) Analysis of changes in DNA copy number in radiation-induced thymic lymphomas of susceptible C57BL/6, resistant C3H and hybrid F1 Mice.Radiat Res 169(4):426–436.

35. Zhang B, Kirov S, Snoddy J (2005) WebGestalt: An integrated system for exploring genesets in various biological contexts. Nucleic Acids Res 33(Web Server issue):W741–748.

36. Foijer F, et al. (2013) Spindle checkpoint deficiency is tolerated by murine epidermalcells but not hair follicle stem cells. Proc Natl Acad Sci USA 110(8):2928–2933.

37. Sheltzer JM, Torres EM, Dunham MJ, Amon A (2012) Transcriptional consequences ofaneuploidy. Proc Natl Acad Sci USA 109(31):12644–12649.

38. Lee PP, et al. (2001) A critical role for Dnmt1 and DNA methylation in T cell de-velopment, function, and survival. Immunity 15(5):763–774.

39. Wagner KU, et al. (1997) Cre-mediated gene deletion in the mammary gland. NucleicAcids Res 25(21):4323–4330.

40. Foijer F, Wolthuis RM, Doodeman V, Medema RH, te Riele H (2005) Mitogen re-quirement for cell cycle progression in the absence of pocket protein activity. CancerCell 8(6):455–466.

41. Szymczak AL, et al. (2004) Correction of multi-gene deficiency in vivo using a single‘self-cleaving’ 2A peptide-based retroviral vector. Nat Biotechnol 22(5):589–594.

13432 | www.pnas.org/cgi/doi/10.1073/pnas.1400892111 Foijer et al.

Supporting InformationFoijer et al. 10.1073/pnas.1400892111SI Materials and MethodsGeneration of Mps1f Conditional and Mps1Δ Mutant Mice. A 129/Svmouse genomic λ-phage library was screened for the genomiclocus of monopolar spindle 1 (Mps1) as described (1). A 14-kbportion of the Mps1 genomic locus encoding exons 1–9 wasisolated and cloned into pBluescript (Life Technologies) viaNotI sites. DNA encoding diphtheria toxin A was cloned into theNotI site as a negative selection marker. A loxP site was clonedinto an NcoI site between exons 2 and 3. A 4-kb selection cas-sette containing thymidine-kinase and the neomycin-resistancegene flanked by two loxP and FRT recombination sites (LFNTcassette) was cloned into a SphI site between exons 4 and 5. Theresulting 22.6-kb targeting vector (Fig. S1A, line 2) was in-troduced into ES cells, and clones were isolated as previouslydescribed (1). Incorporation of the gene-targeting vector byhomologous recombination into ES clones was determined bySouthern blotting as described (1), using a 664-bp 5′ probe anda 665-bp 3′ probe after digestion with NheI or BglI as indicatedin Fig. S1A, line 3. The 5′ probe was generated by digesting a λ-phageclone with SacI and XmaI which contained sequences 5′ of thetargeting vector and then was subcloned into pBluescript (LifeTechnologies). The 5′ probe yields an 11.7-kb wild-type bandand a 9-kb targeted band. The 3′ probe is homologous to se-quences containing exon 9 of Mps1 and was generated by PCRyielding a 13.4-kb wild-type band and a 9.4-kb targeted band.Two correctly targeted ES clones were isolated. The LFNTcassette was removed in these ES clones by transfecting cellswith vectors expressing Cre recombinase (pCrePac) or flippaserecombinase (pFlpe), resulting in the kinetochore-binding do-main deletion (DK) (Fig. S1A, line 5) or conditional (f) (Fig.S1A, line 4) allele, respectively. Southern blotting after re-striction/digestion with NcoI and probing with the 5′ 664-bpprobe yielded 8 kb for wild type, 12 kb for the targeted locus(containing LFNT), 15.2 kb for the DK allele, and 18.6 kb for thef-allele. Three DK+ and two f+ subclones from the two originalclones were injected into blastocysts and successfully generatedfounder lines for the conventional DK and conditional alleles.To assess the consequences of Mps1 truncation in developing

embryos, we generated Mps1Δ/+ ES cells by Cre electroporation,established the Mps1DK allele in the germ line, and intercrossedMps1DK/+ mice. No Mps1DK/DK mice were recovered in 33 litterscomprising 237 pups, showing that the Mps1 truncation allele isembryonic lethal (Mps1+/+ mice comprised 38% of viable pups,and Mps1DK/+ mice comprised 62% of pups, the expected Men-delian ratio) (Fig. S5B). At embryonic day 10.5 (E10.5), embryoswere isolated, and the yolk sac was taken to determine genotypeby PCR as described (2). Mps1DK/DK embryos died at or beforeE10 (Fig. 5C), as previously described with the germ-line deletionof other checkpoint genes (3–5). We conclude that aneuploidygenerated by the Mps1DK allele is lethal to developing embryos.

Generation of pRetrox GFP-Mps1 Full-Length, pRetrox GFP-Mps1DK,and Cre-T2A-GFP. For pRetrox GFP-Mps1 vectors, we first PCRamplified EGFP (Phusion polymerase; Thermo Scientific) withBglII and BamHI+NotI Sites 5′ and 3′ respectively. This PCRfragment was cloned into BamH1-, and NotI- (New EnglandBiolabs) digested pRetrox backbone (Clontech), thus destroyingthe BamH1 site 5′ of the GFP. In the case of full-length Mps1,we then PCR-amplified Mps1 from a human Mps1 cDNA clone(DNA Resource Core, Dana-Farber/Harvard Cancer Center)flanked by BamH1 and NotI sites 5′ and 3′ respectively, whichwas cloned into pRetrox-eGFP digested with BamH1 and NotI.

For Mps1DK, we first PCR-amplified a fragment encompassing the5′ ATG to an endogenous HindII site in exon 1 of human Mps1,which was ligated into a pMSCV backbone (Clontech), harboringthe mouse stem cell virus (MSCV) LTRs, digested with BamH1and HindIII. We next PCR-amplified Mps1 from the start of exon4 to its stop codon and added the remaining exon 1 sequencedownstream of the HindIII site to the 5′ end of this product, re-sulting in a HindIII–NotI fragment. This fragment was ligated intothe product of the previous step, thus creating Mps1DK. We thenPCR-amplified Mps1DK from pMSCV Mps1DK using BglII andNotI sites and cloned it into pRetrox eGFP as we did for full-length Mps1. For Cre-T2A-GFP, we first PCR amplified GFP,adding a T2A sequence flanked by BglII and BamH1-NotI sitesto its 3′ end (6), and cloned it into pRetrox using BamH1 andNotI sites. We next cloned Cre downstream of GFP-T2A usingBamH1 and NotI sites 5′ and 3′, respectively. All primers usedare listed in Dataset S5.

Quantification of Kinetochore-Bound Mps1. To quantify the levelsof GFP-Mps1 bound to kinetochores, fluorescence arbitraryunits on a horizontal line through a kinetochore were analyzed inSoftWorx software (GE Healthcare) using the line-profiling tool.To correct for GFP-tagged protein expression levels, we nor-malized kinetochore-bound values by subtracting the averageGFP expression counts in the cell from the counts at the kinet-ochore (line-profiling tool). We then calculated the fluorescenceratios between GFP-Mps1 and GFP-Mps1DK for both cell types(Fig. S1 E and G).

Comparative Genome Hybridization and Analysis. For array-basedcomparative genomic hybridization (aCGH) experiments DNAwas labeled with Cy3 or Cy5 fluorescent nucleotides according tothe BioPrime array CGH genomic labeling protocol (Life Tech-nologies) and cleaned using the PureLink PCR purification kit(Life Technologies).Many liver samples showed evidence of T-cellinfiltration, disqualifying them as control samples. Therefore weused pooled sex-matched reference samples from uninfiltratedlivers from littermates. Labeled DNA was hybridized to 244Kmouse genome CGH microarrays (Agilent) according to themanufacturer’s protocol and subsequently was scanned, back-ground corrected, and normalized using the loess algorithm withBioconductor limma package (7). Data then were segmented andsplit into regions of estimated equal copy number using a circularbinary segmentation algorithm (Bioconductor) with the DNA-copy package (8). To reduce noise further, we used the “undo”method to remove all splits that are not at least 3 SDs apart. Log2ratios of tumor to/reference were plotted along the chromosomecoordinates using the same software. To create a multisamplematrix, we assigned segment means to mouse genes [NationalCenter for Biotechnology Information (NCBI) M37 and M63] onall segments for each sample using the Bioconductor CNToolspackage. Data were processed further and sorted by chromosomalposition in Microsoft Excel. The aneuploidy score was calculatedby assessing whole-chromosome losses and gains in Agilent Ge-nomic Workbench software, and statistical analyses were per-formed using GraphPad Prism software. To find chromosomeregions showing common gains/losses, we computed segment gainor loss (SGOL) scores by calculating all positive and negative log2ratio values over or under a log2 ratio threshold (j0.3j) using theBioconductor cghMCR package (9).

Foijer et al. www.pnas.org/cgi/content/short/1400892111 1 of 13

RT-PCR, Quantitataive PCR, and Expression Arrays. Synthesis ofbiotin-labeled cRNA was performed using an Illumina Total-Prep RNA amplification kit (Ambion). Amplified biotinylatedcRNA (1.5 μg) was hybridized to llumina Sentrix BeadChips(v6.2) and subsequently labeled with streptavidin-conjugatedCy3 (Amersham) according to the manufacturer’s protocol.After scanning, data were quantile normalized (10) and ana-lyzed using the Bioconductor lumi (11) and limma packages(7). Data were adjusted by P value to yield a sorted list ofdifferentially expression genes (12) and were sorted by chro-mosomal position or frequency of deregulation in MicrosoftExcel. For Gene Ontogeny analysis, all genes with log2 valuesbetween −0.6 and 0.6 (less than 1.5-fold deregulated) andgenes deregulated in fewer than four tumors were excluded.For RT-PCR, reverse transcription was performed using an

OligoDT, random hexamer mixture and Molony murine leu-kemia virus-RT (New England Biolabs) according to themanufacturer’s instructions. Primer sequences can be foundin Dataset S5. For quantitative PCR (qPCR) reactions, 1 μgof total RNA was used for a reverse transcriptase reaction(Superscript II; Life Technologies). The resulting cDNA wasused as a template for qPCR (ABI PRISM 7700 SequenceDetector) in the presence of SYBR-Green (Life Technolo-gies) to label the product. The relative amounts of cDNAwere compared with actin to correct for the amount of totalcDNA. Average values and SDs were calculated as indicatedin figure legends and were compared with the expressionvalues in control mice (normalized to the value of 1). Primersequences are listed in Dataset S5.

1. Dobles M, Liberal V, Scott ML, Benezra R, Sorger PK (2000) Chromosomemissegregationand apoptosis in mice lacking the mitotic checkpoint protein Mad2. Cell 101(6):635–645.

2. Burds AA, Lutum AS, Sorger PK (2005) Generating chromosome instability throughthe simultaneous deletion of Mad2 and p53. Proc Natl Acad Sci USA 102(32):11296–11301.

3. Foijer F, Draviam VM, Sorger PK (2008) Studying chromosome instability in the mouse.Biochim Biophys Acta 1786(1):73–82.

4. Schvartzman JM, Sotillo R, Benezra R (2010) Mitotic chromosomal instability andcancer: Mouse modelling of the human disease. Nat Rev Cancer 10(2):102–115.

5. Holland AJ, Cleveland DW (2009) Boveri revisited: Chromosomal instability,aneuploidy and tumorigenesis. Nat Rev Mol Cell Biol 10(7):478–487.

6. Szymczak AL, et al. (2004) Correction of multi-gene deficiency in vivo using a single‘self-cleaving’ 2A peptide-based retroviral vector. Nat Biotechnol 22(5):589–594.

7. Smyth GK (2005) Limma: Linear Models for Microarray Data (Springer, New York), pp397–420.

8. Venkatraman ES, Olshen AB (2007) A faster circular binary segmentation algorithmfor the analysis of array CGH data. Bioinformatics 23(6):657–663.

9. Aguirre AJ, et al. (2004) High-resolution characterization of the pancreatic adenocarcinomagenome. Proc Natl Acad Sci USA 101(24):9067–9072.

10. Yang YH, et al. (2002) Normalization for cDNA microarray data: A robust compositemethod addressing single and multiple slide systematic variation. Nucleic Acids Res30(4):e15.

11. Du P, Kibbe WA, Lin SM (2008) lumi: A pipeline for processing Illumina microarray.Bioinformatics 24(13):1547–1548.

12. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical andpowerful approach to multiple testing. J R Stat Soc, B 57(1):289–300.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 2 of 13

wt, 505 bp

DK, 284bp

Mps1WT Mps1DK/+

C

NotI 1 2 3 4 5 67

89

NcoI NheI

NheI NotI

NheISphI

1

NotI 1 2 3 4 5 67

89

NcoI NheI

NheI NotIHSV-TK PGK-NeoR

loxP loxP FRTFRT

NheI

NheI

DT-AloxP

SphISphI

2

NotI 1 2 3 4 5 67

89NcoI

NheI NotIHSV-TK PGK-NeoR

loxP loxP FRTFRT

NheI NheINheI

loxP

SphISphI

3

NotI 1 2 3 4 5 67

89NcoI

NheI NotI

loxP FRT

NheI

loxP

SphISphINheI NheI

4

NotI 1 2 5 67

89NcoI

NheI NotI

FRT

NheI

loxP

SphINheI

5

1: Mps1 genomic locus2: Targeting vector3: Targeted Mps1 locus4: Targeted locus following Flipase (Mps1f)

)5: Targeted locus following Cre (Mps1DK

A

B

wt

DK allele

flox allele

DK/+ DK/+ f/+f/+

18.6 kb

15.2 kb

8 kb

Mps1f/f MEFs (99)

Mps1f/f MEFs + CreDox (82)

0 50 100 150 250Time in mitosis (min)

200

Indi

vidu

al c

ells

0:00 6:00

6:00

12:00

12:00

18:00

18:00

24:00

24:00

42:00

42:00

Mps

1WT

Mps

1DK

0:0011μm

E

D

Mps1WT-GFP

Mps1DK-GFP

MCF 10a cells

F Mps1WT-GFP

MCF 10a cells

Mps1DK-GFP

Mps1WT-GFP

MEFs

HG

100%

36%

100%

42%

Mps1DK-GFP

Fig. S1. Engineering a conditional Mps1 allele showing severely decreased kinetochore binding in mitosis. (A) Schematic overview of the Mps1 locus, theconditional targeting vector, and targeted locus before and after Flp and Cre-recombinase expression. (B) Southern blot on tail DNA from founder miceharboring conditional (f) and DK alleles demonstrating germ-line transmission. (C) RT-PCR for Mps1 on mRNA isolated from wild-type and Mps1f/+; Lck-Cre+

thymocytes showing expression of the Mps1DK hypomorphic protein. (D) Time-lapse image stills showing kinetochore localization of retrovirally transducedwild-type GFP-Mps1 during mitosis (Upper) and less binding of GFP-Mps1DK to kinetochores (Lower). DNA was labeled with retroviral H2B-Cherry. Also seeMovies S1 and S2. (E) Quantification on time-lapse image stills (Movies S1 and S2) of kinetochore-bound Mps1 in cells expressing either wild-type GFP-Mps1(Upper) or GFP-Mps1DK (Lower) when expressed in mouse embryonic fibroblasts (MEFs). Indicated percentages refer to the average GFP signal on three ki-netochores within one cell and compared between cells expressing wild-type GFP-Mps1 and GFP- Mps1DK. The dashed red line indicates where the fluorescencesignal was quantified. (F) Time-lapse image stills showing strong kinetochore localization of retrovirally transduced wild-type GFP-Mps1 in pro-metaphase(Upper) and weak kinetochore localization of GFP-Mps1DK (Lower) in human MCF 10A cells. DNA was labeled with retroviral H2B-Cherry. See also Movies S3and S4. (G) Quantification on time-lapse image stills (Movies S3 and S4) of kinetochore-bound Mps1 in cells expressing either wild-type GFP-Mps1 (Upper) orGFP-Mps1DK (Lower) when expressed in MCF 10A cells. The dashed red line indicates where the fluorescence signal was quantified. (H) Distribution of mitoticdurations for control-infected Mps1f/f (blue) and Cre-infected Mps1f/f (red) MEFs. Average values are shown in Fig. 1F.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 3 of 13

B

F

+ + + +- - - Lck-Cre

Mps1f/f thymus samplesMps1W

T

Gen

omic

P

CR

G

C

Mps1f/f p53f/f and Mps1f/f p53f/+ tumors1 2 3 4 5 6 7

Mps1fMps1DK

Mps1WT

Mps1fMps1DK

Mps1WT

8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

25 26 39 40 41 42 43 44 45 46 47 48 49 50 52 53 54 55

Genomic PCR

A

p53Δp53f

p53WT

9 10 11 12 13 14 15 16 17 18 19 20 21p53f/+

Mps1f/f p53f/+ tumors Mps1f/f p53f/f tumors

DLog2 ratio

-5.0 1:1 5.0

Mps1

f/f

p53f/+ C

re+

Trp53 locus7

10111213141516173942434445464754

TID

Mps1

f/f

p53f/f C

re+

Mps1

f/f

Cre

+p53

f/f

Cre

+

Trp53 locus

TID

181920212223242548495052

265355

3637386364

pRetr

ox M

ps1W

T

pRetr

ox M

ps1D

K

36 63 13 23 36 63 13 23

500+ 501 500+ 503Primers

TID M M

3kb

1kb

500b

Age (months)

Per

cent

su r

vivi

ng

0 4 8 120

20

40

60

80

100 Mps1f/+ p53f/f Lck-Cre+ (n=18)p53f/f Lck-Cre+ (n=23, see also Fig 2C)

0

20

40

60

80

100

Age (months)

Per

cent

sur

vivi

ng

0 4 8 12

Mps1f/f p53f/+ MMTV-Cre+ (n=15)

p53f/f MMTV-Cre+ (n=9)

Mps1f/f p53f/f MMTV-Cre+ (n=8)

Mps1f/f MMTV-Cre+ (n=7)

I

E

Genomic PCR

Genotype Chr. 17 # Cells Chr. 15 # CellsMps1f/f Lck-Cre- 1% 161 4% 174Mps1f/f Lck-Cre- 6% 153 6% 156Mps1f/f Lck-Cre- 10% 120 5% 145Mps1f/f Lck-Cre+ 5% 210 5% 196Mps1f/f Lck-Cre+ 18% 211 17% 222

Mps1f/f p53f/+ Lck-Cre+ 9% 141 81% 163Mps1f/f p53f/+ Lck-Cre+ 4% 71 74% 182Mps1f/f p53f/+ Lck-Cre+ 15% 180 17% 184Mps1f/f p53f/+ Lck-Cre+ 18% 62 37% 152Mps1f/f p53f/+ Lck-Cre+ 15% 176 71% 209Mps1f/f p53f/+ Lck-Cre+ 15% 183 73% 199

J

H

150 bp

1 2520600 1560

Kinetochore-binding domain

Kinasedomain

Deleted in Mps1DK

500

501503

RT-PCR primerqPCR primer

Set A RV

Set A FW

Set B FW

Set B RV

Fig. S2. Mps1 truncation accelerates lymphomagenesis in a p53-deficient background. (A) Genotyping PCR showing efficient switching of Mps1f to Mps1DK inuntransformed T cells. (B) Kaplan–Meier curves showing survival of Mps1f/+ p53f/f and p53f/f mice in an Lck-Cre+ background. Note that the p53f/f survival curveis the same curve presented in Fig. 2D. (C) Genomic PCRs showing full p53 deletion in the majority of the Mps1f/f p53f/f and p53f/f lymphomas and frequent p53loss of heterozygosity (LOH) in Mps1f/f p53f/+ lymphomas. Lane numbers are tumor identification numbers. (D and E) aCGH data showing loss of p53 in tumorswith the indicated genotypes. Each rectangle represents a single aCGH probe value. Three probe values are shown, for the deleted fragment (Center) and forthe 5′ (Left) and 3′ (Right) regions flanking the deleted region. Log2 ratios in the range of −5 indicate complete loss of the probe and thus LOH of the wild-typep53 allele in Mps1f/f p53f/+ tumors (D). The numbers on the left are tumor identification numbers (TID). (F) Kaplan–Meier survival curve showing survival ofMps1f/f, Mps1f/f p53f/f, Mps1f/f p53f/+, and p53f/f mice in a mouse mammary tumor virus (MMTV)-Cre+ background. (G) Genomic PCRs showing efficient excisionof Mps1 in the majority of Mps1f/f p53f/f lymphomas. (H) Schematic overview of Mps1 RNA showing where the qPCR primers used in Fig. 2G and the RT-PCRprimers used in Fig. S2I bind. (I) RT-PCR reactions on RNA isolated from p53-control tumors (TID 36 and 63), Mps1f/f p53f/+ (TID 13), and Mps1f/f p53f/f (TID 23)tumors showing full penetrance of the truncated Mps1DK allele in Mps1DK-driven tumors. Primers 500 and 501 amplify murine Mps1 from the start codon to thestop codon; primers 500 and 503 amplify from the start codon to exon 5. Retroviral vectors encoding constructs for human GFP-Mps1DK (pRetrox Mps1DK) andGFP-Mps1WT (pRetrox Mps1WT) (used in Fig. 1B) serve as PCR controls. All RT-PCR products were Sanger sequenced, which confirmed deletion of the exons 3 and4 in Mps1DK tumors. (See Dataset S2 for information about the alignment the sequencing data.) (J) Quantification of interphase FISH measurements onuntransformed Mps1f/f Lck-Cre+ T cells and Mps1f/f p53f/+ Lck-Cre+ T-ALL cells.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 4 of 13

Hie

rach

ical

ly c

lust

erdd

tum

ors

-1.0

1.0

1:1

log2

CG

H ra

tio o

f tum

or D

NA

ov

er re

fere

nce

liver

DN

A

CGH probes sorted chromosomal location

1314

12

1920

1017

24

21

1849

16

23

485250

47

46

7

4443

1525

37

366338

39

22

11

24

4535

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 X Chr.

romuTDI

-1.0

1.0

1:1

log2

rat

io o

f tum

or R

NA

expr

essi

on o

ver R

NA

expr

essi

on

from

non-

trans

form

ed p

rolif

erat

ing

T-ce

lls

C

Expression probes sorted chromosomal location151 2 3 4 5 6 7 8 9 10 11 12 13 14 16 17 18 19

4544421311107

3863464718192021222348

4950

5236

romuTDI Chr.

Exp

ress

ion

arra

y da

ta s

orte

d ch

rom

osom

al lo

catio

n

D

Pathway Category Observed Expected Enrichment Raw p-value Bonferroni-corrected p Annotated in:

mRNA processing 411 90 53.02 1.7 2.30E-07 3.08E-05 Wiki pathways T Cell Receptor Signaling Pathway 137 36 17.67 2.04 1.83E-05 0.0025 Wiki pathways PluriNetWork 283 60 36.5 1.64 6.09E-05 0.0082 Wiki pathways Translation Factors 44 16 5.68 2.82 6.56E-05 0.0088 Wiki pathways Macrophage markers 9 6 1.16 5.17 0.0003 0.0402 Wiki pathways Antigen processing and presentation 67 27 8.64 3.12 1.85E-08 3.64E-06 KEGG pathwayPrimary immunodeficiency 33 16 4.26 3.76 7.43E-07 0.0001 KEGG pathwayCell cycle 115 34 14.83 2.29 1.88E-06 0.0004 KEGG pathwayAllograft rejection 44 18 5.68 3.17 3.34E-06 0.0007 KEGG pathwayIntestinal immune network for IgA production 41 17 5.29 3.21 5.02E-06 0.001 KEGG pathwayGlioma 59 20 7.61 2.63 2.75E-05 0.0054 KEGG pathwayPhagosome 152 38 19.61 1.94 3.64E-05 0.0072 KEGG pathwayGraft-versus-host disease 43 16 5.55 2.88 4.74E-05 0.0093 KEGG pathwayPhosphatidylinositol signaling system 74 22 9.55 2.3 0.0001 0.0197 KEGG pathwaySystemic lupus erythematosus 70 21 9.03 2.33 0.0001 0.0197 KEGG pathwayType I diabetes mellitus 51 17 6.58 2.58 0.0001 0.0197 KEGG pathwayPurine metabolism 154 37 19.86 1.86 0.0001 0.0197 KEGG pathwayAutoimmune thyroid disease 58 18 7.48 2.41 0.0002 0.0394 KEGG pathwayMetabolic process 8383 1355 1079.64 1.26 4.77E-35 2.22E-31 GO biological processGene expression 3644 665 469.31 1.42 9.19E-26 4.28E-22 GO biological processCellular biosynthetic process 3896 668 501.76 1.33 1.55E-18 7.22E-15 GO biological processBiosynthetic process 4058 690 522.62 1.32 2.49E-18 1.16E-14 GO biological processRNA metabolic process 2984 534 384.31 1.39 3.83E-18 1.78E-14 GO biological processRegulation of cellular metabolic process 3683 622 474.33 1.31 1.41E-15 6.57E-12 GO biological processT cell activation 295 89 37.99 2.34 3.06E-15 1.43E-11 GO biological processResponse to stress 1952 362 251.4 1.44 2.56E-14 1.19E-10 GO biological processRegulation of gene expression 2845 494 366.4 1.35 3.30E-14 1.54E-10 GO biological process

E

A

SG

OL

scor

e -

40 -

200

20- 6

0- 8

013 15

161

23

45

67

89

1011

12 1417

1819

Chromosome

Mps1 p53 tumors

SG

OL

scor

e -

10 -

50

5

Chromosome

851

23

4 67 9

1011

1213

1415

1617

1819

p53 tumorsB

Fig. S3. Common copy-number alterations and correlation between tumor copy-number changes and transcriptomes. (A and B) Cumulative SGOL scores fortumors arising in Mps1 p53 (A) and p53 (B) mice. (C and D) aCGH (C) and transcriptome data (D) sorted by chromosomal position. (E) Overview of pathwaysaffected in aneuploid mouse tumors (Mps1f/f p53f/f and Mps1f/f p53f/+) as analyzed using WebGestalt (1).

1. Zhang B, Kirov S, Snoddy J (2005) WebGestalt: An integrated system for exploring gene sets in various biological contexts. Nucleic Acids Res 33(Web Server issue):W741–748.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 5 of 13

Chr. 6R2 = 0.13915

-0.8 -0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 1 R2 = 0.37249

-0.8 -0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 2R2 = 0.59507

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 3 R2 = 0.48397

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 4 R2 = 0.7676

-0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

Chr.5 R2 = 0.57915

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

Chr. 7R2 = 0.32571

-0.8 -0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 8R2 = 0.61488

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 9R2 = 0.32534

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

Chr. 10R2 = 0.63491

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 11R2 = 0.71297

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 16R2 = 0.6093

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 12R2 = 0.49829 -0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 17R2 = 0.48728

-0.8 -0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 13R2 = 0.10012

-0.8 -0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 18R2 = 0.55305

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

-0.4 -0.2 0 0.2 0.4 0.6 0.8

Chr. 19R2 = 0.40168 -0.8

-0.6

-0.4

-0.2

-0.6 -0.4 -0.2 0Chr. 14R2 = 0.62497

-0.2

0.2

0.4

0.6

0.8

-0.2 0 0.2 0.4 0.6 0.8

Chr. 15R2 = 0.4425

-0.2

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8

Ave

rage

log2

sig

nal e

xpre

ssio

n ar

ray

Average log2 signal aCGH array

Fig. S4. Linear regression plots showing correlation strength (coefficient of correlation, R2) between copy-number changes (aCGH) and expression changes(expression arrays) for all chromosomes and all tumors assessed by both aCGH and transcriptome analysis in this study.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 6 of 13

Genotype % Expected %

Mps1+/+ 38% 25%

Mps1DK/+ 62% 50%

Mps1DK/DK 0% 25%

Live pups (n=237; 33 litters)

B C

total number of embryos=96

%

40%

19%

Genotype

41%

Reabsorbed embryos

Mps1+/+

Mps1

Chr. 4; Add.score: +1677 Chr. 14; Add. score: +1118

Chr. 15; Add.score: +1202 Chr. 19; Add. Score: -1025

Col27a1

Cdkn2aChd7

Cd52Ccl21a

TncOma1

Aldh18a1

Psat1Ms4a

Eif3a

CtswCd5Fas

Fbln1Recql4

Cox6aDrosphaCbx5

Cd6

Hsf1Skp2

Krt18Krt8

Krt7Krt5

Myc

Gln3Serp2AdkPrmt5

Cdhr1Clu

Fezf2Epsti1

Cum

ulat

ive

expr

essi

on2l

og-f

old

chan

gein

tum

orpa

nel

Number of deregulated genes per chromosome

-60

-40

-20

0

20

40

400

800

1200

1600

2000

2400

Srm

-40

-20

0

20

40

200

400

600

800

1000

1200

-80

-60

-40

-20

0

20

40

60

200

400

600

800

1000

1200

1400

-60

-40

-20

0

20

40

200

400

600

800

1000

1200

A

DK/+

Fig. S5. Mps1 truncation deregulates cellular metabolism and results in embryonic lethality. (A) Expression changes in individual genes sorted by cumulativeexpression score (the sum of expression changes within all tumors for that gene, additive or Add. score) sorted high to low for Chr4, Chr14, Chr15, and Chr19.(B) Total numbers of pups born from Mps1DK/+ crosses. (C) Embryos (n = 96) harvested at E10.5 indicating embryonic lethality earlier than E10.5 for Mps1DK/DK

mice. The high fraction of reabsorbed embryos also suggests lethality.

Foijer et al. www.pnas.org/cgi/content/short/1400892111 7 of 13

Movie S1. MEFs expressing GFP-tagged full-lengthMps1 (green) coexpressing H2B-Cherry (red) showing kinetochore localization of full-lengthMps1, but notMps1DK.

Movie S1

Movie S2. MEFs expressing GFP-tagged Mps1DK (green) coexpressing H2B-Cherry (red) showing reduced kinetochore localization in Mps1DK.

Movie S2

Foijer et al. www.pnas.org/cgi/content/short/1400892111 8 of 13

Movie S3. MCF 10A cells expressing GFP-tagged full-length Mps1 (green) showing strong kinetochore localization of full-length Mps1.

Movie S3

Movie S4. MCF 10A cells expressing GFP-tagged Mps1DK (green) showing weak kinetochore localization of Mps1DK.

Movie S4

Foijer et al. www.pnas.org/cgi/content/short/1400892111 9 of 13

Movie S5. Mps1f/f MEFs expressing GFP-T2A-Cre showing lagging chromosomes coexpressing H2B-GFP.

Movie S5

Movie S6. Mps1f/f MEFs expressing GFP-T2A-Cre showing lagging chromosomes coexpressing H2B-Cherry.

Movie S6

Foijer et al. www.pnas.org/cgi/content/short/1400892111 10 of 13

Movie S7. Mps1f/f MEFs expressing GFP-T2A-Cre showing lagging chromosomes coexpressing H2B-Cherry. Note: Movie S7 shows the same cell as Movie S6 butshows GFP-T2A-Cre expression to visualize cell morphology.

Movie S7

Movie S8. Control-infected Mps1f/f MEFs expressing H2B-Cherry showing unperturbed mitosis.

Movie S8

Foijer et al. www.pnas.org/cgi/content/short/1400892111 11 of 13

Movie S9. Control-infected Mps1f/f MEFs expressing H2B-GFP showing unperturbed mitosis.

Movie S9

Movie S10. Cre–infected Mps1f/f MEFs expressing H2B-GFP showing failure to form a metaphase plate because of an alignment problem.

Movie S10

Foijer et al. www.pnas.org/cgi/content/short/1400892111 12 of 13

Movie S11. Cre-infected Mps1f/f MEFs expressing H2B-GFP showing failed chromosome alignment followed by tetraploidization.

Movie S11

Dataset S1. Genotyping and pathology information for all mice in this study including genotyping information of tumors

Dataset S1

Dataset S2. Alignments of Mps1 cDNA sequences generated from RNA isolated from tumors expressing full-length Mps1 or Mps1DK,showing loss of the sequence encompassing exons 3 and 4 in Mps1DK tumors

Dataset S2

Dataset S3. Normalized aCGH data for all assessed tumors and average aCGH data per chromosome

Dataset S3

Raw data have been deposited at in the National Center for Biotechnology Information Gene Expression Omnibus database under accession no. GSE57334.

Dataset S4. Normalized expression array data for all assessed tumors, average expression data per chromosome, and gene ontologyanalysis

Dataset S4

Raw data have been deposited at in the National Center for Biotechnology Information Gene Expression Omnibus database under accession no. GSE57334.

Dataset S5. Primer sequences used in this study

Dataset S5

Foijer et al. www.pnas.org/cgi/content/short/1400892111 13 of 13

Alignment: Assembled DNA alignment against reference moleculeParameters: Method: FastScan - Max Qual

Reference molecule: murine Mps1, Region 1 to 2493 Number of sequences to align: 7 Total length of aligned sequences with gaps: 2549 bps Settings: Similarity significance value cutoff: >= 100%

Summary of Percent Matches: Ref: murine Mps1 1 to 2493 ( 2493 bps) -- 2: Pr 494 TID63 1 to 1529 ( 1529 bps) 61% 3: Pr 503 TID63 1 to 968 ( 968 bps) 44% 4: Pr 500 TID63 1 to 1256 ( 1256 bps) 51% 5: Pr 498 TID63 1 to 1360 ( 1360 bps) 90% 6: Pr 501 TID63 1 to 1257 ( 1257 bps) 93% 7: Pr 489 TID63 1 to 367 ( 367 bps) 36%

Areas of significant similarity (in windows 20 bases in length)

murine Mps1

Pr 494 TID63

Pr 503 TID63

Pr 500 TID63

Pr 498 TID63

Pr 501 TID63

Pr 489 TID63

Alignment: Assembled DNA alignment against reference moleculeParameters: Method: FastScan - Max Qual

Reference molecule: murine Mps1, Region 1 to 2493 Number of sequences to align: 7 Total length of aligned sequences with gaps: 2620 bps Settings: Similarity significance value cutoff: >= 100%

Summary of Percent Matches: Ref: murine Mps1 1 to 2493 ( 2493 bps) -- 2: Pr 494 TID18 1 to 1266 ( 1266 bps) 49% 3: Pr 503 TID18 1 to 627 ( 627 bps) 30% 4: Pr 500 TID18 1 to 1352 ( 1352 bps) 61% 5: Pr 501 TID18 1 to 1352 ( 1352 bps) 57% 6: Pr 498 TID18 1 to 1326 ( 1326 bps) 62% 7: Pr 489 TID18 1 to 366 ( 366 bps) 90%

Areas of significant similarity (in windows 20 bases in length)

murine Mps1

Pr 494 TID18

Pr 503 TID18

Pr 500 TID18

Pr 501 TID18

Pr 498 TID18

Pr 489 TID18

Tumor 63, p53f/f; MMTV-Cre+ alligned to murine Mps1

Tumor 18, Mps1f/f p53f/f; Lck-Cre+ alligned to murine Mps1

Dataset S2: Allignments of tumor-isolated Mps1WT and Mps1DK to annotated murine Mps1