Compensation and Specificity of Function Within the E2F Family

L-J Kong, JT Chang, AH Bild and JR Nevins

Department of Molecular Genetics and Microbiology, Institute for Genome Sciences and Policy, Duke University Medical Center, Durham, NC, USA

Functions encoded by single genes in lower organisms are maintained within the family of genes, including often represented by multiple related genes in the functional overlap.In many instances, this complexity mammalian genome. An example is the retinoblastoma of structure and function confounds genetic analyses and E2F families of proteins that regulate transcription that aim to associate discrete functions within individual duringthe cell cycle. Analysis of genefunction using genes.Further complicating the analysis is the use of germline mutations is often confounded by overlapping germline alterations that inactivate a gene of interest.If function resultingin compensation. Indeed, in cells deleted the particular gene is necessary for an essential function, of the E2F1 or E2F3 genes, there is an increase in the the imposed loss of function will either result in an expression of the other family member. To avoid inviable animal or will force a selection whereby another complications of compensatory effects, we have used function compensates for the loss of activity.The latter small-interferingRNAs that target individual E2F pro- is particularly relevant when dealing with a multigene teins to generate a temporary loss of E2F function. We family encoding related proteins with similar functions. find that both E2F1 and E2F3 are required for cells to An example of such complexity can be found in the enter the S phase from a quiescent state, whereas only retinoblastoma (Rb) and E2F families of proteins that E2F3 is necessary for the S phase in growing cells. We regulate transcription during the cell cycle.Whereas also find that the acute loss of E2F3 activity affects the C. elegans encodes a single Rb protein and Drosophila expression of genes encoding DNA replication and mitotic encodes two Rb proteins, three related genes can be activities, whereas loss of E2F1 affects a limited number identified in the mammalian genome (Dimova and of genes that are distinct from those regulated by E2F3. Dyson, 2005).Similarly, there are two E2F genes in We conclude that the long-term loss of E2F activity does the C. elegans and Drosophila genomes in contrast to the lead to compensation by other family members and that eight E2F genes in the mammalian genome (Attwooll the analysis of acute loss of function reveals specific and et al., 2004; Dimova and Dyson, 2005). Furthermore, a distinct roles for these proteins. clear example of the overlapping nature of function can Oncogene (2007) 26, 321–327.doi:10.1038/sj.onc.1209817; be seen in the analysis of Rb function using germline published online 14 August 2006 versus conditional loss-of-function mouse mutants. Mouse embryo fibroblasts (MEFs) isolated from ani- Keywords: E2F; cell cycle; expression signatures mals sustaining a permanent loss of Rb function retain a capacity to arrest in a quiescent state when deprived of growth factors, whereas fibroblasts in which an acute loss of Rb function was induced lose their capacity to arrest in G0 (Sage et al., 2003). Introduction Studies focused on delineating the function of individual members of the E2F family have identified The evolution of complex families of related activities is broad roles for several of the proteins.In particular, a common aspect of mammalian organisms.Functions E2F1–3 appear to function as transcriptional activators, that might be mediated by a single activity in Drosophila E2F4 and 5 appear to function as Rb-dependent or Caenorhabditis elegans are often seen to be encoded transcriptional repressors, and E2F6–8 may function by multiple related genes in the mammalian genome, as Rb-independent repressors of transcription (Dimova likely the result of duplications associated with unequal and Dyson, 2005).Nevertheless, attempts to clearly crossing over.These duplicated genes are then free to define specificity within the subfamilies, such as the evolve to acquire new functional roles, imparting greater E2F1–3 group, have been difficult.For instance, complexity of function within the mammalian organism. although the elimination of any single gene of this As a result, there is often considerable similarity group does not markedly affect cellular proliferation, double knockouts are impaired and the triple knockout Correspondence: Dr JR Nevins, Department of Molecular Genetics cannot grow (Wu et al., 2001). These results would thus and Microbiology, Institute for Genome Sciences and Policy, Duke suggest that these E2F proteins overlap in their University Medical Center, Durham, NC 27710, USA. function.Alternatively, they may possess quite distinct E-mail: [email protected] Received 26 January 2006; revised 8 May 2006; accepted 6 June 2006; functions but also have the capacity to compensate for published online 14 August 2006 the long-term loss of function as occurs in germline Compensation and specificity of function within the E2F family L-J Kong et al 322 knockouts.In an attempt to dissect the functional roles discrete phenotypes resulting from the long-term loss of of the individual E2F proteins, we have made use of E2F function.As an alternative approach to the study short hairpin RNA (shRNA) vectors that were gene- of E2F function, we have made use of shRNAs selected rated to target selectively either E2F1 or E2F3.We have to target either E2F1 or E2F3 to achieve an acute loss of then examined the consequence of an acute loss of E2F function.Each shRNA vector was inserted in a function of these two E2F proteins.The results suggest recombinant adenovirus to facilitate efficient expression that there are quite distinct functions of these proteins of the shRNA and loss of E2F function.Western that are revealed when cells only transiently lose their analysis demonstrates that the E2F1 shRNA specifically activity. reduces the level of the E2F1 protein without affect on the E2F3 level; likewise, the E2F3 shRNA specifically reduces E2F3, but not E2F1 protein levels (Figure 1b). Thus, the use of shRNAs allows for specific targeting of Results individual E2Fs and without evidence for compensatory effects from other family members. Various studies have highlighted a role for both E2F1 Given the previous results that no single E2F was and E2F3 in cell proliferation.Overexpression of either critical for cell cycle entry, but also the evidence for protein is sufficient to induce the S phase in otherwise compensation from long-term loss of function, we made quiescent cells (Cress et al., 1993; Johnson et al., 1993; use of the viruses expressing E2F1 or E2F3 shRNAs to Qin et al., 1994). Moreover, the combined loss of E2F1 evaluate the effects of acute loss of function of these and E2F3 does also impair the ability of cells to enter proteins on the process of cell cycle entry following the S phase, whereas knockout of either gene alone is serum stimulation.Cells were infected with control not sufficient to block the S phase (Humbert et al., 2000; virus, or the shRNA viruses, brought to a quiescent Wu et al., 2001). These results would suggest that either state by serum starvation for 48 h, and then stimulated E2F1 and E2F3 play overlapping roles in the transition to grow by the addition of serum.5-Bromo-2 0-deoxyur- into a cell cycle or that they are distinct, but the long- idine (BrdU) was added to the cultures 15 h after serum term loss of function of either protein leads to stimulation and the cells were harvested 3 h later.Each compensation by the other.In fact, an examination of virus expresses a green fluorescent protein (GFP) the levels of E2F proteins in mouse embryo fibroblasts protein to allow identification of virus-positive cells derived from knockout animals revealed that the level of and BrDU incorporation was detected by immunostain- E2F1 was increased in the E2F3 null MEFs, and ing.Many GFP-positive cells in the control virus likewise, the level of E2F3 was increased in the E2F1 infection were also positive for BrdU.In contrast, cells null MEFs (Figure 1a).These observations suggest the infected with either the shE2F1 virus or the shE2F3 possibility of a compensatory response to the loss of virus, as indicated by GFP, were negative for BrdU E2F function and might be a contribution to the lack of incorporation (Figure 2a). Quantitation of the results through examination of many fields revealed a greater than 90% reduction of BrdU incorporation in the E2F1 or E2F3 shRNA- expressing cells (Figure 2b).The same results were obtained with two independent shE2F1 and shE2F3 recombinant viruses, indicating that the result is not owing to nonspecific off-target effects of the shRNAs. The fact that each of the shRNAs blocks BrDU incorporation indicates that both E2F1 and E2F3 are important for driving quiescent cells into cell cycle, and thus must have distinct functions in this process.This is Figure 1 Effects of germline or acute loss of E2F function. in contrast to previously reported results using knockout (a) Compensation resulting from long-term loss of E2F1 or E2F3 cells, suggesting that the long-term loss of E2F1 or E2F3 function.Nuclear protein extracts were obtained from growing function can be compensated. / primary MEFs of wild type (WT), E2F1 null (E2F1À À) or E2F3 Previous work has suggested that once cells have null (E2F3À/À).Two cell lines are used for each E2F null MEFs. The wild-type and the corresponding E2F null MEFs are made out entered the cell cycle, there might be distinct require- of the embryos from the same female mouse.Proteins were ments for the E2F1 or E2F3 proteins.In particular, separated in 10% SDS–PAGE.E2F proteins were detected by E2F1 DNA-binding activity is detected only in the first Western blot using anti-E2F1or E2F3 antibodies.A parallel gel cycle out of quiescence, whereas E2F3 DNA-binding was probed with SP1 antibody to show equal loading of the proteins.( b) Capacity of shRNAs to eliminate E2F1 and E2F3 activity can be seen to oscillate during succeeding cell activity.Wild-type primary MEFs were infected with control, cycles (Leone et al., 1998). In addition, microinjection of siE2F1 or siE2F3 viruses.Cells were brought to a quiescent state by an E2F3 antibody decreased BrdU incorporation in cells serum starvation for 48 h after virus infection, and then stimulated that had been released from a cell cycle block, whereas to grow by the addition of serum for 16 h.Nuclear proteins were an E2F1 antibody injection did not impair BrdU isolated and E2F proteins were detected by Western blot.A parallel gel was probed with SP1 antibody to show equal loading of the incorporation (Leone et al., 1998). To address this proteins.Two independent siRNA constructs were used for each potentially distinct function in a more direct manner, we E2F. have assessed the effects of acute loss of E2F1 or E2F3

Oncogene Compensation and specificity of function within the E2F family L-J Kong et al 323 to be only necessary for the initial step of cell cycle entry. Previous work has made use of E2F overexpression to evaluate the nature of the target genes regulated by E2F activities.These studies have identified a large number of genes that are regulated by E2F proteins that encode proteins necessary for cell cycle progression, DNA replication and mitosis (Ishida et al., 2001; Moroni et al., 2001; Muller et al., 2001; Polager et al., 2002). Other studies have pointed to mechanisms for transcriptional specificity with the E2F proteins involving specific protein interactions that guide the formation of promoter complexes (Schlisio et al., 2002; Giangrande et al., 2003; Zhu et al., 2004). Nevertheless, a determina- tion of the precise nature of the gene regulatory pathways controlled by individual E2Fs has been difficult and we believe hampered by the fact that knockouts can obscure the true specificity of function owing to compensatory effects.Strategies employing expression of individual E2Fs to activate target genes are potentially compromised by the loss of specificity through overexpression.As such, we have made use of the shRNA viruses to eliminate acutely the function of either E2F1 or E2F3 and then examine the consequences Figure 2 Distinct roles for E2F1 and E2F3 in cell cycle entry and progression.( a) Both siE2F1 and siE2F3 block quiescent cells from on gene expression. entering cell cycle.REF52 cells were infected with control (top row), siE2F1 (middle row) or siE2F3 (bottom row) viruses at an MOI of 100.Cells were brought to a quiescent state by serum starvation for 48 h after virus infection, and then stimulated to grow by the addition of serum to 20%.BrDU was added to the cultures 15 h after serum stimulation, the cells were fixed 3 h later and stained with anti-BrDU antibody followed by Texas Red secondary antibody.BrDU-positive cells were seen as red, cells infected with virus were identified with GFP signal (green) and cell nuclei were shown by DAPI staining (blue).( b) Quantitation for the BrdU incorporation in serum-starvation-synchronized REF52 cells.REF52 cells were infected with control (u6-pro), siE2F1 or siE2F3 viruses at an MOI of 100 and BrDU-labeled as described in (a).Total and BrDU-positive cell numbers are counted.Between 450 and 700 cells were counted for each treatment.The percentages of BrDU-positive cells in all the cells and virus-positive cells are calculated.

function in growing cells.Asynchronous cells were infected with control, shE2F1 or shE2F3 viruses for 60 h, followed by 3 h BrDU labeling.The cells are ﬁxed and stained with anti-BrDU antibody.As shown in Figure 3a, many GFP-positive cells in the control virus infection were also positive for BrdU.Likewise, many GFP-positive cells infected with the shE2F1 virus were also BrdU positive.In contrast, the GFP-positive cells Figure 3 Distinct roles for E2F1 and E2F3 in cell cycle entry and infected with the shE2F3 virus did not incorporate progression.( a) siE2F3 but not siE2F1 blocks DNA synthesis in BrdU.Again, quantitation of the results through cycling cell.Ansynchronized REF52 cells were infected with examination of many ﬁelds revealed a greater than control (top row), siE2F1 (middle row) or siE2F3 (bottom row) viruses at an MOI of 100 for 60 h and then labeled with BrDU for 90% reduction of BrdU incorporation in the E2F3 3 h.Virus infection, BrDU incorporation and nuclei staining were shRNA-expressing cells, but no reduction of BrdU carried out similarly to Figure 2a.( b) Quantitation for the BrdU incorporation in the E2F1 shRNA-expressing cells incorporation in cycling cells.REF52 cells were infected with (Figure 3b).It would thus appear that E2F3 is required control (u6-pro), siE2F1 or siE2F3 viruses at an MOI of 100 and BrDU-labeled as described in (a).Total and BrDU-positive cell for both the initial entry of cells into the cell cycle from a numbers are counted.Between 450 and 700 cells were counted for quiescent state as well as the ability to enter the S phase each treatment.The percentages of BrDU-positive cells in all the during each cell cycle.E2F1, on the other hand, appears cells and virus-positive cells are calculated.

Oncogene Compensation and specificity of function within the E2F family L-J Kong et al 324 Given the apparent role for both E2F1 and E2F3 in the entry of cells into the cell cycle following growth stimulation, and the assumption that they likely play distinct roles in this process, we have focused on the E2F-mediated gene regulatory events associated with cell cycle entry following growth stimulation.Cells were infected with the two shRNA viruses, brought to quiescence, and then stimulated by the addition of serum.Cells were harvested 16 h after stimulation and used for the preparation of RNA and then microarray analysis with Affymetrix GeneChip arrays.To focus on those genes likely regulated by E2F1 or E2F3, we first identified the genes whose expression changed at least 1.3-fold as a function of growth stimulation in cells infected with control virus.Then, within this filtered group, we identified those genes whose expression was altered at least 1.3-fold by the presence of either the E2F1 or E2F3 shRNA.These selections yielded a collection of 149 genes whose expression pattern is depicted in Figure 4a.One point immediately obvious from this analysis is the relatively small number of genes affected by the loss of E2F1 activity relative to the number affected by the loss of E2F3.Although it is possible that this could reflect an inefficiency of the E2F1 shRNA relative to the E2F3 shRNA, this appears unlikely based on the change of both of the two E2F1 probes in Affymetrix GeneChip arrays and the Western blots for E2F1 and E2F3 protein levels (Figure 4b). Not only is there a distinct difference in the number of genes affected by the loss of E2F1 or E2F3 but also the nature of the genes (Supplementary Table 1).The genes affected by the loss of E2F3 activity include many of the previously identified E2F targets and, in particular, those encoding DNA replication, DNA repair, cell cycle regulation and mitotic activities.This category of genes is largely absent from those affected by E2F1 loss of function.A quantitative assessment of the biological functions represented in the E2F1- and E2F3-regulated genes is provided in Supplementary Figure 1, where it is seen that Gene Ontology (GO) categories reflecting mitotic functions (M phase, nuclear division, mitosis, Figure 4 Gene expression alterations resulting from acute loss of E2F1 or E2F3 activity.( a) Microarray expression analysis.A total M phase of mitotic cell cycle) are significantly enriched of 149 genes were selected based on an increased expression of at for the E2F3-regulated genes.In contrast, the genes least 1.3-fold after serum stimulation and decreased expression of affected by loss of E2F1 activity do not show an at least 1.3-fold owing to the loss of E2F1/3 activity. The genes enrichment for this group of genes.Perhaps most were ordered according to their response to shE2F1, so that the genes most sensitive to the loss of E2F1 occur at the top of the plot. important is the observation that the genes identified The first three columns show the gene expression levels in the three through the acute loss of E2F1 or E2F3 function replicates of the samples infected with control virus without serum, appears to define a much more discrete and distinct the next two columns show the levels after the addition of serum, grouping than previously seen in the analysis of over- the following three include cells infected with siE2F1 virus and the expression or long-term loss of function. final two with siE2F3 virus.( b) Effective knockdown of E2Fs in the samples that used for microarry analysis.Wild-type MEFs were infected with control (left five lanes), siE2F1 (top row right six lanes) or siE2F3 (bottom row right five lanes) viruses, serum starved for 48 h and then stimulated by the addition of 20% serum. Discussion Cells were harvested 0 h and 16 h after stimulation.E2F protein levels were evaluated by Western blot using anti-E2F1 (top row) or anti-E2F3 (bottom row) antibodies.The protein levels of the E2Fs The E2F family of transcriptional regulatory proteins are measured using Quantity One software from BioRad.The has been shown to be critical for proper regulation of efficacies of siE2F1 knockdown were 85.3, 84.3 and 85.0% as cell proliferation (Dyson, 1998; Nevins, 1998).The measured by signal intensity.Those for siE2F3 knockdown were deregulation of E2F activity, either as a result of cyclin 88.2 and 86.4%. D amplification, loss of p16 activity or loss of Rb function, is a general property of the oncogenic process

Oncogene Compensation and specificity of function within the E2F family L-J Kong et al 325 and contributes to the development of most if not all requirement of E2F3 for the S-phase entry both from human cancers (Sherr, 1996; Nevins, 2001).The com- quiescence as well as during the cell cycle, as these plexity of E2F function is substantial in mammals, activities are required each time a cell enters the S phase. composed of 10 distinct gene products encoded by eight The genes specifically regulated by E2F1, that might be independent loci, that can be divided into three expected to play a role in the initial cell growth response, subfamilies based on their sequence homology – the do not define such a clear relationship, although several E2F1–3 genes, the E2F4 and 5 genes and the E2F6–8 represent GO categories of cellular metabolism (Supple- genes.This sequence-based division also reflects their mentary Figure 1), perhaps reflecting a role in generat- functional properties.For example, the E2F1–3 genes ing components necessary for cell growth. are tightly regulated by cell growth and during cell cycle, In summary, the results presented here, and taken in whereas the E2F4–8 genes are constitutively expressed. the context of a variety of previous studies, suggest that Furthermore, E2F1–3 act as positive regulators of the evolution of the E2F family of transcription factors transcription, whereas E2F4 and E2F5, when bound has left considerable overlap in the function of the to p130 or Rb, act as transcriptional repressors.The individual proteins.Loss of function of a single family E2F6–8 proteins appear to function as Rb-independent member can be compensated by a related protein transcriptional repressors (Attwooll et al., 2004). allowing survival of the organism.Nevertheless, our Although these broad functional distinctions have findings demonstrate that individual family members do been fairly well established, the specific functional roles play critical roles in cell proliferation, coinciding with and distinctions among the E2F family members have specific aspects of gene regulation.It is the ability to been more difficult to assess.Generally, E2F1–3 appear inactivate acutely the function of members of complex to be most important for allowing and maintain cell gene families that provides an opportunity for dissecting proliferation, E2F4 and 5 appear to be important for otherwise overlapping functions and has revealed coupling cell cycle with decisions of cell differentiation distinct roles for two of the E2F proteins, both in the and quiescence, and E2F1 has been linked with the context of cell proliferation control as well as regulation induction of apoptosis.Nevertheless, these distinctions of specific groups of genes associated with these have been blurred in many instances with apparent processes. overlaps seen in the function of the individual E2Fs.The work we present here suggests that at least some of this overlap might be ascribed to methods of analysis that Materials and methods have employed germline gene disruptions as the primary method of assessing function.Our work strongly shRNA virus construction and infection suggests that the long-term loss of E2F function can Two of the small-interfering RNA (siRNA) target sequences result in compensatory mechanisms arising to ensure a were selected for each gene of interest.The target sequences for viable cell or organism.Because of the compensation, E2F1 siRNAs are as follows: GACTGTGACTTTGGG the discrete roles of the individual E2F proteins are GACCT and GATCTCCCTTAAGAGCAAAC.The target often obscured.In contrast, the use of methods to sequences for E2F3 siRNAs are: GACTTCATGTGTAGTT achieve acute loss of function appears to more clearly GATT and GGTTAGCTTATGTTACATAT.These sequences were cloned as hairpin structures under U6 promoter define roles for individual E2F proteins. as described earlier (Yu et al., 2002), and then constructed into The studies presented here now point to functional Adtrack vector (He et al., 1998). U6 promoter was cloned into distinctions between the E2F1 and E2F3 proteins not Adtrack vector to make empty control virus.Recombinant appreciated previously.First, although studies of knock- viruses are produced, purified and titrated according to outs generally concluded that no single E2F was published protocols (Nevins et al., 1997; He et al., 1998). essential for cell cycle entry following growth stimula- Virus infections were carried out as described earlier by Cook tion (Humbert et al., 2000; Wu et al., 2001), the acute et al.(2002). loss of either E2F1 or E2F3 blocked the S phase following stimulation of quiescent cells.Thus, it would Western analysis appear that each of these E2F proteins perform essential Primary MEFs were isolated and cultured as described by but distinct roles in this process.In contrast, only E2F3 Rempel et al.(2000).Nuclear protein extracts were obtained as appeared to be essential for entry to the S phase once described earlier by Leone et al.(1998).A total of 8 mg nuclear cells had entered the growth cycle.Thus, E2F1 appears protein extracts were separated in 10% sodium dodecyl to be important for a cell cycle entry process, whereas sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). E2F proteins were detected by immunoblot using anti-E2F1 E2F3 is important for events that occur at this time as (c20) or E2F3 (c18) polyclonal antibodies (Santa Cruz well as subsequent cell cycles.Second, the nature of the Biotechnology, Santa Cruz, CA, USA) at 1:1000 or 1:2000 gene expression patterns influenced by the acute loss of dilutions, respectively. E2F1 or E2F3 activity was sharply distinct.E2F3- dependent genes included many of the DNA replication, BrDU incorporation mitotic and cell cycle regulatory genes (Supplementary REF52 cells were cultured in Dulbecco’s modified Eagle’s Table 1 and Supplementary Figure 1), whereas E2F1 medium (DMEM) with 5% fetal bovine serum (FBS) and 5% did not appear to be involved in the expression of bovine calf serum, infected with control virus, or the shRNA these genes.Importantly, a role for E2F3 in the control viruses at relative multiplicity of infection (MOI) of 100.To of these genes would thus be consistent with the evaluate cell cycle entry, cells were brought to a quiescent state

Oncogene Compensation and specificity of function within the E2F family L-J Kong et al 326 by serum starvation for 48 h after virus infection, and then in gene expression analyses, we converted the log values that stimulated to grow by the addition of serum.BrDU was added RMA yielded back into a normal scale for further calculations to the cultures 15 h after serum stimulation at final concentra- by taking its exponentiation with a base of two.As each tion of 10 mM, and the cells were harvested 3 h later.For experiment was repeated multiple times, we combined repli- cycling cells, all the experiments were carried out similarly, cates and calculated a single average expression value for each except that we infected asynchronous cells with viruses for 60 h gene in an experimental condition.The fold-response of a gene followed by 3 h BrDU labeling.Cells were fixed and BrDU to serum is the averaged value of that gene 16 h after serum incorporation was visualized by immunostaining as described stimulation divided by the value in quiescence.The fold- by Smith et al.(2000).Cells infected with virus were identified response of a gene to loss of E2F1 or E2F3 is its averaged with GFP signal. value in cells treated with shE2F1/3 16 h after serum stimulation divided by its value in untreated cells at the same RNA isolation and microarray analysis time point.Thus, genes that are normally induced by E2F1 will Primary wild-type MEFs were trypsinized and infected with receive low scores (less than one) because its values are lower in control, siE2F1 or siE2F3 viruses at relative MOI of 100.For cells containing the shE2F1 construct.Its expression is lost each infection, 8 Â 106 cells were incubated with virus for 1 h in concomitant to loss of E2F1 expression.For generation of the 300 ml DMEM and plated into DMEM with 0.2% FBS for color bar shown in Figure 3a, the change in expression (RMA 48 h, and then stimulated by the addition of 20% serum.Cells normalized data) of aforementioned selected genes in serum- were harvested 16 h after stimulation.We used half of the cells stimulated cells (control, siE2F1, siE2F3) was compared to the for nuclear protein extraction to evaluate E2F protein levels by average expression of the respective genes in quiescent cells. Western blots, and the other half to isolate RNA.RNA was Matlab was used to visualize the results from this analysis. prepared using RNeasy kit as described by the manufacturer (Qiagen, Valencia, CA, USA) for microarray analysis with Affymetrix GeneChip arrays MOE430A as described earlier by Acknowledgements Ishida et al.(2001). We thank Drs David Turner and Yang Shi for providing U6 Analysis of microarray data promoter and technical advices.We thank Dr Rachel Rempel We computed and normalized the gene expression values from for providing MEFs.This work was supported by grants from the microarrays using the Robust Multichip Average (RMA) the NIH (CA104663, CA106520, CA112952).Finally, we algorithm provided by the Bioconductor package (Irizarry thank Kaye Culler for her assistance in the preparation of the et al., 2003; Gentleman et al., 2004). For analyses of fold-change manuscript.

References

Attwooll C, Denchi EL, Helin K.(2004).The E2F family: Ishida S, Huang E, Zuzan H, Spang R, Leone G, West M et al. specific functions and overlapping interests. The EMBO J (2001).Role for E2F in the control of both DNA replication 23: 4709–4716. and mitotic functions as revealed from DNA microarray Cook JG, Park CH, Burke TW, Leone G, DeGregori J, Engel analysis. Mol Cell Biol 21: 4684–4699. A et al. (2002).Analysis of Cdc6 function in the assembly of Johnson DG, Schwarz JK, Cress WD, Nevins JR.(1993). mammalian prereplication complexes. Proc Natl Acad Sci Expression of transcription factor E2F1 induces quiescent USA 99: 1347–1352. cells to enter S phase. Nature 365: 349–352. Cress WD, Johnson DG, Nevins JR.(1993).A genetic analysis Leone G, DeGregori J, Yan Z, Jakoi L, Ishida S, Williams RS of the E2F1 gene distinguishes regulation by Rb, p107, and et al. (1998).E2F3 activity is regulated during the cell cycle adenovirus E4. Mol Cell Biol 13: 6314–6325. and is required for the induction of S phase. Genes Dev 12: Dimova DK, Dyson NJ.(2005).The E2F transcriptional 2120–2130. network: old acquaintances with new faces. Oncogene 24: Moroni MC, Hickman ES, Denchi EL, Caprara G, Colli E, 2810–2826. Cecconi F et al. (2001).Apaf-1 is a transcriptional target for Dyson N.(1998).The regulation of E2F by pRB-family E2F and p53. Nat Cell Biol 3: 552–558. proteins. Genes Dev 12: 2245–2262. Muller H, Bracken AP, Vernell R, Moroni MC, Christians F, Gentleman RC, Carey VJ, Bates DM, Bolstad BM, Dettling M, Grassilli E et al. (2001).E2Fs regulate the expression of Dudoit S et al. (2004).Bioconductor: open software genes involved in differentiation, development, proliferation, development for computational biology and bioinformatics. and apoptosis. Genes Dev 15: 267–285. Genome Biol 5: R80. Nevins JR.(1998).Toward an understanding of the functional Giangrande PH, Hallstrom TC, Tunyaplin C, Calame K, complexity of the E2F and Retinoblastoma families. Cell Nevins JR.(2003).Identification of the E box factor TFE3 Growth Differ 9: 585–593. as a functional partner for the E2F3 transcription factor. Nevins JR.(2001).The Rb/E2F pathway and cancer. Hum Mol Cell Biol 23: 3707–3720. Mol Genet 10: 699–703. He T-C, Zhou S, da Costa LT, Yu J, Kinzler KW, Vogelstein Nevins JR, DeGregori J, Jakoi L, Leone G.(1997).Functional B.(1998).A simplified system for generating recombinant analysis of E2F. Meth Enzymol 283: 205–219. adenoviruses. Biochem Biophys Res Comm 95: 2509–2514. Polager S, Kalma Y, Berkovich E, Ginsberg D.(2002).E2Fs Humbert PO, Verona R, Trimarchi JM, Rogers C, Dandapani S, up-regulate expression of genes involved in DNA replica- Lees JA.(2000).E2f3 is critical for normal cellular pro- tion, DNA repair and mitosis. Oncogene 21: 437–446. liferation. Genes Dev 14: 690–703. Qin X-Q, Livingston DM, Kaelin WG, Adams PD.(1994). Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Deregulated transcription factor E2F-1 expression leads to Speed TP.(2003).Summaries of Affymetrix GeneChip probe S-phase entry and p53-mediated apoptosis. Proc Natl Acad level data. Nucleic Acids Res 31: e15. Sci USA 91: 10918–10922.

Oncogene Compensation and specificity of function within the E2F family L-J Kong et al 327 Rempel RE, Saenz-Robles MT, Storms R, Morham S, Smith DS, Leone G, DeGregori J, Ahmed MN, Qumsiyeh MB, Ishida S, Engel A et al. (2000).Loss of E2F4 activity leads Nevins JR.(2000).Induction of DNA replication in adult to abnormal development of multiple cellular lineages. rat neurons by deregulation of the retinoblastoma/E2F G1 Mol Cell 6: 293–306. cell cycle pathway. 11: 625–633. Sage J, Miller AL, Perez-Mancera PA, Wysocki JM, Jacks T. Wu L, Timmers C, Maiti B, Saavedra HI, Sang L, Chong GT (2003).Acute mutation of retinoblastoma gene function is et al. (2001).The E2F1–3 transcription factors are essential sufficient for cell cycle re-entry. Nature 424: 223–228. for cellular proliferation. Nature 414: 457–462. Schlisio S, Halperin T, Vidal M, Nevins JR.(2002).Interaction Yu JY, DeRuiter SL, Turner DL.(2002).RNA interference by of YY1 with E2Fs, mediated by RYBP, provides a expression of short-interfering RNAs and hairpin RNAs in mechanism for specificity of E2F function. EMBO J 21: mammalian cells. Proc Natl Acad Sci USA 99: 6047–6052. 5775–5786. Zhu W, Giangrande P, Nevins JR.(2004).E2Fs link the Sherr CJ.(1996).Cancer cell cycles. Science 274: 1672–1677. control of G1/S and G2/M. EMBO J 23: 4615–4626.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene