Specific tumor suppressor function for E2F2in Myc-induced T cell lymphomagenesisRene Opavsky*†, Shih-Yin Tsai*†, Martin Guimond*‡, Anjulie Arora*†, Jana Opavska*†, Brian Becknell*‡,Michael Kaufmann*†, Nathaniel A. Walton*†, Julie A. Stephens§, Soledad A. Fernandez§, Natarajan Muthusamy‡,Dean W. Felsher¶, Pierluigi Porcu‡, Michael A. Caligiuri*‡�, and Gustavo Leone*†�**

*Human Cancer Genetics Program, Department of Molecular Virology, Immunology, and Medical Genetics, College of Medicine and Public Healthand †Department of Molecular Genetics, College of Biological Sciences, §Center for Biostatistics, ‡Division of Hematology and Oncology, Department ofInternal Medicine, and �The Comprehensive Cancer Center, Ohio State University, Columbus, OH 43210; and ¶Division of Oncology, Department ofMedicine, Stanford University, CCSR 1105B, 269 Campus Drive, Stanford, CA 94305-5151

Edited by Tak Wah Mak, University of Toronto, Toronto, ON, Canada, and approved August 10, 2007 (received for review July 6, 2007)

Deregulation of the Myc pathway and deregulation of the Rbpathway are two of the most common abnormalities in humanmalignancies. Recent in vitro experiments suggest a complex cross-regulatory relationship between Myc and Rb that is mediatedthrough the control of E2F. To evaluate the functional connectionbetween Myc and E2Fs in vivo, we used a bitransgenic mousemodel of Myc-induced T cell lymphomagenesis and analyzed tumorprogression in mice deficient for E2f1, E2f2, or E2f3. Whereas thetargeted inactivation of E2f1 or E2f3 had no significant effect ontumor progression, loss of E2f2 accelerated lymphomagenesis.Interestingly, loss of a single copy of E2f2 also accelerated tumor-igenesis, albeit to a lesser extent, suggesting a haploinsufficientfunction for this locus. The combined ablation of E2f1 or E2f3,along with E2f2, did not further accelerate tumorigenesis. Myc-overexpressing T cells were more resistant to apoptosis in theabsence of E2f2, and the reintroduction of E2F2 into these tumorcells resulted in an increase of apoptosis and inhibition of tumor-igenesis. These results identify the E2f2 locus as a tumor suppressorthrough its ability to modulate apoptosis.

transcription � cancer

MYC is often amplified in human cancers, and mouse modelsof cancer have demonstrated a causal role for MYC over-

expression in hematopoietic, mammary, and other cancer types(1–3). Deregulation of the Rb/E2F gene networks also representscommon events in cancer (4). Like Myc, E2F can positively andnegatively regulate the expression of hundreds of targets whosegene products are involved in a wide spectrum of biologicalprocesses, with a bias for genes that control cell cycle, apoptosis, anddifferentiation (5–10).

Based on amino acid sequence analysis and structure–functionstudies in vitro, E2F family members can be artificially grouped intoactivator (E2F1–3) and repressor (E2F4–8) subclasses (11). Be-cause of the intense interest in E2Fs as major regulators of the cellcycle, individual E2F family members have also been extensivelystudied in vivo by gene-targeting approaches in mice. E2f1�/� miceare viable and suffer from impaired thymocyte apoptosis, defectivenegative selection, and testicular atrophy. E2f2�/� mice are alsoviable and have a mild increase in hematopoietic and autoreactiveT cells. Much later in life, a portion of E2f1�/� and E2f2�/� micedevelop hematopoietic malignancies (12–15). These mutant phe-notypes might reflect the particular bias for the expression of E2f1and E2f2 in hematopoietic tissues. Although disruption of the E2f3gene in a mixed genetic background yields viable mice, its disrup-tion in pure strains results in embryonic lethality at around em-bryonic day 12.5 (G.L., unpublished observation). Surprisingly,embryos deficient for each of these E2Fs have no apparent defectin cellular proliferation, raising the possibility of functional redun-dancy among members of the activator subclass of E2Fs. There alsoappears to be functional redundancy among members of therepressor subclass, because disruption of E2f4, E2f5, or E2f6 in mice

has little consequence on the proliferative capacity of cells, but thecombined disruption of E2f4 and E2f5 or E2f4 and E2f6 results inthe inappropriate proliferation and expression of target genes inresponse to specific antiproliferative signals (16, 17).

Recent observations in cell culture systems indicate extensivecross-regulation between the action of Myc and E2Fs in coordi-nating the control of cellular proliferation. Myc action can befunneled by a number of concerted mechanisms to control E2Factivity (5, 18), including through the regulation of their expression(7). Although Myc can certainly influence E2F activities, it is alsotrue that Myc can be influenced by E2Fs. How this complexcross-regulatory relationship between Myc and E2F is effectivelyorchestrated in vivo remains poorly understood. In this study, weused a mouse model of Myc-induced T cell lymphomagenesis alongwith mice deficient for each of the E2F activators to directlyexamine the connection between Myc action and these E2Fs in vivo.

ResultsE2f2 Locus Harbors Tumor Suppressor Function. To examine theconnection between Myc and E2F transcription factors in vivo, weused a conditional bitransgenic mouse model of MYC-induced Tcell lymphomagenesis (19). In this system, expression of E�SR-tTAmediates the transcription of the Teto-MYC transgene in B and Tcells and results in the development of predominantly immature Tcell lymphomas. These tumors invade the spleen, lymphatics, bonemarrow, and blood and eventually lead to the death of mice by 4months of age.

To explore the possibility that E2f1, E2f2, and E2f3 play a role inMYC-induced lymphomagenesis, we first determined whether theseE2Fs are expressed in hematopoietic cell lineages. As assessed byreal-time RT-PCR assays, E2f1, E2f2, and E2f3 are expressed in allof the main hematopoietic organs, including bone marrow, spleen,thymus, and lymph nodes (Fig. 1A). Most other organs testedexpressed significantly lower levels of E2f1 and E2f3 and essentiallylittle or no E2f2. It is not clear whether the particular highexpression of E2f2 in hematopoietic organs is a reflection of thehigh levels of E2f2 transcripts in these compartments or is just areflection of the low basal levels of E2f2 in other tissues. We alsoexamined the expression of these three E2fs in T cell tumors thatdeveloped in E�SR-tTA;Teto-MYC mice. This analysis revealed a

Author contributions: R.O. and G.L. designed research; R.O., S.-Y.T., M.G., A.A., J.O., M.K.,N.A.W., and G.L. performed research; B.B., D.W.F., and P.P. contributed new reagents/analytic tools; R.O., J.A.S., S.A.F., N.M., M.A.C., and G.L. analyzed data; and R.O. and G.L.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviation: MS, median survival.

**To whom correspondence should be addressed. E-mail: gustavo.leone@osumc.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0706307104/DC1.

© 2007 by The National Academy of Sciences of the USA

15400–15405 � PNAS � September 25, 2007 � vol. 104 � no. 39 www.pnas.org�cgi�doi�10.1073�pnas.0706307104

significant reduction in E2f2 expression in most tumors tested butlittle effect on the expression of E2f1, E2f3a, and E2f3b (Fig. 1B anddata not shown). The basis for the reduction in E2f2 expressionremains to be elucidated, but we speculated that it might be relatedto its potential role in the MYC-induced tumorigenic process.

To rigorously assess the physiological role of E2F activators inMYC-induced lymphomagenesis, we initially generated cohorts ofE�SR-tTA;Teto-MYC mice that lacked either one or both alleles ofE2f1 or E2f2. These mice were monitored for tumor formation overa period of 1 year, as described (19). Time of death was noted andplotted on Kaplan–Meier survival graphs. As shown previously,E�SR-tTA;Teto-MYC mice succumbed to tumors by 110 days of agewith a median survival (MS) of 83 days [confidence interval (C.I.)78–85 days; n � 109]. The inactivation of E2f1 had no significanteffect on the MS of these tumor mice (Fig. 1C). Strikingly, theinactivation of E2f2 significantly accelerated tumor onset andprogression, resulting in their death at a median age of 67 days (C.I.,65–70 days; n � 128; P � 0.001; Fig. 1C). Interestingly, loss of evenone copy of E2f2 significantly accelerated disease progression whencompared with the control wild-type cohort (E2f2�/�, P � 0.001),albeit to a lesser extent than when both alleles were deleted. Tumorsderived from E2f2 heterozygous animals retained the wild-typeallele, as assessed by Southern blot analysis [supporting information(SI) Fig. 5 and data not shown], suggesting that small changes in thelevels of E2F2 protein could have a substantial impact in tumori-genesis. Characterization of tumors by cell surface marker expres-sion confirmed that control and E2f2-deficient tumors were positivefor CD3 and TCR� and negative for the B cell marker B220 ormyeloid marker CD11b (data not shown), indicating tumorswere of T cell origin. In each case, tumors consisted of eitherCD4/CD8 double-positive cells or CD4 single-positive cells (SIFig. 6). Together, these results formally demonstrate a haplo-insufficient tumor suppressor function for the E2f2 locus in T celllymphomagenesis.

Germ-line inactivation of E2f3 in an FVB strain backgroundresults in embryonic lethality at around embryonic day 12.5,precluding tumor studies with these mice (G.L., unpublished ob-servations). To circumvent the problem of embryonic lethality, weanalyzed tumor development in E�SR-tTA;Teto-MYC mice con-taining the Teto-Cre transgene and a conditional allele of E2f3(E2f3LoxP). In these mice, tTA driven from the E�SR-tTA transgeneresults in the expression of both the Teto-MYC and Teto-Cretransgenes. Thus, MYC expression and Cre-mediated ablation ofE2f3 can be achieved within the same subset of hematopoietic cells.As shown in Fig. 1C, there was no significant difference in the MStime between the E�SR-tTA;Teto-MYC;E2f3LoxP/LoxP and E�SR-tTA;Teto-MYC;Teto-Cre;E2f3LoxP/LoxP groups of mice (P � 0.375).Southern blot analysis and PCR-based genotyping confirmed thecomplete ablation of E2f3 in tumors arising in E�SR-tTA;Teto-MYC;Teto-Cre; E2f3LoxP/LoxP mice (Fig. 1F and data not shown),suggesting that loss of E2f3 function was not selected against duringtumorigenesis. Moreover, Cre expression itself did not appear toaffect tumor outcome, because cohorts of mice containing theTeto-Cre transgene had identical MS times as those lacking theTeto-Cre transgene (data not shown). These results strongly suggestthat E2f3 does not significantly contribute to MYC-induced lym-phomagenesis. Because E2f3 was conditionally deleted in hemato-poietic lineages, as opposed to the global deletion of E2f1 or E2f2,we cannot rule out the formal possibility that inactivation of E2f3

black line) and E�SR-tTA;Teto-MYC;Teto-Cre;E2f1�/�;E2f2�/�;E2f3LoxP/LoxP ([�/�;�/�;LoxP/LoxP]; red line). (F) Conditional deletion of the E2f3LoxP allele in Tcell lymphomas derived from E�SR-tTA;Teto-MYC;Teto-Cre;E2f3LoxP/LoxP orE�SR-tTA;Teto-MYC;Teto-Cre;E2f3�/LoxP mice as analyzed by Southern blot-ting using an E2f3-specific probe. Nondeleted E2f3LoxP/LoxP genomic DNAserved as a control. The position of bands corresponding to E2f3 knockout (�),E2f3LoxP (LoxP), and E2f3 wild-type (�/�) alleles is indicated.

Fig. 1. Loss of E2f2 accelerates Myc-induced T cell lymphomagenesis. (A)Real-time RT-PCR analysis of the expression of E2f1, E2f2, E2f3a, and E2f3b innormal mouse tissues (b.m., bone marrow; l.n., lymph node). The relativevalues were determined by comparing the expression of the indicated mRNAsto the expression of GAPDH. The data are shown as induction (n-fold) of geneexpression in tissues relative to expression in the liver that was set to one. Onerepresentative example of three independent experiments is shown. Errorbars represent mean � SD. Values that exceeded the y-axis scale are shown onthe top of interrupted bars. (B) Real-time RT-PCR analysis of E2f1, E2f2, E2f3a,and E2f3b expression in normal mouse thymocytes (controls) and MYC-induced T cell lymphomas (tumors). (C) Kaplan–Meier survival curves (timefrom birth to death) of the E�SR-tTA (tetracycline controlled transactivator);Teto-MYC (�/�) and E�SR-tTA;Teto-MYC cohorts of mice containing het-erozygous (�/�) or homozygous (�/�) mutations of indicated E2fs. Survivalcurve for E�SR-tTA;Teto-MYC is shown by solid black line. Curves for thecohorts of mice containing deletion of E2f2 allele are shown by red lines,whereas curves for the cohorts of mice that did not involve deletion of E2f2 areshown in blue. Dashed lines and solid lines indicate heterozygous and ho-mozygous deletion of E2fs, respectively. The number of mice used in eachcohort is indicated by n. Student’s t test was used for statistical analyses asdescribed in Materials and Methods, and P values are shown. (D) Kaplan–Meier survival curves of the E�SR-tTA;Teto-MYC ([�/�;�/�]; black line), E�SR-tTA;Teto-MYC;Teto-Cre;E2f1�/�;E2f3LoxP/LoxP ([�/�;LoxP/LoxP]; red line),E�SR-tTA;Teto-MYC;E2f1�/�;E2f2�/� ([�/�;�/�]; red line), or E�SR-tTA;Teto-MYC;Teto-Cre;E2f2�/�;E2f3LoxP/LoxP ([�/�;LoxP/LoxP]; red line) cohorts of mice.(E) Kaplan–Meier survival curves of the E�SR-tTA;Teto-MYC ([�/�;�/�;�/�];

Opavsky et al. PNAS � September 25, 2007 � vol. 104 � no. 39 � 15401

GEN

ETIC

S

in the germ line would have a different phenotypic consequence onT cell lymphomagenesis than observed here.

E2f1 and E2f3 Are Dispensable for MYC-Induced Lymphomagenesis.Functional compensation among E2F family members, as demon-strated in fibroblasts cultured in vitro (20), may explain the lack ofan effect on tumor outcome imparted by the loss of either E2f1 orE2f3. We therefore analyzed disease progression in mice deleted forboth E2f1 and E2f3. Surprisingly, E�SR-tTA;Teto-MYC;E2f1�/�;E2f3LoxP/LoxP and E�SR-tTA;Teto-MYC;Teto-Cre;E2f1�/�;E2f3LoxP/LoxP mice developed T cell lymphomas with similar kineticsas wild-type mice (Fig. 1D). Southern blot analysis of tumor DNAconfirmed the complete deletion of E2f3 in tumors derived frommice expressing Cre (SI Fig. 7A and data not shown). From thesein vivo studies, we conclude that E2f1 and E2f3 do not play ameasurable role in MYC-induced lymphomagenesis.

E2f2’s Tumor Suppressor Function Is Independent of E2f1 and E2f3.The tumor studies described above demonstrate specificity amongE2fs in the manifestation of tumor outcome; however, because ofthe incredible functional plasticity among E2F family members(17), it is difficult to ascertain the molecular basis for this specificity.Although no significant change in the expression of other E2Ffamily members was observed in tumors deficient for E2f2 (data notshown), it is possible that E2F1 and/or E2F3 action might have beenrerouted in E2f2-deficient cell to perform functions not normallyperformed in cells containing all three activator E2Fs. These‘‘acquired functions’’ could be responsible for diminishing or ac-centuating the manifestation of tumor outcome resulting from theloss of E2f2. To test this possibility, we created cohorts of E�SR-tTA;Teto-MYC;E2f2�/� mice that were also deficient for eitherE2f1, E2f3, or both. As shown in Fig. 1D, E�SR-tTA;Teto-MYC;E2f1�/�;E2f2�/� andE�SR-tTA;Teto-MYC;Teto-Cre;E2f2�/�;E2f3LoxP/LoxP cohorts had a similar MS time as E�SR-tTA;Teto-MYC;E2f2�/� mice (66, 67, and 67 days, respectively). Surprisingly,the simultaneous inactivation of all three E2F activators in E�SR-tTA;Teto-MYC;Teto-Cre;E2f1�/�;E2f2�/�;E2f3LoxP/LoxP mice re-sulted in a similar progression of disease as in E2f2-deficient mice(MS time of 65 and 67 days, respectively; Fig. 1E). Once again,Southern blot analysis confirmed the complete deletion ofE2f3LoxP/LoxP in the subset of tumors expressing Cre (SI Fig. 7 B andC). Based on these data, we conclude that E2f2’s tumor suppressorrole in T cell lymphomagenesis is independent of E2f1 and E2f3.

Cell-Autonomous Tumor Suppressor Function of E2F2. The globalinactivation of E2f2 precluded us from making any conclusionsrelating to where the critical action of E2F2 for suppressinglymphomagenesis might reside. To examine whether loss of E2f2accelerated the tumorigenic process in a cell-autonomous manner,we used an adoptive transfer strategy to introduce E2f2-deficientfetal liver cells into wild-type irradiated animals. To this end,lethally irradiated recipient FVB mice were injected with fetal livercells isolated from E�SR-tTA;Teto-MYC and E�SR-tTA;Teto-MYC;E2f2�/� 15.5-day-old embryos. Injected mice were then mon-itored for tumor formation over a period of 1 year (25 mice pergenetic group; see SI Fig. 8A). Approximately 60% of mice (14/25)from the group that received E�SR-tTA;Teto-MYC cells developedT cell lymphomas, as confirmed by FACS-based immunopheno-typing and histological analysis (SI Fig. 8B and data not shown) andsuccumbed to tumors with a MS time of 160 days. The other 40%of mice remained healthy during the 300-day observation period. Incontrast, most mice (22/25) that received E�SR-tTA;Teto-MYC;E2f2�/� fetal liver cells developed T cell lymphomas and hada MS time of 110 days (SI Fig. 8A). Lymphomas that lacked E2f2were immunophenotypically indistinguishable from those that hadE2f2. Statistical comparison between these two groups showed asignificant difference in their MS (P � 0.019). These studiesdemonstrate that the tumor suppressor function of E2f2 resides

within the hematopoietic compartment and is likely cell-autonomous.

Tumor Analysis in E2f2-Deficient Mice. Loss of E2f2 has been shownto lead to autoimmune disease in older mice due to enhancedTCR-stimulated proliferation and the accumulation of autoreactiveeffector/memory T lymphocytes (14). We therefore exploredwhether E2f2’s tumor suppressor function described above might becausally related to a T cell differentiation defect, which could bedivided into four substages depending on the expression of CD25and CD44 markers: DN1 (CD44�CD25�), DN2 (CD44�CD25�),DN3 (CD44�CD25�), and DN4 (CD44�CD25�) (21). To this end,we analyzed the expression of CD44, CD25, CD4, and CD8 inthymocytes derived from 21-day-old wild-type E2f2�/�, E�SR-tTA;Teto-MYC, and E�SR-tTA;Teto-MYC;E2f2�/� mice. Loss ofE2f2 in either a nontumor or tumor setting had no appreciableeffect on the distribution of DN1–DN4 cells or the proportion ofCD4 or CD8 single or CD4–CD8 double-positive cells (Fig. 2A anddata not shown). We did observe, however, that in contrast to themonoclonal nature of tumors derived from E�SR-tTA;Teto-MYCmice (19), the vast majority of tumors in the E2f2�/� and E2f2�/�

background were oligoclonal, with homozygously deleted micehaving tumors composed of up to five different clones that ex-pressed different TCR receptor isoforms (Fig. 2B).

In view of the well established role of E2Fs in the control ofcellular proliferation and apoptosis, we investigated whetherchanges in these two processes may represent the basis for thetumor suppressor function of E2F2. To this end, we measuredproliferation and apoptotic indices in control and E2f2-deficientlittermate animals at early and late stages of tumor development.Analysis of precancerous (21 days of age) or terminally sick mice

Fig. 2. Decreased apoptosis in mice deficient for E2f2 during tumorigenesis.(A) Development of double-negative thymocytes in 21-day-old E2f2�/� (non-transgenic; white bars), E2f2�/� (nontransgenic; black bars), E�SR-tTA;Teto-MYC;E2f2�/� (E�SR-tTA;Teto-MYC; white bars), E�SR-tTA;Teto-MYC;E2f2�/�

(E�SR-tTA;Teto-MYC; black bars) as assessed by FACS. Staining with anti-CD4and -CD8 antibody was used to determine CD4�CD8� double-negative pop-ulation within the live lymphoid gate. Double-negative populations weresubsequently analyzed for CD44 and CD25 surface expression. Data are pre-sented as an average percentage � SD for DN1 (CD44�), DN2 (CD44�CD25�),DN3 (CD44�CD25�), and DN4 (CD44�CD25�). (B) Graphic representation oftumor clonality in E�SR-tTA;Teto-MYC mice of the following genotypes:E2f2�/�, E2f2�/�, and E2f2�/�, as determined by analysis of different tissuesusing a panel of monoclonal antibodies recognizing different TCR V� chains.(C) BrdU incorporation and apoptosis assays of E�SR-tTA;Teto-MYC;E2f2�/�

(white bars) or E�SR-tTA;Teto-MYC;E2f2�/� (black bars) mice at final stages ofdisease as determined by FACS using anti-BrdU or anti-Annexin V antibodies.The number of mice used for each cohort is indicated by n. Student’s t test wasused for statistical analyses, and P values are shown.

15402 � www.pnas.org�cgi�doi�10.1073�pnas.0706307104 Opavsky et al.

failed to reveal any substantial difference in the proliferation ofcontrol and E2f2-deficient T cells (Fig. 2C and data not shown). Incontrast, AnnexinV staining revealed that E2f2-deficient T cellswere more resistant to apoptosis than control cells at both early andlate stages of tumor development, suggesting that a loss of E2f2 mayconfer a selective advantage for the expansion of MYC-expressingT cells.

Molecular Characterization of T Cell Lymphomas in E2f2-DeficientMice. To identify relevant downstream activities that might beresponsible for the observed acceleration of tumorigenesis inE2f2-deficient mice, we analyzed the expression of cell cycle inhib-itors that have previously been implicated in cancer biogenesis.First, we performed real-time RT-PCR expression analysis of theMYC-transgene itself and two known MYC-target genes, Ornithinedecarboxylase (Odc) and Nucleolin, in thymic tumor masses fromterminally sick E�SR-tTA;Teto-MYC and E�SR-tTA;Teto-MYC;E2f2�/� mice. This analysis confirmed that the expression ofthe MYC transgene and its two target genes was similarly activatedin all tumors examined when compared with T cells from aged-matched normal control mice (Fig. 3A and data not shown).

Previous work using the E�-MYC model of B cell lymphomademonstrated that MYC-induced B cell lymphomagenesis requiresthe inactivation of apoptotic checkpoints, and that this is frequentlyachieved through disabling the Arf-p53 pathway (22). We thereforeassessed the status of p53 in tumors arising in E�SR-tTA;Teto-MYCand E�SR-tTA;Teto-MYC;E2f2�/� mice. Direct sequencing of p53cDNA prepared from 10 tumors did not reveal any mutation in its

coding sequence (SI Text). We then analyzed the expression ofp19ARF, whose induction had been shown to be associated with theoverexpression of MYC (23). As might have been expected, weobserved a dramatic induction of p19ARF expression in all tumorsanalyzed, but this induction was independent of the status of E2f2(Fig. 3A). Consistent with the conclusion that the p53 pathway is notdifferentially impacted by the presence or absence of E2f2, theexpression of one of its target genes, p21Cip1, did not change inresponse to the loss of E2f2 (Fig. 3A). Because the expression ofp73, a proapoptotic member of the p53 family, has been implicatedas a downstream target of E2F1 during TCR activation-induced celldeath of peripheral T cells (24), we examined its expression innormal and tumor tissues from E2f2-deficient mice. We found thatp73 transcripts were significantly decreased in tumor samples, butthis decrease did not depend on the status of E2f2 (Fig. 3A). Incontrast, the expression of the antiapoptotic version of p73,deltaNp73, was unchanged between normal and tumor samples (SIFig. 9). These results suggest that E2F2’s tumor suppressor functionis independent of the p53 apoptotic axis.

The ability of Myc to impact the cell cycle at least partiallydepends on its ability to down-regulate p27Kip1 (25), a cyclin-dependent kinase inhibitor known to contribute to the regulationof the cell cycle and to be down-regulated in a number of tumorsettings (26, 27). Consistent with a posttranscriptional mechanismof regulation, p27Kip1 protein levels but not its mRNA levels, weresubstantially decreased in tumor cells derived from E�SR-tTA;Teto-MYC or E�SR-tTA;Teto-MYC;E2f2�/� mice (Fig. 3 A and B).Although this down-regulation of p27Kip1 likely represents animportant event in MYC-induced T cell lymphomagenesis, itsregulation would appear not to be linked to E2F2’s tumor suppres-sor function.

Reintroduction of E2f2 Activity into E2F2-Deficient Cells InhibitsTumorigenesis. To further examine the role of E2f2 in lym-phomagenesis, we evaluated T cells isolated from late-stage tumorsthat were reconstituted with exogenous E2F2. Tumor cells derivedfrom E�SR-tTA;Teto-MYC;E2f2�/� mice readily adapted to in vitrogrowth and could be efficiently infected with retroviral expressionvectors (Fig. 4 A and B). Flow cytometric analysis in multipleexperiments showed that 20–80% of tumor cells infected with acontrol or E2F2-expressing MSCV-IRES-GFP vector were GFP-positive (Fig. 4B). Subcutaneous injection of these cells resulted inthe formation of tumors within 2 weeks. Essentially all of the tumorcells that emerged from the injection of E2F2-transduced cells wereGFP-negative (�99% GFP-negative; Fig. 4 B and C). In contrast,tumors that emerged from the injection of control-transduced cellsretained the same percentage of GFP-positive cells as observedbefore the injection. These experiments reveal a profound biasagainst the formation of tumors originating from E2F2 overex-pressing T cells.

Alternatively, control and E2F2-transduced cells were plated ontissue culture plates, and their growth was monitored over thecourse of �8 days. In these assays, we could measure a moderatebut consistent reduction in the proliferation of cells infected withE2F2-expressing vectors (Fig. 4D). Importantly, comparison ofGFP-positive and -negative cells by FACS analysis revealed aprofound decrease over time in the percentage of GFP-positivecells transduced with E2F2-vectors (Fig. 4E). In contrast, GFP-positive and -negative cells in control-treated samples proliferatedequally well. BrdU incorporation assays indicated that the numberof GFP-positive cells entering S phase was not influenced by theover-expression of E2F2 (Fig. 4F and data not shown). AnnexinVassays, however, revealed that the ratio of GFP-positive/-negativeapoptotic cells was markedly increased in populations transducedwith E2F2-expressing vectors but not with control vectors (Fig. 4G).These results suggest a strong bias against the proliferation of cellsexpressing the E2F2 protein that is based on its ability to potentlyinduce apoptosis.

Fig. 3. Expression of cell cycle-regulated genes in E2f2-deleted cells. (A)Real-time RT-PCR analysis of MYC, ODC, p19ARF, p73, p21Cip1, and p27Kip1

expression in normal thymocytes (E2f2�/�), E2f2-deficient thymocytes(E2f2�/�), tumors derived from E�SR-tTA;Teto-MYC;E2f2�/� or E�SR-tTA;Teto-MYC;E2f2�/� mice, as indicated. Sequences of primers used fortested genes are shown in SI Table 1. (B) Western blot analysis of p27Kip1

expression in normal thymocytes (N) and E�SR-tTA;Teto-MYC tumors nonde-leted or deleted for E2f2 as indicated.

Opavsky et al. PNAS � September 25, 2007 � vol. 104 � no. 39 � 15403

GEN

ETIC

S

In parallel experiments, we could show that overexpression ofE2F1 and E2F3a in E2f2-deficient cells could also induce apoptosisand preclude the growth and tumorigenicity of tumor cells (SI Fig.10 A–F and data not shown). These results suggest that usingoverexpression approaches, any of the three E2F activators canengage apoptotic pathways and thus eliminate tumor cells. We viewthese results to indicate that gene ablation strategies can revealfunctional specificity with greater fidelity than by overexpressionstrategies.

DiscussionThe E2f2 Locus Harbors Tumor Suppressor Function. Overexpressionof the MYC oncogene and inactivation of the RB tumor suppressorpathway are hallmarks of human cancers. Recent in vitro experi-ments suggest a complex cross-regulatory relationship between Myc

and Rb that is mediated through the control of E2F activities. Here,we used a bitransgenic mouse model of MYC-induced T celllymphomagenesis and mice deficient for E2f1, E2f2, or E2f3 toevaluate the functional relationship between Myc and E2Fs in vivo.These experiments demonstrate a unique tumor suppressor role forE2F2 in T cell lymphomagenesis. Adoptive transfer experimentsshow that E2f2’s tumor suppressor function resides within thehematopoietic compartment and is therefore likely to be cell-autonomous. Loss of even one allele significantly acceleratedtumorigenesis, indicating that tumor progression is sensitive tosmall changes in total E2F2 protein. This raises the possibility thatpolymorphisms in the genome that result in lower levels of E2F2protein, directly or indirectly, may place individuals at a higher riskfor cancer development.

E2f1 and E2f3 Are Not Required for Lymphomagenesis. Based on theirshared abilities to control cell proliferation and apoptosis, a func-tional connection between the Myc and E2F pathways has beenlong speculated. Most recently, work in mouse embryo fibroblastssuggested that two important functions of Myc in the control ofproliferation and apoptosis are mediated, at least in part, by E2Fs(28). This work showed that in fibroblasts, the execution of Myc’sproliferative arm requires E2f2 and E2f3, and the execution of itsapoptotic arm requires E2f1. This bifurcation of Myc’s function atthe level of E2F suggested that E2f1 could have tumor suppressorfunction and E2f2 and E2f3 could have oncogenic functions. Thesepredictions were not born out by the in vivo studies of T celllymphomagenesis presented here. In fact, mice lacking both E2f1and E2f3 developed T cell lymphomas with similar kinetics as micecontaining a full complement of E2Fs, suggesting that E2f1 andE2f3 do not play a measurable role in MYC-induced lymphomagen-esis. The observation that T cells devoid of E2f1, E2f2, and E2f3proliferated and were fully transformed is quite surprising given ourprevious work showing that fibroblasts deficient for these E2Fswere unable to proliferate in vitro (20). We do not yet know whetherthis difference between E2f1/2/3-deficient T cells and fibroblastsreflects a tissue-specific requirement for E2Fs or a consequence ofMYC overexpression in T cells. Clearly, important differences mustexist between fibroblasts and T cells, and a generalized outcomestemming from the action of E2Fs across different cell contexts maybe difficult to predict.

The observation that loss of E2f1 had little bearing on MYC-induced T cell lymphomagenesis is also in direct contradiction to arecent study by Baudino et al. (29), where the authors show that lossof E2f1 dramatically delayed MYC-induced B cell lymphomagen-esis. These two different tumor outcomes resulting from theinactivation of E2f1 could reflect inherent differences in the biologyof B and T cells. Alternatively, differences in the two E2f1-nullalleles (12, 13) or in genetic backgrounds (30, 31) housing envi-ronments, methods of tumor analysis, and size of genetic cohortscould account for the observed discrepancies.

Tumor Suppressor Role of E2F2 in Apoptosis. Two observationsindicate that the underlying basis for E2f2’s tumor suppressorfunction is in the control of apoptosis rather than in cell prolifer-ation. First, we observed a decreased number of apoptotic tumorcells in E2f2-null mice. Second, reexpression of E2F2 in E2f2-deficient tumor cells resulted in an increase of apoptosis andabrogation of tumor cell expansion, demonstrating that the criticalcomponents required to signal and execute apoptosis remain intactin these tumor cells. In contrast, these loss-of-function and over-expression studies failed to reveal any E2f2-dependent differencesin the ability of MYC-expressing tumor cells to replicate their DNA.

The role of E2F2 in apoptosis early on during tumorigenesiscould be important in limiting the number of cells susceptible toMYC-induced oncogenesis. This hypothesis would predict that lossof E2f2 might increase the tumor-prone T cell population respon-sive to MYC overexpression and thus facilitate tumor development.

Fig. 4. Proapoptotic tumor suppressor function of E2f2. (A) Western blotanalysis of T cells derived from E�SR-tTA;Teto-MYC;E2f2�/� tumors infectedwith MSCV-IRES-EGFP empty vector (con) or MSCV-E2F2-IRES-EGFP (E2F2)using anti-E2f2 antibody. Tubulin served as loading control. (B) FACS analysisof unselected cells infected with the indicated retroviruses before injectioninto nude mice (Upper). Representative examples of FACS analysis of individ-ual tumors that developed in nude mice (Lower). The percentage of GFP-positive and -negative cells is indicated within the FACS diagrams. (C) Analysisof tumors that developed in mice injected with E�SR-tTA;Teto-MYC;E2f2�/�

tumor T cells that were infected either with control or E2F2 retroviruses. Thedata are presented as the average ratio of percent of GFP-positive cells in eachindividual tumor relative to the percent of GFP-positive cells before injection.n indicates a number of tumors analyzed for each group. (D) In vitro prolif-eration assay of unselected cells infected with the indicated retroviral con-structs. Cells were plated at a concentration of 0.2 � 106 per ml (day 0) andcounted every 24 h for 7 days. (E) Cells were treated as in D, and the percentageof GFP-positive cells was determined by FACS. The infection efficiency at day0 (28% for the MSCV-IRES-EGFP vector control and 39% for MSCV-E2F2-IRES-EGFP) was set to 100%. The values obtained for percentage of GFP-positivecells at each time point were plotted relative to the percentage at day 0. (F andG) The percentage of BrdU- and Annexin V-positive cells infected with theindicated retroviruses was measured 2 and 4 days after infection. For D–G,representative examples from three independent experiments are shown.

15404 � www.pnas.org�cgi�doi�10.1073�pnas.0706307104 Opavsky et al.

Consistent with this notion, the vast majority of E2f2�/� andE2f2�/� mice developed tumors that were oligoclonal in nature,with homozygously deleted tumors expressing up to five differentTCR receptor isoforms (Fig. 2B). This is in contrast to the typicalmonoclonal nature of tumors found in E2f2�/� mice. These datasuggest a role for E2f2 in both the early and late stages of tumorprogression.

Specificity of Function Among E2Fs. The tumor analysis in E2f-deficient mice clearly demonstrates a specific role for E2f2 in T celllymphomagenesis. The molecular basis for this tumor suppressorfunction, however, is less clear. On the one hand, a decrease in E2F2protein levels but not in E2F1 and E2F3 through the targetedinactivation of E2f2 is sufficient to suppress apoptosis and permittumor progression. On the other hand, overexpression of E2F2 orE2F1 and, to lesser degree, E2F3a, is able to induce apoptosis inMYC-overexpressing tumor cells. Thus, loss-of-function and over-expression studies appear to lead to contradictory results. Becauseoverexpression of E2Fs can potentially compete with and alter thebinding of other E2Fs to target promoters and/or cofactors, webelieve that gene ablation approaches are more adept at revealingphysiological differences between the function of E2F familymembers.

What is the underlying reason for the unique tumor suppressorfunction of E2f2 in MYC-induced lymphomagenesis that is observedin vivo? One possibility is that endogenous levels of E2F2 proteinregulate the expression of a specific set of apoptotic-related genesthat can be similarly achieved only by other E2Fs when overex-pressed at supraphysiological levels. The alternative explanation isthat E2f2’s tumor suppressor role is not intrinsic to the function ofits protein product but rather depends on the magnitude ofexpression imparted by its locus. In other words, the size of the‘‘total pool’’ of E2F activity may be the critical variable thatdetermines whether an apoptotic response can be surmounted inface of an oncogenic insult. Although it is possible that the basis forthe unique role of E2f2 in MYC-induced lymphomagenesis maystem from quantitative differences in the expression of ‘‘activator’’E2Fs in T cells, three main reasons argue against this latterpossibility. First, if ‘‘total’’ E2F activity was determining tumorsuppression, it would be expected that loss of all three E2Factivators would accelerate lymphomagenesis further than thatobserved by the simple loss of E2f2. This prediction was notrealized; rather, a deficiency of E2f1/E2f2/E2f3 accelerated lym-phomagenesis to the same extent as by loss of E2f2 (Fig. 1 D andE). Second, the absence of ‘‘enhanced’’ tumorigenesis in T cellsdeficient for all three E2F activators is not simply because E2f1/3are not expressed to sufficient levels to have a function in T cells.In fact, E2F1 has been shown, as has E2F2, to contribute to normalhematopoiesis in a number of settings. For example, it has been

previously shown that in contrast to E2f1�/�E2f2�/� orE2f1�/�E2f2�/� cells, E2f1�/�E2f2�/� bone marrow cells are un-able to contribute to development of multiple hematopoietic lin-eages (33), suggesting that a single allele of E2f1 or E2f2 contributesignificantly to this process. Third, in unpublished work from ourlaboratory, we have observed that loss of E2f2 in tissues thatnormally express much lower levels of it than in T cells alsoaccelerates tumorigenesis (G.L., unpublished observations). Itwould thus seem that the tumor suppressor role of E2f2 is not simplydue to its abundance in T cells but rather is likely a reflection of itsspecific function in T cell lymphomagenesis.

Whether E2F2 might also play a role in tumor maintenanceremains to be investigated. Further experiments will be necessaryto decipher the exact mechanism by which E2F2 exerts its tumorsuppressor action in vivo. In summary, these studies reveal speci-ficity among E2F activators in MYC-induced lymphomagenesis,highlighting an apoptotic role for E2f2 in this process.

Materials and MethodsGeneration and Maintenance of Mice. The E2f1�/�, E2f2�/�, andTeto-Cre mice were generous gifts from Michael Greenberg (Chil-dren’s Hospital, Boston, MA), Stuart Orkin (Harvard MedicalSchool, Boston, MA), and Andreas Nagy (Samuel LunenfeldResearch Institute, Toronto, ON, Canada), respectively. The gen-eration of the conditional E2f3 knockout mice (E2f3LoxP/LoxP) hasbeen described (20). Genotyping of mice was performed by PCRfrom genomic DNA isolated from mouse tails. All tumor studieswere performed with mice bred into FVB (fifth generation). Micefor tumor studies generated using standard genetic procedures weremonitored for tumor formation over a period of 1 year, as de-scribed (19).

Statistical Analysis. Kaplan–Meier curves were generated, and MStimes with 95% confidence intervals (32) were calculated. Propor-tional hazards assumptions were confirmed, and the log-rank testwas found to be appropriate to compare the survival curves in allcases. Bonferroni adjustments for multiple comparisons were used.Although this is a conservative method of adjustment, the P valuesfound in these data were on the extreme ends, and a less-conservative method would have lead to the same conclusions. Inthe case where the proportional hazard assumption was not met,such as for the E2f1�/�E2f3�/� cohort (Fig. 1D), the survival curveswere compared between groups before and after the cross-over ofthe survival curves.

This work was funded by National Institutes of Health Grants R01CA85619 and P01 CA097189 (both to G.L.) and by a translational awardby the Leukemia and Lymphoma Society of America (to G.L.). R.O. issupported by a T32 CA106196 fellowship in Cancer Genetics, and G.L.is the recipient of the Pew Charitable Trusts Scholar Award and theLeukemia and Lymphoma Society Scholar Award.

1. Arvanitis C, Felsher DW (2005) Cancer Lett 226:95–99.2. Jonkers J, Berns A (2004) Cancer Cell 6:535–538.3. Pelengaris S, Khan M, Evan G (2002) Nat Rev Cancer 2:764–776.4. Sherr CJ, McCormick F (2002) Cancer Cell 2:103–112.5. Adhikary S, Eilers M (2005) Nat Rev Mol Cell Biol 6:635–645.6. Evan GI, Vousden KH (2001) Nature 411:342–348.7. Sears R, Nevins JR (2002) J Biol Chem 277:11617–11620.8. Attwooll C, Denchi EL, Helin K (2004) EMBO J 23:4709–4716.9. Cobrinik D (2005) Oncogene 24:2796–2809.

10. Dimova DK, Dyson NJ (2005) Oncogene 24:2810–2826.11. Trimarchi JM, Lees JA (2002) Nat Rev Mol Cell Biol 3:11–20.12. Field SJ, Tsai FY, Kuo F, Zubiaga AM, Kaelin WG, Livingston DM, Orkin SH, Greenberg

ME (1996) Cell 85:549–561.13. Yamasaki L, Jacks T, Bronson R, Goillot E, Harlow E, Dyson NJ (1996) Cell 85:537–548.14. Murga M, Fernandez-Capetillo O, Field SJ, Moreno B, Borlado LR, Fujiwara Y, Balomenos

D, Vicario A, Carrera AC, Orkin SH, et al. (2001) Immunity 15:959–970.15. Zhu JW, Field SJ, Gore L, Thompson M, Yang H, Fujiwara Y, Cardiff RD, Greenberg M,

Orkin SH, DeGregori J (2001) Mol Cell Biol 21:8547–8564.16. Gaubatz S, Lindeman GJ, Ishida S, Jakoi L, Nevins JR, Livingston DM, Rempel RE (2000)

Mol Cell 6:729–735.17. Giangrande PH, Zhu W, Rempel RE, Laakso N, Nevins JR (2004) EMBO J 23:1336–1347.18. Gartel AL, Radhakrishnan SK (2005) Cancer Res 65:3980–3985.

19. Felsher DW, Bishop JM (1999) Mol Cell 4:199–207.20. Wu L, Timmers C, Maiti B, Saavedra HI, Sang L, Chong GT, Nuckolls F, Giangrande P,

Wright FA, Field SJ, et al. (2001) Nature 414:457–462.21. Cantrell DA (2002) Nat Rev Immunol 2:20–27.22. Eischen CM, Weber JD, Roussel MF, Sherr CJ, Cleveland JL (1999) Genes Dev 13:2658–2669.23. de Stanchina E, McCurrach ME, Zindy F, Shieh SY, Ferbeyre G, Samuelson AV, Prives C,

Roussel MF, Sherr CJ, Lowe SW (1998) Genes Dev 12:2434–2442.24. Lissy NA, Davis, Irwin M, Kaelin WG, Dowdy SF (2000) Nature 407:642–645.25. Sherr CJ, Roberts JM (1999) Genes Dev 13:1501–1512.26. Kang-Decker N, Tong C, Boussouar F, Baker DJ, Xu W, Leontovich AA, Taylor WR,

Brindle PK, van Deursen JM (2004) Cancer Cell 5:177–189.27. Keller UB, Old JB, Dorsey FC, Nilsson JA, Nilsson L, MacLean KH, Chung L, Yung C,

Spruck C, Boyd K, et al. (2007) EMBO J 26:2562–2574.28. Leone G, Sears R, Huang E, Rempel R, Nuckolls F, Park CH, Giangrande P, Wu L,

Saavedra HI, Field, SJ, et al. (2001) Mol Cell 8:105–113.29. Baudino TA, Maclean KH, Brennan J, Parganas E, Yang C, Aslanian A, Lees JA, Sherr CJ,

Roussel MF, Cleveland JL (2003) Mol Cell 11:905–914.30. Wikonkal N, Remenyik E, Knezevic D, Zhang W, Liu M, Zhao H, Berton TR, Johnson DG,

Brash DE (2003) Nat Cell Biol 5:655–660.31. Wloga EH, Criniti V, Yamasaki L, Bronson RT (2004) Nat Cell Biol 6:565–567.32. Brookmeyer R, Crowley J (1982) Biometrics 38:29–41.33. Li FX, Zhu JW, Hogan CJ, DeGregori J (2003) Mol Cell Biol 23:3607–3622.

Opavsky et al. PNAS � September 25, 2007 � vol. 104 � no. 39 � 15405

GEN

ETIC

S