E2F family members are differentially regulated by reversible acetylation.

The six members of the E2F family of transcription factors play a key role in the control of cell cycle progression by regulating the expression of genes involved in DNA replication and cell proliferation. E2F-1, -2, and -3 belong to a structural and functional subfamily distinct from those of the other E2F family members. Here we report that E2F-1, -2, and -3, but not E2F-4, -5, and -6, associate with and are acetylated by p300 and cAMP-response element-binding protein acetyltransferases. Acetylation occurs at three conserved lysine residues located at the N-terminal boundary of their DNA binding domains. Acetylation of E2F-1 in vitro and in vivo markedly increases its binding affinity for a consensus E2F DNA-binding site, which is paralleled by enhanced transactivation of an E2F-responsive promoter. Acetylation of E2F-1 can be reversed by histone deacetylase-1, indicating that reversible acetylation is a mechanism for regulation also of non-histone proteins.

The six members of the E2F family of transcription factors play a key role in the control of cell cycle progression by regulating the expression of genes involved in DNA replication and cell proliferation. E2F-1, -2, and -3 belong to a structural and functional subfamily distinct from those of the other E2F family members. Here we report that E2F-1, -2, and -3, but not E2F-4, -5, and -6, associate with and are acetylated by p300 and cAMPresponse element-binding protein acetyltransferases. Acetylation occurs at three conserved lysine residues located at the N-terminal boundary of their DNA binding domains. Acetylation of E2F-1 in vitro and in vivo markedly increases its binding affinity for a consensus E2F DNA-binding site, which is paralleled by enhanced transactivation of an E2F-responsive promoter. Acetylation of E2F-1 can be reversed by histone deacetylase-1, indicating that reversible acetylation is a mechanism for regulation also of non-histone proteins.
Ordered progression through the G 1 and S phases of the eukaryotic cell cycle is tightly coupled to the expression of genes whose products are involved in cell growth and DNA replication. In mammalian cells, in Drosophila, and presumably in other multicellular eukaryotes, this temporal control of transcription is primarily regulated by the members of the E2F family of transcription factors (reviewed in Refs. [1][2][3][4]. Several E2F-responsive genes are activated in mid to late G 1 and play a crucial role in cell proliferation. Among these genes, those coding for Cdc6, Mcm proteins, thymidine kinase, dehydrofolate reductase, Orc1, DNA polymerase ␣, and proliferating cell nuclear antigen were demonstrated to be induced by exogenous E2F (5)(6)(7). Furthermore, binding sites for E2F are present also in the promoters of several other important cell cycle regulators such as cdc-2 (8), cyclin A (9), cyclin E (10), and B-myb (11).
E2F was originally identified as a cellular transcription factor required for the transactivation of the adenovirus E2 promoter by the E1A oncoproteins (12). E2F consists of a family of related proteins of which six distinct members have been found to date in mammalian cells (reviewed in Refs. 1 and 3). Sequence-specific DNA recognition and interaction is mediated by a DNA binding domain conserved in E2F-1 to -6. All six E2Fs contain also a leucine zipper dimerization domain and bind DNA preferentially as heterodimers with a subunit encoded by the DP family of genes, which are distantly related to the E2F family; heterodimerization with DP-1 and DP-2 markedly enhances E2F affinity for DNA. In vivo studies have confirmed that the endogenous E2F activity is generated from the combined properties of multiple E2F-DP proteins (13,14). The E2F transactivation domain consists of about 60 amino acids located at the C terminus of E2F-1 to -5 but lacking in E2F-6, the most recently discovered E2F family member. Accordingly, E2F-6 appears to behave as an inhibitor of E2F transcription, apparently by competing for promoter binding with the other E2Fs (15)(16)(17).
The correct timing of expression of E2F-dependent genes requires a precise regulation of E2F transactivation potential. E2F activity is controlled by members of the retinoblastoma protein (RB) 1 family, namely the tumor suppressor pRB (18 -20), p107 (21,22), and p130 (23). A short, highly conserved domain embedded in the E2F transactivation domain mediates binding to the pRB family proteins, resulting in inhibition of E2F transactivation. This inhibitory effect is relieved in mid to late G 1 , when pRB family members become phosphorylated by cyclin-dependent kinases. Phosphorylation of pRB determines dissociation from E2F, transcriptional activation of E2F-regulated genes, and consequent progression through the cell cycle (reviewed in Refs. 1-3).
E2F-1, E2F-2, and E2F-3 appear to belong to a structurally and functionally distinct E2F subfamily. This partition is also corroborated by their pattern of expression during the cell cycle and their nuclear localization. Thus, E2F-1, E2F-2, and E2F-3 are induced in late G 1 and, due to a N-terminal nuclear localization signal, are found exclusively in the nucleus when overexpressed (3). In contrast, the levels of E2F-4 and E2F-5 do not undergo major fluctuations, and these factors do not have a nuclear localization signal. The behavior of E2F-6 is still not completely clear, but it does not appear to be regulated by pRB (15,17).
Although the precise mechanism remains to be elucidated, a converging body of evidence suggests that histone-modifying enzymes may play a pivotal role in mediating and modulating E2F transactivation and inhibition by pRB. E2F was found to interact in vitro and to be synergistically activated by the cAMP-response element-binding protein (CBP; Ref. 24), a transcriptional co-activator that, similar to the closely related p300 protein, has histone acetyltransferase (HAT) activity (25,26).
In contrast, inhibition of E2F-responsive promoters by pRB was recently shown to involve the recruitment of histone deacetylase-1 (HDAC-1; Refs. [27][28][29]. HATs and HDACs are thought to activate and repress transcription, respectively, by relaxing or compacting the chromatin scaffold at cellular promoters. Acetylated histones are a characteristic feature of transcriptionally active chromatin, whereas hypoacetylated histones accumulate within transcriptionally silenced domains (reviewed in Refs. 30 -33). In addition to histones, few non-histone proteins have recently been discovered to be targets for acetylation by HATs, including TFIIE␤ and TFIIF (34), p53 (35), GATA-1 (36), and the erythroid Krü ppel-like factor (37). Also for these proteins, acetylation is specific for lysine residues. The transcription factor acetyltransferase (35) activity of HAT enzymes has opened an important new perspective to our current understanding of transcriptional regulation of gene expression and have suggested novel mechanisms for the fine tuning of transcription factor activity.
By a combination of biochemical as well functional assays, here we show that E2F-1, E2F-2, and E2F-3 (but not E2F-4, E2F-5, and E2F-6) are acetylated by the p300/CBP acetyltransferases and that acetylation markedly increases the specific DNA binding activity of E2F-1 and stimulates its transcriptional activity. We also show that acetylation of E2F-1 can be reversed by HDAC-1, thus providing for the first time evidence that also non-histone proteins are substrates for histone deacetylase enzymes.
GST in Vitro Binding Assay for HAT Activity-GST pull-down and HAT assays were performed as already described (38).
Acetylation Assay-Glutathione agarose-immobilized proteins were analyzed for acetylation as described (48), using 200 g of Dignam nuclear extract, immunoprecipitated p300/CBP, or recombinant CBP (aa 1098 -1877). After acetylation, proteins were washed three times with RIPA buffer to remove nuclear proteins and unbound [ 14 C]acetyl-CoA. Acetylated proteins were detected by autoradiography after separation by SDS-PAGE.
In Vitro Binding Assay-Binding of GST-proteins to 35 S-labeled p300 was performed as already published (38).
Electrophoretic Mobility Shift Assay-DNA binding was detected as described previously (44).

A Nuclear Histone Acetyltransferase (HAT) Acetylates E2F
Family Members-A current model to explain the role of p300/ CBP HAT and pRB-associated HDAC-1 in the regulation of E2F transactivation predicts that recruitment of these proteins alters the nucleosomes assembled at the promoters of E2Fresponsive genes. To determine whether histone acetylation is indeed involved in the regulation of E2F functions, we investigated the ability of E2F-1 to associate with a nuclear HAT. Recombinant E2F-1 purified from bacteria as a GST fusion protein was used in in vitro binding experiments after incubation with Dignam nuclear extract. The E2F-1-bound proteins were subsequently incubated with purified histones in the presence of radioactive acetyl-CoA to determine their HAT activity. The reaction products were separated by SDS-PAGE and detected by autoradiography. As shown in Fig. 1A, an E2F-1-associated nuclear HAT activity acetylated all the four core histones (H3, H2b, H2a, and H4; Fig. 1A, lower panel). Interestingly, also the band corresponding to GST-E2F-1 was significantly labeled, suggesting that E2F-1 is itself a target for acetylation (Fig. 1A, upper panel).
The site of acetylation within the 437 amino acids of E2F-1 was mapped using a panel of E2F-1 GST fusion fragments spanning the full-length protein. The fragments were incubated with Dignam nuclear extract and radioactive acetyl-CoA and treated as above (Fig. 1B). Acetylation was detected in the N-terminal half of the protein and restricted to the region between amino acids 109 and 127 (Fig. 1C).
Protein acetylation by HATs is specific for lysine residues (50). Three lysines (Lys-117, Lys-120, and Lys-125) are present in the E2F-1 region encompassing amino acids 109 -127. As shown in Fig. 1D, arginine substitution at all three lysines abolished acetylation, whereas mutations of two out of the three residues still allowed acetylation, although to a lower extent. These results indicate that all three lysines can serve as substrate for acetylation and that these are the only acetylated residues in the protein. An alignment of the sequences of the other E2F proteins corresponding to aa 109 -127 of E2F-1 revealed that lysines 117, 120, and 125 of E2F-1 are conserved also in E2F-2 and E2F-3 but not in E2F-4, E2F-5, and E2F-6 ( Fig. 2A). This prompted us to investigate whether these other members of the E2F family were also acetylated. As shown in Fig. 2B, and in agreement with the conservation of the three lysines in E2F-2 and E2F-3, these proteins were also acetylated. In contrast, we were not able to detect any acetylation of E2F-4, E2F-5, and E2F-6, suggesting that acetylation is restricted to a subset of E2F family members. In addition, we also found that DP-1 was not acetylated in this assay (data not shown).
E2F-1 Is Acetylated by p300/CBP-As apparent in the lower panel of Fig. 1A, E2F-1 is acetylated by a HAT activity specific for all the four core histones. Among the nuclear HATs to date identified and characterized, p300 and the closely related CBP proteins are the only enzymes with this substrate specificity (reviewed in Ref. 51). In addition, earlier studies using truncated proteins have shown that E2F-1 is able to bind CBP in vitro (24). Therefore, we assessed the ability of all six fulllength E2F family members and lysine mutants to interact with p300 in vitro. Recombinant E2F proteins were incubated with 35 S-labeled p300, and the amount of bound p300 was detected by autoradiography after separation by electrophoresis (Fig. 3A). We discovered that only E2F-1, E2F-2, and E2F-3 were able to bind p300 efficiently, a finding in good agreement with the results of our acetylation studies. Interestingly, the triple lysine mutant that is not acetylated showed unaltered affinity for p300 as compared with wild type E2F-1. This demonstrates that the substitution of arginines for lysines has not destroyed the overall structure of the protein and that binding to p300 is not sufficient to induce modification of any other lysine residue present in the protein.
To ascertain that endogenous p300/CBP HATs could acetylate E2F, we immunoprecipitated p300 and CBP from nuclear extract and showed that both immunoprecipitates contained proteins capable of acetylating E2F-1 as well as the four core histones (Fig. 3B). Acetylation was substrate-specific, as unrelated proteins such as bovine serum albumin (BSA) remained unmodified. Acetylation of E2F-1 was also obtained with a recombinant fragment of CBP containing the HAT domain, thus confirming the immunoprecipitation results (Fig. 3C). These data suggest that E2F-1 is acetylated by p300/CBP. The p300 binding and acetylation results obtained for the different E2F members and for the E2F-1 mutants are summarized in Fig. 4.
E2F Acetylation Can Be Reversed by HDAC-1-The action of HATs on histones can be reversed by HDACs, the prototype of which is HDAC-1 (52). So far, no evidence is available that HDACs can also act on non-histone substrates. Therefore, we set out to investigate whether E2F could be a substrate for this class of enzymes. A fragment of E2F-1 containing lysines 117, 120, and 125 (aa 41-127) was acetylated in vitro by immunoprecipitated p300 and CBP in the presence of [ 14 C]acetyl-CoA. Subsequently, the radioactive acetyl-CoA was removed, and the modified protein was incubated with HDAC-1 immunoprecipitated from Dignam nuclear extract. After separation by SDS-PAGE, acetylation was detected by autoradiography. Similar to histone acetylation, acetylation of E2F-1 by p300 and CBP could be reversed by a subsequent incubation with immunoprecipitated HDAC-1 (Fig. 5). The enzymatic activity of the latter enzyme was confirmed by its ability to deacetylate p300and CBP-acetylated histones (data not shown). The fact that immunoprecipitated CBP appears to acetylate E2F-1 more strongly than immunoprecipitated p300 depends on the properties of the two antibodies used in the assay and does not reflect a better efficiency of acetylation.
Acetylation of E2F Increases DNA Binding and Transcriptional Activation-What is the functional relevance of E2F acetylation? To date, only few non-histone proteins have been demonstrated to be acetylated (34). For some of these proteins, including p53 (35) and GATA-1 (36), acetylation was shown to lead to increased DNA binding. Since the three acetylated lysines in E2F-1, E2F-2, and E2F-3 are located at the Nterminal boundary of the DNA-binding domain of these proteins (53), we explored the possibility that acetylation altered the affinity of E2F for its binding site. E2F-1 was incubated with recombinant CBP in the absence or presence of acetyl-CoA. The latter incubation was also carried out separately using [ 14 C]acetyl-CoA to monitor acetylation efficiency (data not shown). After incubation, increasing amounts of the reactions were assayed for binding to a consensus E2F DNA-binding site. Fig. 6A shows that acetylation of E2F-1 led to a marked increase in the affinity for its binding site as compared with the mock-acetylated protein.
To determine whether these in vitro observations are also relevant in vivo, we investigated the E2F DNA binding activity in nuclear extracts of cells overexpressing CBP. Both the binding activity of endogenous E2F and that of transfected E2F-1 were clearly augmented by transfection of an expression plasmid for CBP (Fig. 6B).
The functional consequences of the increased DNA binding were also tested on the transcriptional activation of an E2Fresponsive promoter (Fig. 7A). To prevent interference due to the recruitment of histone deacetylases to the promoter through pRB, these experiments were performed in the pRBnegative Saos-2 cell line. Co-transfection of a CBP expression plasmid in these cells resulted in a clear increase in E2F-1 transactivation (Fig. 7B). In contrast, transactivation by the E2F-1 mutant with substitutions of lysines 117, 120, and 125 to arginines (which binds to but is not acetylated by p300/CBP) and of E2F-4 (which neither binds to nor is acetylated by p300/CBP) were both only minimally affected by CBP co-transfection (Fig. 7B). Altogether, these data highlight the importance of E2F-1 acetylation in response to CBP co-activation. Interestingly, the triple mutant showed increased basal transactivation in the absence of CBP as compared with wild type E2F-1. This result is possibly related to the different stability of this mutant protein (54). DISCUSSION The E2F family of transcription factors plays a critical role in the regulation of the cell proliferation by controlling transcription of several genes, whose timely expression is necessary for the ordered progression through the cell cycle. The responsiveness of their promoters to E2F ensures that this whole set of genes is simultaneously activated at the exact time their products are required in the cell cycle (reviewed in Refs. 1 and 3). Given this pivotal role, it is not surprising that E2F activity is strictly kept in check by a complex inhibitory mechanism that starts to be elucidated. Silencing of E2F transactivating potential is in part mediated by pRB proteins and is abrogated in mid to late G 1 to ensure that the growth-suppressive properties of pRB become inactivated (reviewed in Ref. 55). Since the pRBbinding domain of E2F is embedded in its transactivating domain, it has been proposed that inhibition by pRB is the consequence of a masking effect that negatively affects the physical interaction of E2F with the basal transcriptional machinery, in particular with TFIID and TFIIA (56 -58).
Several data have suggested that pBR represses E2F activity also through an active mechanism (reviewed in Ref. 1), which has been recently demonstrated to involve HDAC-1. Three independent studies have shown that HDAC-1 is associated with E2F-1 via pRB and that its deacetylase activity is required for inhibition of E2F activity (27)(28)(29). The evidence that chromatin-modifying enzymes are involved in the regulation of E2F has led to a model according to which repression or activation of E2F-responsive genes is also dependent on the acetylation state of the nucleosomes at their promoters. Consistently, it has been earlier demonstrated that E2F-1 can interact in vitro with a fragment of CBP (24), a general transcriptional coactivator having histone acetyltransferase activity (25,26).
The findings presented here add more complexity to the notion that acetylases and deacetylases are involved in the regulation of E2F function. In fact, they indicate that not only modification of chromatin by these enzymes but also modification of E2F itself is likely to play a role in the transcriptional regulation of E2F-controlled genes.
E2F-1 is specifically acetylated at lysines at positions 117, 120, and 125 by p300 and CBP. The acetylation of E2F-1 by these enzymes is direct, since acetylation of a recombinant E2F-1 substrate containing the three lysines was detected both by using p300 and CBP immunoprecipitates and a recombinant fragment of CBP. Interestingly, substitution of lysines 117, 120, and 125 of E2F-1 with arginines abolishes its acetylation but not its binding to p300/CBP. Consistent with the binding and acetylation data, transfection of an expression vector for CBP in pRB-negative Saos-2 cells markedly increased transcriptional activation of an E2F-responsible promoter by E2F-1 but not by E2F-4. In the same assay, the construct in which the three lysines being the targets for acetylation were substituted with arginines did not respond to CBP co-transfection. The same mutant, however, was still proficient in the transactivation of the E2F-responsive promoter. This evidence suggests that E2F-1 acetylation is probably not required per se for transcriptional activation, whereas it is important to mediate p300/ CBP co-activation.
The three lysines acetylated by p300/CBP are located at the N-terminal boundary of the DNA-binding domain of E2F-1, and acetylation leads to a marked increase in affinity of E2F-1 for its DNA consensus binding site. This result is functionally FIG. 6. Acetylation of E2F-1 increases DNA binding. A, acetylation increases DNA binding in vitro. GST-E2F-1 (aa 1-284) was incubated with the enzymatically active GST-CBP (aa 1098 -1877) protein, without or with acetyl-CoA, as indicated. After incubation, increasing amounts of the reactions, containing the indicated nanograms of GST-E2F-1 protein, were used for electrophoretic mobility shift assays using an E2F probe. The numbers on top of each bar on the right side of the histogram indicate fold binding of the samples incubated with acetyl-CoA over the respective samples incubated in the absence of acetyl-CoA. B, acetylation increases DNA binding in vivo. Overexpression of CBP increases DNA affinity of endogenous and transfected E2F. U2OS cells were transfected with expression vectors for E2F-1 and/or CBP, as indicated. Transfections of E2F-1 were always carried out in the presence of a DP1-expression plasmid. Nuclear extracts were analyzed by electrophoretic mobility shift assays using an E2F probe.

FIG. 7. Acetylation of E2F-1 increases transcriptional activation.
A, schematic representation of the reporter plasmid used in the transfection experiments. The plasmid contains six E2F-binding sites upstream of the luciferase reporter gene (not drawn to scale). B, overexpression of CPB HAT enhances wild type E2F-1 transactivation. Transcriptional activation of an E2F-responsive promoter was studied in Saos-2 cells after transfection of different amounts of expression vectors for wild type E2F-1, E2F-1 K117R/K120R/K125R and E2F-4, in the presence or absence of an expression plasmid for CBP, as indicated. All E2F transfections also contained 1 or 5 ng of an expression plasmid for DP1. similar to the findings reported for p53 and GATA-1. For p53, acetylation occurs in the C-terminal domain of the protein, a region that interacts with the DNA-binding domain and locks it in an active conformation (35). In GATA-1, the lysines acetylated by p300/CBP lie within the DNA-binding region of the factor (36). Similar to our results on E2F-1, the acetylation of p53 and GATA-1 dramatically increases their affinity to their respective consensus DNA-binding sites. In this respect, it is noteworthy that, although the net negative charge of these proteins is increased by acetylation, the overall functional consequence is a marked increase in DNA binding. It can be envisaged that the increase in DNA affinity results from a conformational change induced by the addition of acetyl groups, which in turn favors the formation of more stable protein-DNA interactions.
The finding that only E2F-1, E2F-2, and E2F-3 bind to p300/ CBP, contain the lysine residues at positions 117, 120, and 125, and are substrates for acetylation further highlights the notion that these three factors belong to a subgroup of the E2F family. Contrary to E2F-4 and E2F-5, these factors are not present in G 0 or early G 1 ; they bind directly to cyclin A; they only bind pRB and not p107/p130; and they display a different pattern of expression during development (reviewed in Refs. 1 and 3). To what extent these functional differences could be related to the differential acetylation of these factors will clearly need further consideration. In this respect, it is interesting to observe that not only E2F-1 is a substrate for acetylation by p300/CBP but that acetylated E2F-1 is also a substrate for deacetylation for HDAC-1. This observation again supports and reinforces the notion that activation and suppression of E2F transcription by acetylation is likely to be a tunable process for the regulation of this DNA binding factor. Although this is the first, and so far the only (59) demonstration that acetylation of non-histone proteins is in fact reversible and that HDAC enzymes can also act on a non-histone substrate, it might be envisaged that reversible acetylation will turn out to be a general mechanism for post-translational protein regulation.
In conclusion, we have provided evidence that the distinct structural organization of the E2F family members determines a different susceptibility to acetylation by the p300/CBP acetyltransferases and that acetylation of E2F-1 markedly activates its biochemical function. This is a novel mechanism of regulation of E2F function, distinct both from the modulation of chromatin conformation mediated by histone acetylation and deacetylation (24, 26 -29) and from the repression of the transactivation domain of the protein exerted by pRB and acting also in the absence of chromatin (58,60). Further studies will be required to assess the relative contribution of these three mechanisms in the regulation of cell cycle progression.
While performing the experiments described in this manuscript, we became aware that similar results were obtained by T. Kouzarides and co-workers while studying acetylation of E2F-1 by the P/CAF histone acetyltransferase (54).