Identification of E2F-3B, an alternative form of E2F-3 lacking a conserved N-terminal region

Abstract

We have identified a novel form of the full-length E2F-3 protein that we term E2F-3B. In contrast to full-length E2F-3, which is expressed only at the G1/S boundary, E2F-3B is detected throughout the cell cycle with peak levels in G0 where it is associated with Rb. Transfection and in vitro translation experiments demonstrate that a protein identical to E2F-3B in size and iso-electric point is produced from the E2F-3 mRNA via the use of an alternative translational start site. This alternative initiation codon was mapped by mutagenesis to codon 102, an ACG codon. Mutation of the ACG codon at position 102 abolished E2F-3B expression, whereas the conversion of ACG 102 to a consensus ATG led to the expression of a protein indistinguishable from E2F-3B. Given these results, E2F-3B is missing 101 N-terminal amino acids relative to full-length E2F-3. This region includes a moderately conserved sequence of unknown function that is present only in the growth-promoting E2F family members, including E2F-1, 2 and full-length E2F-3. These observations make E2F-3B the first example of an E2F gene giving rise to two different protein species and also suggest that E2F-3 and E2F-3B may have opposing roles in cell cycle control.

Introduction

E2F refers to a family of transcriptional regulatory proteins that appear to serve dual roles in cellular growth control. E2F proteins bind DNA as heterodimers with members of the related DP family of proteins (Bandara et al., 1993; Helin et al., 1993; Rogers et al., 1996; Wu et al., 1995; Zhang and Chellappan, 1995). As heterodimers, the E2F/DP proteins serve to drive transcription of genes required for DNA synthesis and cell cycle progression. In contrast, when E2F/DP heterodimers form complexes with members of the Rb family, they serve as transcriptional repressors and inhibit cell growth (for recent reviews see Dyson (1998); Johnson and Schneider-Broussard (1998); Nevins (1998)). Six members of the E2F protein family have been identified (Beijersbergen et al., 1994; Buck et al., 1995; Ginsberg et al., 1994; Helin et al., 1992; Hijmans et al., 1995; Itoh et al., 1995; Ivey-Hoyle et al., 1993; Kaelin et al., 1992; Lees et al., 1993; Morkel et al., 1997; Sardet et al., 1995; Shan et al., 1992; Trimarchi et al., 1998). All E2F proteins share homologous regions including a central DNA binding domain, a DP dimerization domain, a transcriptional activation domain (with the exception of E2F-6) and within the transcriptional activation domain a region that binds to a member of the Rb family.

The six E2F proteins can be divided into three subgroups based upon structural and functional distinctions. E2F-1, 2, and 3 share a moderately conserved hydrophobic N-terminal region (NTR) of approximately 100 amino acids of unknown function. E2F-1, -2 and -3 also share a cyclin A binding domain (Kitagawa et al., 1995; Krek et al., 1994; 1995; Xu et al., 1994), a nuclear localization sequence (Allen et al., 1997; Lindeman et al., 1997; Magae et al., 1996; Muller et al., 1997; Verona et al., 1997) and an Sp1-binding domain (Karlseder et al., 1996; Lin et al., 1996; Shin et al., 1996) that are not present in other members of the E2F family. E2F-1, 2 and 3 bind exclusively to pRb (in vivo), but not to the other Rb family members, p107 and p130 (Lees et al., 1993). In contrast, E2F-4 and 5 lack an extended NTR and interact with all three members of the Rb family (Beijersbergen et al., 1994; Ginsberg et al., 1994; Hijmans et al., 1995; Wu et al., 1995). E2F-6 alone represents the third E2F subgroup and appears to be a transcriptional repressor independent of an interaction with Rb (Cartwright et al., 1998; Gaubatz et al., 1998; Morkel et al., 1997; Trimarchi et al., 1998). Structurally, E2F-6 lacks a transcriptional activation domain and an interaction domain for members of the Rb family.

Emerging data suggest that individual members of the E2F family may play either a primary role in growth stimulation or in growth inhibition. First, E2F family subgroups differ markedly in their pattern of expression. E2F-4 and 5 are expressed constitutively, whereas the expression of E2F-1, 2 and 3 is restricted to cells at the G1/S boundary and into the S phase (Flores et al., 1998; Hsiao et al., 1994; Ikeda et al., 1996; Johnson et al., 1994; Leone et al., 1998; Moberg et al., 1996; Neuman et al., 1994; Sardet et al., 1995; Sears et al., 1997). Second, ectopic expression suggests fundamentally different biological activities for individual E2F proteins. For example, overexpression of either E2F-1, 2, or 3 in quiescent REF52 fibroblasts promotes S phase entry; whereas overexpression of E2F-4 or E2F-5 does not or does so less efficiently (DeGregori et al., 1997; Johnson et al., 1993; Lukas et al., 1996; Mann and Jones, 1996). In addition, in vivo E2F-1 is a much stronger transcriptional activator than either E2F-4 or -5 (Pierce et al., 1998) and in vitro E2F-1 selects a broader range of DNA sequences than E2F-4 and binds more avidly to all sequences examined (Tao et al., 1997). These results potentially explain why E2F-1 induces cell growth efficiently. In contrast, E2F-4 and 6 may specialize in growth inhibition since the overexpression of E2F-4 has recently been shown to promote the differentiation of nerve growth factor-treated PC12 cells (Persengiev et al., 1999). Similarly, the over-expression of E2F-6 appears to block cell growth progression of stimulated fibroblasts (Cartwright et al., 1998; Gaubatz et al., 1998).

In the present work, we have identified a novel member of the E2F protein family which appears to be an alternatively initiated form of the E2F-3 protein, designated E2F-3B. In contrast to growth-promoting E2F family members, including full-length E2F-3, E2F-3B is produced at its highest level in quiescent cells where it is associated with Rb. Interestingly, several lines of evidence presented here suggest that E2F-3B is produced by alternative translational initiation at codon 102, an ACG. Thus, E2F-3B is missing the entire N-terminal region of the growth-promoting members of the E2F family.

Results

Deoxycholate treatment reveals a novel form of E2F-3

In previous work (Dong et al., 1998; Flores et al., 1998), we demonstrated that the dominant E2F complex detected by electrophoretic mobility shift assay (EMSA) in density-arrested fibroblasts contained E2F-4 and the p130 protein. However, in mouse fibroblasts the E2F-Rb complex is not apparent in EMSAs (Hurford et al., 1997). To address the possibility that forms of E2F were present, but not apparent in our EMSAs, extracts of cells in the G0 and G1/S phases of the cell cycle were treated with deoxycholate (DOC). DOC treatment disrupts E2F/RB complexes and releases free E2F that binds DNA in EMSAs (Chellappan et al., 1991). The EMSA shown in Figure 1a demonstrates that treatment of a BALB/c-3T3 GO nuclear extract with DOC resulted in the generation (or enhancement) of a single predominant E2F band not characteristic of any E2F complexes that we have defined in the past. Antibody supershift experiments were performed (Figure 1a, lanes 3-6) demonstrating that this novel form of E2F was abolished by the addition of an antibody directed against the E2F-3 C-terminal 18 amino acids. This result was not expected because we (Flores et al., 1998) and others (Leone et al., 1998; Moberg et al., 1996) have demonstrated that the full-length E2F-3 protein was not expressed at a measurable level in quiescent or G1 cells as assayed by Western blot. While DOC-treatment of G1/S extracts (Figure 1b) did not reveal any complexes not already apparent in untreated extracts, Figure 1B does make it clear that there are two complexes in G1/S cells recognized by the E2F-3 C-terminal antibody (compare lanes 2 and 4 of panels a and b). The slower moving E2F-3 band corresponds to full-length E2F-3 which we have previously described (Dong et al., 1998; Flores et al., 1998). The faster moving E2F-3 band was designated E2F-3B.

Figure 1
figure1

Treatment with DOC reveals a novel form of E2F present in BALB/c-3T3 G0 cells. (a) Nuclear extracts of density-arrested BALB/c-3T3 (G0) cells were treated with 0.2% DOC and 20 μg subjected to EMSA. Lane 1, untreated BALB/c-3T3 G0 extract. Lane 2, DOC-treated BALB/c-3T3 G0 extract. Lanes 3–6, antibodies were added to DOC-treated G0 extract to identify complexes. Lane 3, E2F-4 antibody (sc-866X). Lane 4, E2F-3 antibody (C18, sc-878X). Lane 5, E2F-2 antibody (sc-632X). Lane 6, E2F-1 antibody (sc-193X). The positions of putative complexes are indicated. ?refers to an unidentified complex(es) in this and subsequent figures. This film was exposed approximately 48 h. (b) The same as for (a) except extracts of BALB/c-3T3 cells following 15 h PDGF stimulation (85% G1/S phase) were used. (c) Whole cell extracts of cells transiently overexpressing specific E2F/DP-1 heterodimers were compared with E2F-3B in an EMSA. Lane 1, DOC-treated BALB/c-3T3 G0 extract (predominantly E2F-3B). Lane 2, extract of C-33A cells transfected with 20 μg empty pcDNA3 vector extract (mock), Lanes 3–8, extracts of C-33A cells transfected with 10 μg of pcDNA3 vector containing E2F-1, 2, 3, 4, 5 or E2F-3NT cDNAs cotransfected with 10 μg CMV-neo-HA-DP-1. This film was exposed approximately 16 h. Twenty μg of protein were loaded in each lane. E2F-2 was very poorly expressed in this experiment, but in other experiments it comigrates with E2F1 (not shown)

The native electrophoretic mobility of E2F-3B was compared with E2F-1 through 5 by transfecting plasmids that expressed these proteins (together with DP-1) into C-33A cells (Figure 1c). An additional construct, E2F-3NT, which expresses an artificial N-terminally truncated form of E2F-3 (see Figure 2a for a schematic) was also included (Lees et al., 1993). Complexes containing full-length E2F-1, 2 and 3 migrated with indistinguishable mobilities, E2F-4 migrated more slowly while E2F-5 complexes migrated more quickly. Figure 1c reveals that E2F-3B (lane 1) migrated very similar to E2F-3NT and much more rapidly than do the full-length proteins.

Figure 2
figure2

E2F-3B is recognized by an E2F-3 C-terminal antibody, but not by an E2F-3 N-terminal antibody. (a) Schematic structure of E2F-3NT compared with E2F-3. Various known functional domains and regions are indicated. Cln A and NLS refer to overlapping cyclin A-binding and a nuclear localization sequences. Antibody epitopes are indicated. A putative structure for E2F-3B is proposed (justified below). (b) DOC-treated BALB/c-3T3 G0 extracts, overexpressed E2F-3NT and E2F-3 were subjected to antibody supershift assay. Antibodies were added as indicated (1 μl). C is an antibody directed against the E2F-3 C-terminal 18 amino acids (sc-878x) and N is an antibody directed against the E2F-3 N-terminal 20 amino acids (sc-879x). This film was exposed approximately 16 h. Twenty μg of protein was loaded in each lane. (c) BALB/c-3T3 G0 extracts were treated with DOC to release E2F-3B and the indicated antibodies added. Lane 1, no antibody. Lane 2, N-terminal E2F-3 antibody. Lane 3, C-terminal E2F-3 antibody. Lane 4, antibody 1372 raised against E2F-3 amino acids 351–414. Lane 5, antibody 1373 raised against E2F-3 amino acids 351–414. Lanes 6 and 7, preimmune antibody preparation for 1372 and 1373, respectively. This film was exposed approximately 48 h. Twenty μg of protein was loaded in each lane

To test the hypothesis that E2F-3B could represent an E2F-3 protein lacking its N-terminal region we used an antibody directed against the E2F-3 N-terminal 20 amino acids that should not recognize an N-terminally truncated form (see Figure 2a for a schematic) side by side with the C-terminal antibody. The data in Figure 2b reveal that the N-terminal E2F-3 antibody recognized neither E2F-3B nor E2F-3NT. In contrast, the N-terminal antibody quantitatively abolished full-length E2F-3, demonstrating that the N-terminal antibody was fully functional. To exclude the possibility that the C-terminal antibody cross-reacts with a novel member of the E2F family additional antibodies were generated specific to a region of E2F-3 (highlighted in Figure 2a) that is not conserved within the E2F family. Figure 2c demonstrates that antibody preparations 1372 and 1373 both quantitatively shifted E2F-3B, whereas, preimmune antibody preparations had no effect. In control experiments, the E2F-3 antibodies 1372 and 1373 did not cross-react with other members of the E2F family, but each did quantitatively abolish full-length E2F-3 (not shown).

E2F-3B is missing the NTR characteristic of the growth-promoting members of the E2F family and associates with Rb in G0

A Western blot was performed on extracts from BALB/c-3T3 G0 and G1/S cells using the C-terminal antibody as a probe to determine if E2F-3B could be detected. Full-length E2F-3 was translated in vitro and included in the Western blot as a control. As shown in Figure 3a, G0 cells possessed an immunoreactive band at a lower molecular weight than full-length E2F-3. This band diminished in G1/S cells, although it did not disappear. As predicted from the EMSA results and reported earlier (Flores et al., 1998; Moberg et al., 1996), G0 cells had little full-length E2F-3, whereas cells at G1/S had a dramatic increase in full-length E2F-3. The mobility of E2F-3B suggested that it was missing a significant portion of the full-length E2F-3 protein. To obtain a more accurate estimate of the point of deletion the mobility of E2F-3B was compared with in vitro translated E2F-3 NT. Comparing the mobilities of E2F-3B (from a BALB/c-3T3 G0 extract) and E2F-3NT (Figure 3b lanes 2 and 3) revealed that the two proteins are very similar in size. Since E2F-3NT begins translation at amino acid 132 of E2F-3, this result suggests that E2F-3B is missing approximately 130 N-terminal amino acids.

Figure 3
figure3

Western blot analysis indicates that E2F-3B is expressed at its highest levels in G0, lacks the entire N-terminal region of full-length E2F-3 and associates with Rb. (a) Extracts of BALB/c-3T3 G0 cells, cells in G1/S and 1 μl of in vitro translated (IVT) full-length E2F-3 were compared by Western blot using the C-terminal E2F-3 antibody as probe. Mol wt markers are indicated. Lanes 2 and 3, 100 μg of protein was loaded in each lane. (b) The relative electrophoretic mobilities of E2F-3B, IVT E2F-3 and IVT E2F-3NT were determined by Western blot using the C-terminal E2F-3 antibody as probe. (c) To enrich extracts for E2F DNA binding activity a DNA pull down on 200 μg of protein was employed prior to Western blot with C-terminal and N-terminal E2F-3 antibodies. Lane 1, one μl of IVT full-length E2F-3. Lanes 2 and 4, control DNA pull down excluding biotinylated oligonucleotide (−). Lanes 3 and 5, DNA pull down including biotinylated oligonucleotide (+). Lanes 1, 4 and 5 were probed with N-terminal antibody. Lanes 2 and 3 were probed with C-terminal antibody. (d) Extracts (200 μgs) of BALB/c-3T3 G0 and G1/S cells were subjected to immunoprecipitations with Rb (14051, Pharmingen) and DP-1 (MS-255 and MS-266, NeoMarkers) monoclonal antibodies. Immunoprecipitates were subjected to Western blots using the C-terminal E2F-3 antibody as probe. Lane 1, G0 extract, Rb IP. Lane 2, G1/S extract, Rb IP. Lane 3, G0 extract, DP-1 (255) IP. Lane 4, G1/S extract DP-1(255) IP. Lane 5, G0 extract DP-1 (266) IP. Lane 6, G1/S extract DP-1 (266) IP. Lane 7, 1 μl of in vitro-translated full-length E2F-3

Western blots using the N-terminal E2F-3 antibody were not initially informative, since this antibody cross-reacted with other proteins in crude extracts. To prepare partially purified E2F, and thus eliminate cross-reacting proteins, a DNA pull down was employed. A 5′-biotinylated oligonucleotide and a complementary oligonucleotide were used to form a double-stranded DNA containing a high affinity E2F binding site. This biotinylated oligonucleotide probe was incubated with DOC-treated extracts of asynchronous cells to allow binding to E2F (asynchronous cells have both E2F-3 and 3B). The biotinylated DNA/E2F complex was recovered using streptavidin-agarose beads. The partially purified E2F fraction was eluted from the beads and subjected to Western blot using the C-terminal and N-terminal antibodies in parallel. The data in Figure 3c reveal that the C-terminal E2F-3 antibody recognized two forms of E2F-3, the full-length protein and E2F-3B. The N-terminal E2F-3 antibody recognized only the full-length form.

It has been established that E2F-1, 2, and 3 do not interact with p107 or p130 in vivo, and thus, it is likely that in G0 cells E2F-3B is bound to Rb and that the complex simply does not appear in the EMSA. To determine if this were so, immunoprecipitations were performed using a Rb monoclonal antibody followed by Western blot using the C-terminal E2F-3 antibody to detect both E2F-3 and E2F-3B. Two DP-1 monoclonal antibodies served as controls since DP-1 should interact with both E2F-3 and E2F-3B in G0 and in S. Figure 3d reveals that E2F-3 was associated with Rb in BALB/c-3T3 G0 cells, but that the association was lost during G1/S (consistent with a model in which phosphorylation of Rb abolishes the Rb/E2F-3B interaction during S phase). In contrast, both DP-1 antibodies co-immunoprecipitated E2F-3B in BALB/c-3T3 G0 and E2F-3B and full-length E2F-3 in G1/S. We were unable to detect any association of E2F-3B with p130 or p107 (data not shown).

E2F-3B protein is expressed in differentiated human cells, whereas full-length E2F-3 protein expression is down regulated without apparent change in the E2F-3 message

To determine if E2F-3 and E2F-3B were expressed in human cells E2F-3 and E2F-3B expression was measured in HL60 cells (a human promyelocytic leukemia cell line) that can be induced to differentiate into macrophage-like cells (Ikeda et al., 1996) by the addition of PMA. Twenty-four hours following PMA, 80–90% of the cells had differentiated into macrophage-like cells and adhered to the tissue culture flasks. Flow cytometry revealed that 38% of the cells were in S phase prior to PMA treatment, whereas adherent PMA-treated cells displayed less than 2% of cells in S phase (data not shown).

Whole cell extracts (Ikeda et al., 1996) were prepared from growing and differentiated cells and subjected to Western blot (as controls we also included Western blots of E2F-1 and 4). Figure 4a reveals that both E2F-3 and E2F-3B were present in asynchronously growing cells, whereas only E2F-3B was expressed in the differentiated cells (a pattern identical to the G0 versus G1/S expression pattern in mouse fibroblast). The E2F-1 polypeptide disappeared in differentiated cells and E2F-4 levels increased 2–3-fold (the multiple E2F-4 species were differentially phosphorylated forms of E2F-4, data not shown).

Figure 4
figure4

E2F-3B is expressed in differentiated HL60 cells whereas full-length E2F-3 is not, independent of apparent changes in the E2F-3 message. (a) Protein extracts from proliferating (pro) and differentiated (dif) HL60 cells were analysed on separate Western blots (the relative mobilities of the three blots are not comparable since the E2F-3 gel ran longer). E2F-3 was detected with the C-terminal antibody. E2F-1 and 4 were detected using antibodies sc-193 and sc-866, respectively. One-hundred μg of protein was loaded in each lane. (b) Proliferating (pro) and differentiated (dif) HL60 cells were analysed by Northern blot. Five μg of polyadenylated RNA per lane was resolved on an agarose-formaldehyde gel. DNA fragments corresponding to the nonconserved C-terminal regions of E2F-1, E2F-3 and E2F-4 were used to probe the membrane as indicated

Several possible mechanisms could explain the generation of E2F-3B. While, members of the DP family have been shown to have alternative transcripts (Rogers et al., 1996; Zhang and Chellappan, 1995), there have been no reports of alternative transcripts for any member of the E2F family. However, to test the simple hypothesis that E2F-3 and E2F-3B could arise from alternative transcripts we analysed the E2F-3 mRNA in growing and differentiated HL60 cells. Figure 4b reveals that there was a single E2F-3 message in growing cells and that this message did not change in size or intensity following differentiation. This result supports the hypothesis that E2F-3 and E2F-3B both arise from the same transcript. As controls, the blot was stripped and reprobed with E2F-4 and E2F-1 cDNA fragments. Figure 4b reveals that the E2F-4 message was constitutive, whereas the E2F-1 message disappeared in the differentiated cells, accounting for the loss of E2F-1 protein.

A protein indistinguishable from E2F-3B originates by alternative translational initiation

Transient transfections were used to test two additional hypotheses; first, that E2F-3B is the product of a specific proteolysis and second, that E2F-3B is the product of alternative translational initiation. C-33A cells were transfected with pcDNA3-E2F-3, a plasmid that expresses full-length E2F-3 under the direction of a CMV promoter (WT). Comparison of Figure 5a lanes 1 and 3 demonstrates that the cells transfected with this plasmid overexpress predominantly full-length E2F-3 and some E2F-3B relative to untransfected cells (untransfected C-33A cells express little E2F-3B and virtually no full-length E2F-3). This result is consistent with the conclusion that E2F-3 and E2F-3B both originate from the same mRNA, and are not the products of different genes or alternatively spliced mRNAs. To determine if E2F-3 and 3B are alternative translation products or if E2F-3B is produced by proteolysis of E2F-3, a second plasmid that deletes the first 94 codons of E2F-3 including the translation start codon (Δ1–94) was transfected. If E2F-3B is the product of post-translational processing, then this deletion must abolish the production of both full-length E2F-3 and E2F-3B. If, on the other hand, E2F-3B is the product of an alternative translational start site, then the Δ1–94 plasmid could express E2F-3B, but could not express full-length E2F-3. Transfection of the Δ1–94 plasmid clearly resulted in E2F-3B overexpression in the absence of overexpression of full-length E2F-3 (see lane 2 of Figure 5a).

Figure 5
figure5

E2F-3B is produced by alternative translational initiation. (a) C-33A cells were transfected and after 36 h extracts were prepared and 100 μgs Western blotted using E2F-3 C-terminal antibody. Cells were transfected with the following plasmids: Lane 1, 20 μg empty pcDNA-3 vector (Mock). Lane 2, 10 μg pcDNA-3 containing E2F-3 deleted of its usual ATG Δ1–94 and 10 μg empty vector. Lane 3, 10 μg pcDNA-3 expressing full-length E2F-3 cDNA (WT) and 10 μg empty vector. Lane 4, 10 μg CMV-neo vector expressing E2F-3NT cDNA (E2F-3 NT) and 10 μg empty pcDNA-3 vector. Lane 5, 10 μg CMV-neo vector expressing E2F-3NT and 10 μg pcDNA-3 expressing full-length E2F-3 cDNA (WT) Lane 6, 10 μg pcDNA-3 containing E2F-3 deleted of its usual ATG (Δ1–94) and 10 μg pcDNA-3 expressing full-length E2F-3 cDNA (WT). Lane 7 represents in vitro translated E2F-3 that serves as a molecular weight reference and an internal control for the Western blot technique. (b) Same as panel A except that transfections included 10 μg of an HA-DP1-expressing plasmid and extracts were subjected to EMSA. Lane 1, 10 μg empty pcDNA-3 vector (Mock) and 10 μg of an HA-DP1-expressing plasmid. Lane 2, 10 μg pcDNA-3 expressing full-length E2F-3 cDNA (WT) and 10 μg HA-DP-1 plasmid. Lane 3, 10 μg pcDNA-3 containing E2F-3 deleted of its usual ATG (Δ1–94) and 10 μg HA-DP-1. Lane 4, 10 μg CMV-neo vector expressing E2F-3NT cDNA (E2F-3 NT) and 10 μg HA-DP-1. (c) BALB c-3T3 cells were transfected with the E2F-responsive reporter E2F1-LUC (Johnson et al., 1994) an internal control plasmid and either pcDNA3 empty vector, pcDNA3-E2F-3 or pcDNA3-E2F-3 Δ1–94, as indicated. Transfected cells were serum starved (0.5% calf serum) and lysed 48 h following transfection (Cress and Nevins, 1996) and luciferase levels (both reporter and internal control) determined. Relative activities are normalized using the renilla luciferase as transfection controls and setting pcDNA3 empty vector equal to onefold activation

To confirm that the polypeptide produced by the Δ1–94 plasmid was functional, transfections were repeated including an HA-DP-1 expression vector (Helin and Harlow, 1994) and whole cell extracts were evaluated by EMSA. Figure 5b reveals that the Δ1–94 plasmid produced a functional DNA-binding complex and that this complex had an electrophoretic mobility indistinguishable from E2F-3NT (and thus E2F-3B). As a second test of functionality, we cotransfected an E2F-responsive reporter plasmid E2F-1 LUC (Johnson et al., 1994) together with empty pcDNA3 vector, pcDNA3-E2F-3 or pcDNA3-E2F-3 Δ1–94 (which produces E2F-3B in the absence of full-length E2F-3 expression). The results presented in Figure 5c indicate that E2F-3B (produced by the pcDNA3-E2F-3 Δ1–94 construct) was only modestly diminished in its ability to activate transcription relative to full-length E2F-3.

The second ATG of the E2F-3 cDNA occurs at codon 232 and translational initiation at this codon would encode a protein lacking the DNA binding domain. Thus, the E2F-3 sequence does not contain an ATG codon that could possibly serve as the initiating codon for E2F-3B. Taken together, these facts suggest that E2F-3B must initiate translation at a non-ATG codon. To test this hypothesis the E2F-3 cDNA sequence was scanned for non-ATG codons that have been demonstrated to serve as initiation sites for other proteins and converted these codons to authentic ATG codons (Peabody, 1989). Figure 3b revealed that E2F-3B was approximately the size of E2F-3NT, and thus, mutagenesis focused on codons in the general vicinity of codon 132. Specifically, constructs were created that replaced putative initiation codons ACG 181, AAG 154, CTG 119, CTG 112 and ACG 102 (and all upstream sequences) with a consensus ATG initiation codon (see schematic in Figure 6a). These proteins were then translated in vitro and compared by SDS–PAGE with E2F-3B produced from the Δ1–94 construct. The data shown in Figure 6b,c reveal that initiation at codon 102 generated a protein with an electrophoretic mobility identical to E2F-3B, whereas all other constructs generated smaller proteins. To further test the hypothesis that ACG 102 serves as the initiation codon of E2F-3B, this codon was mutated to GGT (in the context of the Δ1–94 construct) to determine if its alteration would abolish E2F-3B expression. Figure 6d reveals that this mutation (GGT 102) abolished the expression of E2F-3B, and thus, ACG 102 serves at the E2F-3B initiation codon.

Figure 6
figure6

E2F-3B initiates at codon 102 relative to full-length E2F-3. (a) Schematic representing a series of E2F-3 constructs used to locate the E2F-3B non-ATG initiation codon. The schematic represents only the first 200 codons of each clone and is only approximately to scale. (b) The indicated proteins were translated in vitro in the presence of 35S-methionine and compared by SDS–PAGE. Lane 1, pcDNA-3 E2F-3 (full-length E2F-3). Lane 2, pcDNA-3 E2F-3 ATG 119. Lanes 3, 5 and 7, pcDNA-3 Δ1–94 (E2F-3B). Lane 4, pcDNA-3 ATG 132. Lane 6, pcDNA-3 ATG 154. Lane 8, pcDNA-3 ATG 181. Lane 9, un-programmed lysate (UPL), which received only empty pcDNA-3 vector. To partially compensate for the lower efficiency use of the E2F-3B non-ATG initiation codon, fivefold less reaction volumes were loaded. Specifically, lanes 3, 5, 7 and 9 received 10 μl, whereas lanes 1, 2, 4, 6 and 8 were loaded with only 2 μl. (c) Additional E2F-3 variants were assayed as described in panel B. Lanes 1, 3 and 5 were loaded with 5 μl of in vitro translated pcDNA-3 Δ1–94 (E2F-3B). Lane 2, 1 μl ATG 102. Lane 4, 1 μl ATG 112. Lane 6, 1 μl ATG 119. Lane 7, 5 μl unprogrammed lysate. (d) The putative E2F-3B initiation codon ACG 102 was changed to GGT and in vitro translation of this altered sequence was compared with the wt sequence and the ATG 102 mutant. Lane 1, 5 μl in vitro translated pcDNA-3 GGT 102. Lane 2, 5 μl in vitro translated pcDNA-3 E2F-3B. Lane 3, 5 μl ATG 102. Lane 4, 5 μl un-programmed lysate. Quantitation of the image (including correction for the amount of protein loaded) indicated that ATG 102 was translated 14-fold more efficiently than the normal WT sequence ACG 102. GGT 102 did not produce detectable protein

Initiation at a non-ATG codon is considered a rare event. To further verify that the E2F-3B-sized protein produced by initiation at codon 102 in transfections is the same as the endogenous E2F-3B the proteins were compared by denaturing iso-electric focusing. E2F-3 has a calculated pI of 5.29 and E2F-3B has a calculated pI of 5.41. In contrast, E2F-1, -2, -4 and -5 have much lower pIs of 4.79, 4.75, 4.66 and 4.94 respectively. Since E2F-3 and 3B are estimated to have considerably higher iso-electric points (pI) than other members of the E2F family do, they should be easily resolved from other family members (or related but non-identical proteins) by iso-electric focusing. To test this hypothesis, five members of the E2F family were translated in vitro in the presence of 35S-methionine and protein expression measured by SDS–PAGE (Figure 7a). These 35S-labeled proteins were then compared on an iso-electric-focusing gel (Figure 7b). The data in Figure 7b reveal the predicted result, E2F-3 and 3B migrate with similar mobility whereas E2F-1, 2 and 5 migrate to a much lower pI. E2F-4 apparently migrates off the gel under these conditions since it is highly phosphorylated (data not shown). Next we used iso-electric focusing to compare endogenous E2F-3B with E2F-3B produced by transfected Δ1–94 plasmid. For this experiment extracts of density-arrested fibroblasts, which contain only E2F-3B (and no full-length protein, Figure 3a), served as a source of endogenous E2F-3B. PDGF-stimulated fibroblasts that express primarily full-length E2F-3 served as a source of endogenous E2F-3. These extracts were then compared by iso-electric focusing with C-33A cells transfected with either empty vector (which express very little of either E2F-3 or 3B), with full-length E2F-3 and with Δ1–94. Following iso-electric focusing, proteins were transferred to membranes and detected by Western blot using the C-terminal antibody that recognizes both E2F-3 and 3B. The data in Figure 7c reveals that the endogenous E2F-3 and 3B proteins have pIs that are indistinguishable from the proteins produced by the plasmids. The multiple bands in the iso-electric focusing gels are likely to be the result of post-translational modifications that slightly change the pI of E2F-3 or E2F-3B (but apparently do not affect mobility in SDS–PAGE gels). Given the resolution of this system it is expected that each distinct band represents a difference in a single charge. To demonstrate this point the artificial E2F-3 protein E2F-3 NT, which initiates at codon 132 and has a calculated pI of 5.0, was compared to full-length E2F-3. With regard to charge E2F-3 NT differs from E2F-3 and 3B by one histidine and two arginine residues. Lanes 6 and 7 in Figure 7c clearly reveal that the difference of only three charged residues results in distinct mobility for E2F-3 NT. We conclude that endogenous E2F-3B and the protein produced by initiation at ACG 102 are identical.

Figure 7
figure7

Initiation at ACG codon 102 results in the expression of a protein indistinguishable from endogenous cellular E2F-3B. (a) E2F-3B and five full-length members of the E2F family were translated in vitro in the presence of 35S-methionine, separated by SDS–PAGE and then detected using a PhosphoImager. Each lane received 2 μl of the indicated in vitro transcription/translation reaction. Lane 1, UPL. Lane 2, represents in vitro translated E2F-3B (pcDNA-3 ATG 102). Lanes 2–4, represent in vitro translated full-length E2F-1 through E2F-5 (each translated from a pcDNA-3 vector). (b) In vitro-translated proteins (from a) were separated on a denaturing isoelectric focusing gel and detected by autoradiography. Lanes 1–6, each lane received 5 μl of the indicated in vitro transcription/translation reaction and the positions of bands are indicated. The positions of iso-electric point markers, carbonic anhydrase (pI=6.0), β-lactoglobin (pI=5.2) and glucose oxidase (pI=4.2) are indicated. (c) Endogenous E2F-3B from quiescent cells was compared with E2F-3B produced by the Δ1–94 plasmid. Protein extracts (100 μg) were resolved on a denaturing isoelectric focusing gel and E2F-3 proteins detected by Western blot. Lane 1, C33-A cells transfected with empty pcDNA-3 (mock). Lane 2, density-arrested BALB/c-3T3 cells (G0). Lane 3, density-arrested BALB/c-3T3 cells synchronized at G1/S by 15-h treatment with PDGF (G1/S). Lane 4, C33-A cells transfected with pcDNA-3 expressing full-length E2F-3 (E2F-3). Lane 5, C33-A cells transfected with pcDNA-3 Δ1–94 (E2F-3B). Lanes 6 and 7, 2 μl in vitro-translated full-length E2F-3 (IVT E2F-3) and E2F-3 NT (IVT ATG 132), respectively. The relative contributions of E2F-3 and E2F-3B in G0 and S fractions are the same as in Figure 3a (data not shown)

Discussion

We have identified an additional member of the E2F protein family, E2F-3B. It is intriguing that the switch from E2F-3B to E2F-3 full-length mirrors the general switch from expression of E2Fs lacking the NTR (E2F-4 and 3B) in G0 to expression of E2Fs containing an NTR (E2F-1, -2 and -3) during S phase. However, since the function of the E2F-3 NTR is unknown, we can only speculate at the functional relevance of the switch from E2F-3B to E2F-3 during the cell cycle. The data presented in Figure 5c make it clear that the NTR of E2F-3 is not required for transcriptional activation since E2F-3B efficiently activates the E2F-1 promoter. This result suggests that the NTR of E2F-3 is not essential for transcriptional activation (at least in the context of a luciferase reporter and overexpressed protein). These results, however, do not rule out a subtle role for the NTR in transcriptional activation at physiological levels of expression. Alternatively, the NTR of E2F-3 could serve an entirely novel function. Future experiments will address what function (if any) is provided by the NTR of E2F-3.

During the review of this manuscript a paper describing E2F-3 deficient mice was published (Humbert et al., 2000). The E2F-3 gene was found to be critical for normal development since E2F-3 deficient pups arose at one-quarter the expected frequency. Furthermore, it was found that the loss of E2F-3 dramatically impaired the mitogen-induced, transcriptional activation of numerous E2F-responsive genes in fibroblasts derived from E2F-3 deficient embryos. The results with E2F-3-deficient mice are in contrast to E2F-1 and E2F-5 deficient animals, which have no apparent defects in cell proliferation (Field et al., 1996; Lindeman et al., 1998; Yamasaki et al., 1996). The primary defect in E2F-1 deficient animals appears to be in apoptosis (especially in the thymus) and these animals develop a wide variety of tumors late in life (Field et al., 1996; Yamasaki et al., 1996). The primary defect in E2F-5 deficient mice appears to be in differentiation since the animals secrete excess cerebral spinal fluid and die from hydrocephalis (Lindeman et al., 1998). The observation that E2F-3 deficient mice have defects in both differentiation and proliferation suggests an important role of E2F-3 in growth control and in development. However, this data raises the question of the relative contributions of E2F-3 and E2F-3B to the phenotype of the E2F-3 deficient mice since both E2F-3 and 3B are included in the reported knockout. We speculate that E2F-3B may be primarily responsible for the defect in differentiation, whereas the full-length E2F-3 protein may be critical in the proliferative function of the E2F-3 gene.

We have presented several lines of evidence which are consistent with a model in which the E2F-3B polypeptide originates by alternative translational initiation at ACG codon 102 of the E2F-3 mRNA. Initiation at a non-ATG codon is not unprecedented among cell cycle-regulating transcription factors. There are three forms of c-Myc (c-myc1, c-myc2 and c-mycS) which appear to be produced by the use of alternative translational start sites, one of which is a non-ATG start codon (Hann, 1994, 1995; Hann et al., 1994, 1988, 1992; Spotts et al., 1997). C-myc1 arises from a non-ATG codon (CUG) and its levels, like E2F-3B, are highest in quiescent cells (Hann et al., 1988, 1992). C-myc2 arises from a downstream ATG codon. In general this is the most abundant form of c-Myc, yet in starved cells the c-myc1 levels are nearly equal to c-myc2 (Hann et al., 1992). C-mycS is produced by translational initiation at two closely spaced ATG codons far downstream of the c-myc2 ATG (Spotts et al., 1997). The c-myc1 and c-mycS isoforms appear to function primarily as dominant-negative molecules and, in the limited number of assays which have been examined, they block various functions of c-myc2 (Hann et al., 1994; Xiao et al., 1998). Likewise, different isoforms of the growth regulating human fibroblast growth factor 2 (FGF-2) are synthesized by the use of four alternative non-AUG codons and a single canonical AUG codon (Arnaud et al., 1999). The various modes of action of FGF-2 are the direct consequence of which initiation codon is used. Isoforms initiated at the upstream non-AUG codons localize to the nucleus and lead to the immortalization of bovine aortic endothelial cells, whereas the shorter isoform that initiates at the canonical AUG are cytoplasmic and leads to transformation of the same cells (Bugler et al., 1991; Couderc et al., 1991).

The E2F-3B non-ATG initiation codon is clearly different from the c-myc and FGF-2 non-ATG codons since it lies downstream of a consensus ATG. There are only a few precedents that we have found for a non-ATG codon downstream of an alternative ATG start codon to be a strong initiation codon (Boyd and Thummel, 1993; Madigan et al., 1996). Furthermore, we have found four examples of proteins that initiate exclusively at non-ATG initiation codons (Falvey et al., 1995; Imataka et al., 1997; Riechmann et al., 1999; Xiao et al., 1991). Presently, we can only speculate at the mechanism that controls the expression levels of full-length E2F-3 relative to E2F-3B. Figure 5a makes it apparent that both full-length E2F-3 and 3B can be produced from the full-length mRNA, thus (under these conditions) the translational machinery apparently leaks through the upstream E2F-3 ATG and then utilizes the downstream ACG 102 codon. Perhaps the relative leakiness of the upstream initiation codon is regulated in a cell cycle-dependent manner. Alternatively, the leakiness of the upstream codon could be constant and the relative accumulation of the two proteins could be determined by differential protein stability. Finally, although we observe only one band in our Northern analysis, we have not formally excluded alternative transcription units as means to regulate the relative levels of expression of E2F-3 versus 3B. Future work will be necessary to distinguish between these and other viable models.

Materials and methods

Cell culture

BALB/c-3T3 cells were maintained in Dulbecco’s Modified Eagle’s Media containing 5% calf serum, arrested in G0 by growing to confluence and stimulated to re-enter DNA synthesis by the addition of fresh media containing 5% calf serum and 10 ng/ml platelet-derived growth factor (PDGF). Eighty-five per cent of these cells reached the G1/S boundary 15 h following PDGF treatment (Flores et al., 1998) as determined by DNA content (propidium iodide staining). HL60 cells (ATCC CCL-240) were maintained in RPMI 1640 supplemented with 10% heat-inactivated fetal calf serum (FCS) and were induced to differentiate as previously described (Ikeda et al., 1996). C-33A cells were maintained in DMEM supplemented with 10% FCS.

Vectors, mutagenesis, transfections, and in vitro translations

E2F-3 and its derivatives were expressed using pcDNA-3 (Invitrogen). PcDNA3-E2F-3 was a gift from Alex Miron (Duke University). HA-DP-1 was expressed using a CMV-neo Bam vector which was a gift from Christian Helin (European Institute of Oncology). The calcium phosphate-based procedure used for transient transfections of E2F/DP heterodimeric pairs in C-33A cells has been previously described (Cress and Nevins, 1994, 1996; Flores et al., 1998). E2F-3 deletion construct Δ1–94 was prepared by the polymerase chain reaction (PCR) and the PCR fragment was cloned into the pcDNA3-E2F-3 plasmid using a vector HindIII site and an EcoDI restriction site at codon 221 of E2F-3. Other mutations were created using oligonucleotide-directed mutagenesis as previously described (Cress et al., 1993) and all mutants were sequenced by the Moffitt Molecular Core Facility. Oligonucleotides were obtained from Gibco/BRL (Bethesda, MD, USA) and specific sequences are available upon request. In vitro translations were done using a TNT-coupled transcription/translation system (Promega, Madison, WI, USA) as previously described (Cress et al., 1993). BALB/c-3T3 transfections were performed using Fugene 6 (Roche Molecular Biochemicals, Mannheim, Germany) in 60 mM dishes using 2 μg E2F1-LUC, 100 ng of pcDNA3 expression vector and 200 ng Renilla luciferase internal control (Promega, Madison, WI, USA).

Biochemical assays

Nuclear (BALB/c-3T3 cells) and whole cell extracts (transfected C-33A cells) were prepared as previously described (Ikeda et al., 1996). Western blots and electrophoretic mobility shift assays (EMSAs) were performed as previously described (Flores et al., 1998). Immunoprecipitation (IP) experiments used Dynal paramagnetic beads (Dynal, Success Lake, NY, USA). Luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). Northerns were done using GeneScreen (NEN Research Products, Boston, MA, USA) according to the manufacturer’s protocol. Polyadenylated RNA was purified using a Poly Atract System 1000 (Promega, Madison, WI, USA).

Denaturing iso-electric focusing gels were run as described in Current Protocols in Protein Science (Coligan et al., 1995) with the following modifications. All gels were 10 cm long and contained 8 M Urea, 2% Triton X-100, 4.5% Acrylamide (19 : 1), 2.6% ampholyte mixture pH 3 to 10, and 2.0% ampholyte mixture pH 4 to 6.5 (ampholytes were from Amersham/Pharmacia, Pharmalyte brand). Running buffers were 20 mM Phosphoric acid and 20 mM NaOH. They were prepared fresh for each gel and the NaOH solution was degassed before use. Low ionic strength samples were suspended in IEF load buffer (8 M urea, 2% Triton X-100, 2% ampholyte pH 3 to 10.5, 0.7 M β-mercaptoethanol) before loading. Gels were run for 3–4 h at a constant power of 5 W. Gels were analysed either by autoradiography of radiolabeled samples or by Western blot. If analysed by Western blot, gels were first washed five times for 10 minutes with 100 ml of 50% Methanol, 1% SDS, and 5 mM Tris pH 7.9. After washing, the proteins were transferred overnight to PVDF membrane (Millipore) and developed using standard procedures.

For DNA pull-down purification of E2F, 200 μgs of nuclear extract were treated with 0.4% deoxycholate (DOC) to disrupt E2F/RB complexes (apparently E2F-3B/RB complexes do not bind DNA well in this assay or in the EMSA). The DOC was neutralized with Nonidet P-40, and the extract incubated with a biotinylated double-stranded oligonucleotide (the 5′-biotinylated oligonucleotide was 5′-TCG ACC CCT TTT CCC GCC AAA AGG GG-3′ and complementary strand was 5′-TCG ACC CCT TTT GGC GGG AAA AGG GG-3′). The biotinylated DNA and bound E2F were ‘pulled down’ using streptavidin-agarose beads.

Production of antibodies against an E2F-3 internal region

The following oligonucleotide primers were used in the PCR to generate a fragment which was cleaved with BamHI and EcoRI and cloned into pGEX2T: forward primer; 5′-CCG CCG GGA TCC ATG AAA ACA AAC AAC CAA-3′ and reverse primer; 5′-CCG CCG GAA TTC TCA GTT CAC AAA CGG TCC-3′). GST-E2F- 3351–414 fusion protein was expressed in bacteria and purified by affinity chromatography. Aves Labs (Tigard, OR, USA) used the purified fusion protein to raise antibodies in two laying hens and provided purified IgY from preimmune and post-immune eggs.

References

  1. Allen KE, de la Luna S, Kerkhoven RM, Bernards R and La Thangue NB. . 1997 J. Cell. Sci. 110: 2819–2831.

  2. Arnaud E, Touriol C, Boutonnet C, Gensac MC, Vagner S, Prats H and Prats AC. . 1999 Mol. Cell. Biol. 19: 505–514.

  3. Bandara LR, Buck VM, Zamanian M, Johnston LH and La Thangue NB. . 1993 EMBO J. 12: 4317–4324.

  4. Beijersbergen RL, Kerkhoven RM, Zhu L, Carlee L, Voorhoeve PM and Bernards R. . 1994 Genes Dev. 8: 2680–2690.

  5. Boyd L and Thummel CS. . 1993 Proc. Natl. Acad. Sci. U S A. 90: 9164–9167.

  6. Buck V, Allen KE, Sorensen T, Bybee A, Hijmans EM, Voorhoeve PM, Bernards R and La Thangue NB. . 1995 Oncogene 11: 31–38.

  7. Bugler B, Amalric F and Prats H. . 1991 Mol. Cell. Biol. 11: 573–577.

  8. Cartwright P, Muller H, Wagener C, Holm K and Helin K. . 1998 Oncogene 17: 611–623.

  9. Chellappan SP, Hiebert S, Mudryj M, Horowitz JM and Nevins JR. . 1991 Cell 65: 1053–1061.

  10. Coligan J, Dunn B, Ploegh H, Speicher D, Wingfield P and Chanda V. . Current protocols in protein science. John Wiley and Sons. Inc.

  11. Couderc B, Prats H, Bayard F and Amalric F. . 1991 Cell. Regul. 2: 709–718.

  12. Cress WD, Johnson DG and Nevins JR. . 1993 Mol. Cell. Biol. 13: 6314–6325.

  13. Cress WD and Nevins JR. . 1994 J. Virol. 68: 4213–4219.

  14. Cress WD and Nevins JR. . 1996 Mol. Cell. Biol. 16: 2119–2127.

  15. DeGregori J, Leone G, Miron A, Jakoi L and Nevins JR. . 1997 Proc. Natl. Acad. Sci. USA 94: 7245–7250.

  16. Dong F, Cress Jr WD, Agrawal D and Pledger WJ. . 1998 J. Biol. Chem. 273: 6190–6195.

  17. Dyson N. . 1998 Genes Dev. 12: 2245–2262.

  18. Falvey E, Fleury-Olela F and Schibler U. . 1995 EMBO J. 14: 4307–4317.

  19. Field SJ, Tsai FY, Kuo F, Zubiaga AM, Kaelin Jr WG, Livingston DM, Orkin SH and Greenberg ME. . 1996 Cell 85: 549–561.

  20. Flores AM, Kassatly RF and Cress WD. . 1998 Oncogene 16: 1289–1298.

  21. Gaubatz S, Wood JG and Livingston DM. . 1998 Proc. Natl. Acad. Sci. USA 95: 9190–9195.

  22. Ginsberg D, Vairo G, Chittenden T, Xiao ZX, Xu G, Wydner KL, DeCaprio JA, Lawrence JB and Livingston DM. . 1994 Genes Dev 8: 2665–2679.

  23. Hann SR. . 1994 Biochimie 76: 880–886.

  24. Hann SR. . 1995 Adv. Exp. Med. Biol. 375: 107–116.

  25. Hann SR, Dixit M, Sears RC and Sealy L. . 1994 Genes. Dev. 8: 2441–2452.

  26. Hann SR, King MW, Bentley DL, Anderson CW and Eisenman RN. . 1988 Cell 52: 185–195.

  27. Hann SR, Sloan-Brown K and Spotts GD. . 1992 Genes Dev. 6: 1229–1240.

  28. Helin K and Harlow E. . 1994 J. Virol. 68: 5027–5035.

  29. Helin K, Lees JA, Vidal M, Dyson N, Harlow E and Fattaey A. . 1992 Cell 70: 337–350.

  30. Helin K, Wu C-L, Lees JA, Dynlacht BD, Ngwu C and Harlow E. . 1993 Genes. Dev. 7: 1850–1861.

  31. Hijmans EM, Voorhoeve PM, Beijersbergen RL, van’t Veer LJ and Bernards R. . 1995 Mol. Cell. Biol. 15: 3082–3089.

  32. Hsiao KM, McMahon SL and Farnham PJ. . 1994 Genes Dev 8: 1526–1537.

  33. Humbert PO, Verona R, Trimarchi JM, Rogers C, Dandapani S and Lees JA. . 2000 Genes Dev. 14: 690–703.

  34. Hurford Jr RK, Cobrinik D, Lee MH and Dyson N. . 1997 Genes Dev. 11: 1447–1463.

  35. Ikeda MA, Jakoi L and Nevins JR. . 1996 Proc. Natl. Acad. Sci. USA 93: 3215–3220.

  36. Imataka H, Olsen HS and Sonenberg N. . 1997 EMBO J. 16: 817–825.

  37. Itoh A, Levinson SF, Morita T, Kourembanas S, Brody JS and Mitsialis SA. . 1995 Cell. Mol. Biol. Res. 41: 147–154.

  38. Ivey-Hoyle M, Conroy R, Huber HE, Goodhart PJ, Oliff A and Heimbrook DC. . 1993 Mol. Cell. Biol. 13: 7802–7812.

  39. Johnson DG, Ohtani K and Nevins JR. . 1994 Genes Dev. 8: 1514–1525.

  40. Johnson DG and Schneider-Broussard R. . 1998 Front. Biosci. 3: 447–448.

  41. Johnson DG, Schwarz JK, Cress WD and Nevins JR. . 1993 Nature 365: 349–352.

  42. Kaelin Jr WG, Krek W, Sellers WR, DeCaprio JA, Ajchenbaum F, Fuchs CS, Chittenden T, Li Y, Farnham PJ, Blanar MA, Livingston DM and Flemington EK. . 1992 Cell 70: 351–364.

  43. Karlseder J, Rotheneder H and Wintersberger E. . 1996 Mol. Cell. Biol. 16: 1659–1667.

  44. Kitagawa M, Higashi H, Suzuki-Takahashi I, Segawa K, Hanks SK, Taya Y, Nishimura S and Okuyama A. . 1995 Oncogene 10: 229–236.

  45. Krek W, Ewen ME, Shirodkar S, Arany Z, Kaelin Jr WG and Livingston DM. . 1994 Cell 78: 161–172.

  46. Krek W, Xu G and Livingston DM. . 1995 Cell 83: 1149–1158.

  47. Lees JA, Saito M, Vidal M, Valentine M, Look T, Harlow E, Dyson N and Helin K. . 1993 Mol. Cell. Biol. 13: 7813–7825.

  48. Leone G, DeGregori J, Yan Z, Jakoi L, Ishida S, Williams RS and Nevins JR. . 1998 Genes. Dev. 12: 2120–2130.

  49. Lin SY, Black AR, Kostic D, Pajovic S, Hoover CN and Azizkhan JC. . 1996 Mol. Cell. Biol. 16: 1668–1675.

  50. Lindeman GJ, Dagnino L, Gaubatz S, Xu Y, Bronson RT, Warren HB and Livingston DM. . 1998 Genes Dev. 12: 1092–1098.

  51. Lindeman GJ, Gaubatz S, Livingston DM and Ginsberg D. . 1997 Proc. Natl. Acad. Sci. USA 94: 5095–5100.

  52. Lukas J, Petersen BO, Holm K, Bartek J and Helin K. . 1996 Mol. Cell. Biol. 16: 1047–1057.

  53. Madigan SJ, Edeen P, Esnayra J and McKeown M. . 1996 Mol. Cell. Biol. 16: 4222–4230.

  54. Magae J, Wu CL, Illenye S, Harlow E and Heintz NH. . 1996 J. Cell. Sci. 109: 1717–1726.

  55. Mann DJ and Jones NC. . 1996 Curr Biol. 6: 474–483.

  56. Moberg K, Starz MA and Lees JA. . 1996 Mol. Cell. Biol. 16: 1436–1449.

  57. Morkel M, Wenkel J, Bannister AJ, Kouzarides T and Hagemeier C. . 1997 Nature 390: 567–568.

  58. Muller H, Moroni MC, Vigo E, Petersen BO, Bartek J and Helin K. . 1997 Mol. Cell. Biol. 17: 5508–5520.

  59. Neuman E, Flemington EK, Sellers WR and Kaelin Jr WG. . 1994 Mol. Cell. Biol. 14: 6607–6615.

  60. Nevins JR. . 1998 Cell. Growth Differ 9: 585–593.

  61. Peabody DS. . 1989 J. Biol. Chem. 264: 5031–5035.

  62. Persengiev SP, Kondova II and Kilpatrick DL. . 1999 Mol. Cell. Biol. 19: 6048–6056.

  63. Pierce AM, Schneider-Broussard R, Philhower JL and Johnson DG. . 1998 Mol. Carcinog 22: 190–198.

  64. Riechmann JL, Ito T and Meyerowitz EM. . 1999 Mol. Cell. Biol. 19: 8505–8512.

  65. Rogers KT, Higgins PD, Milla MM, Phillips RS and Horowitz JM. . 1996 Proc. Natl. Acad. Sci. USA 93: 7594–7599.

  66. Sardet C, Vidal M, Cobrinik D, Geng Y, Onufryk C, Chen A and Weinberg RA. . 1995 Proc. Natl. Acad. Sci. USA 92: 2403–2407.

  67. Sears R, Ohtani K and Nevins JR. . 1997 Mol. Cell. Biol. 17: 5227–5235.

  68. Shan B, Zhu X, Chen PL, Durfee T, Yang Y, Sharp D and Lee WH. . 1992 Mol. Cell. Biol. 12: 5620–5631.

  69. Shin EK, Tevosian SG and Yee AS. . 1996 J. Biol. Chem. 271: 12261–12268.

  70. Spotts GD, Patel SV, Xiao Q and Hann SR. . 1997 Mol. Cell. Biol 17: 1459–1468.

  71. Tao Y, Kassatly RF, Cress WD and Horowitz JM. . 1997 Mol. Cell. Biol 17: 6994–7007.

  72. Trimarchi JM, Fairchild B, Verona R, Moberg K, Andon N and Lees JA. . 1998 Proc. Natl. Acad. Sci. USA 95: 2850–2855.

  73. Verona R, Moberg K, Estes S, Starz M, Vernon JP and Lees JA. . 1997 Mol. Cell. Biol. 17: 7268–7282.

  74. Wu CL, Zukerberg LR, Ngwu C, Harlow E and Lees JA. . 1995 Mol. Cell. Biol. 15: 2536–2546.

  75. Xiao JH, Davidson I, Matthes H, Garnier JM and Chambon P. . 1991 Cell 65: 551–568.

  76. Xiao Q, Claassen G, Shi J, Adachi S, Sedivy J and Hann SR. . 1998 Genes Dev 12: 3803–3808.

  77. Xu M, Sheppard KA, Peng CY, Yee AS and Piwnica-Worms H. . 1994 Mol. Cell. Biol. 14: 8420–8431.

  78. Yamasaki L, Jacks T, Bronson R, Goilot E, Harlow E and Dyson NJ. . 1996 Cell 85: 537–548.

  79. Zhang Y and Chellappan SP. . 1995 Oncogene 10: 2085–2093.

Download references

Acknowledgements

We thank Kip Wharton, Gene Ness, and Rhonda Croxton for critical comments on the manuscript. We thank Nancy Olashaw, Tere Muñoz-Antonia, Ed Seto and Ken Wright for helpful discussions. The Moffitt Flow Cytometry and Molecular Biology Core Facilities performed flow cytometry and DNA sequencing, respectively. This work was supported by NCI Grant #CA78214 to WDC, by a graduate fellowship to Y He from the American Heart Association, Florida/Puerto Rico Affiliate and by the H. Lee Moffitt Cancer Center and Research Institute. MJ Thomas is a Fellow of the Leukemia and Lymphoma Society.

Author information

Correspondence to W Douglas Cress.

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • E2F
  • E2F-3
  • retinoblastoma
  • cell cycle

Further reading