Dynamic Recruitment of NF-Y and Histone Acetyltransferases on Cell-cycle Promoters*

Regulation of transcription during the cell-cycle is under the control of E2 factors (E2Fs), often in cooper-ation with nuclear factor Y (NF-Y), a histone-like CCAAT-binding trimer. NF-Y is paradigmatic of a constitutive, ubiquitous factor that pre-sets the promoter architecture for other regulatory proteins to access it. We analyzed the recruitment of NF-Y, E2F1/4/6, histone acetyltransferases, and histone deacetylase (HDAC) 1/3/4 to several cell-cycle promoters by chromatin immunoprecipitation assays in serum-starved and restimulated NIH3T3 cells. NF-Y binding is not constitutive but timely regulated in all promoters tested, being displaced when promoters are repressed. p300 association correlates with activation, and it is never found in the absence of NF-Y, whereas PCAF/hGCN5 is often found before NF-Y association. E2F4 and E2F6, together with HDACs, are bound to repressed promoters, including the G 2 /M Cyclin B2. As expected, an inverse relationship between HDACs association and histones H3/H4 acetylation is observed.

In addition to E2F sites, most cell-cycle regulated promoters, particularly those of regulators of the cycle, also contain CCAAT boxes. Genes activated in G 0 (PDGF␤-R), G 1 /S (E2F1, CDC25A, and PCNA), S (Cyclin A and CDC25B), S/G 2 (topoisomerase II␣ and PLK), and G 2 /M (CDC25C and Cyclin B1/B2) all contain CCAAT boxes, invariably shown to be crucial for the proper regulation of these genes (8 -17). In vivo footprinting experiments indicated that the CCAAT boxes are protected (10,18,19). A combination of electrophoretic mobility shift assays (EMSAs) and transfections with highly diagnostic dominant negative vectors implicated nuclear factor Y (NF-Y) as the common CCAAT activator (20). NF-Y is composed of three subunits, NF-YA, NF-YB, and NF-YC, all necessary for DNA binding. NF-YA and NF-YC possess large Q-rich activation domains. NF-YB and NF-YC dimerization is a prerequisite for NF-YA association and sequence-specific DNA binding. NF-YB and NF-YC contain histone fold motifs (HFM) found in all core histones, which mediate dimerization and DNA binding (Ref. 21 and references therein). In all systems of inducible transcription tested so far (heat shock, cholesterol biosynthesis, endoplasmic reticulum stress, and CYP genes), ChIP experiments determined that NF-Y is bound in vivo before gene activation (22,23) 2 ; indeed, NF-Y is bound to a transcribing Cyclin B1 promoter during mitosis in HeLa cells (24). In G 1 /S promoters, NF-Y cooperates with E2Fs (8). In G 2 /M promoters, NF-Y-binding CCAAT-boxes are usually found in multiple copies near the CDE-CHR elements (10 -12, 15-17). These latter conserved boxes have been identified by genetic experiments as the crucial determinants of proper regulation, that is repression in G 0 , G 1 , and early S (19). However, biochemical characterization of the CDE-CHR interacting protein(s) has been elusive. In the case of Cdc2, circumstantial evidence suggests that E2Fs might partake in promoter regulation through the CDE-CHR (25).
A further level of regulation is exerted by coactivators and corepressors containing histones modifying activities, histone acetyltransferases (HATs), and HDACs. The former serves as a bridge for transcription factors and holo-enzyme interactions, having been recruited to promoters through the activation domains of the DNA-binding activators. Not surprisingly, connections between NF-Y and HAT coactivators have emerged. Interactions of the HFM subunits with hGCN5 were reported (26). p300/CBP-associated factor (PCAF) mediates the trichostatin A induction of the MDR1 promoter through NF-Y (27). Treatment with HDAC inhibitors induces the TGF␤-RII promoter by increasing the affinity of PCAF for the NF-Y trimer (28). Binding to p300 activates the HSP70 promoter in Xenopus oocytes, in the absence of heat shock or addition of heat shock factor 1 (29). In this latter study, p300 was able to acetylate NF-YB in vitro, but the function of this modification is currently unknown. Negative regulation through NF-Y, HDAC activity, and histone deacetylation also surfaced (28,30,31). These latter results suggest that NF-Y might be instrumental in the establishment of repressive as well as activating complexes on the regulated promoters. We decided to use the powerful means of chromatin immunoprecipitation to gain a more complete view of the association of NF-Y, HATs, HDACs, and E2Fs to cell-cycle regulated CCAAT-containing promoters.

MATERIALS AND METHODS
Cell Line and Culture Conditions-NIH3T3 mouse fibroblasts were grown in RPMI under standard conditions with 10% fetal calf serum. Cells were arrested by incubation for 60/72 h in 0.5% fetal calf serum, and restimulation was obtained by addition of fresh medium with 10% fetal calf serum. Cell-cycle progression was monitored by DNA content through propidium iodide staining and FACScan analysis. For R-roscovitine treatment, the drug (Cyclacell, Dundee, UK) was used at a 20 M concentration for 16 h. Under these conditions all cells were blocked in G 1 , as reported in Ref. 32.
Chromatin Immunoprecipitations-Formaldehyde cross-linking and chromatin immunoprecipitation was performed as described in Ref. 3. NIH3T3 cells (0.5/1.10 8 ) were washed in phosphate-buffered saline and incubated for 10 min with 1% formaldehyde. After quenching the reaction with glycine 0.1 M, the cross-linked material was sonicated into chromatin fragments of an average length of 500/800 bp. Chromatin was kept at Ϫ80°C. Immunoprecipitations were performed with ProtG-Sepharose (KPL) and 3-5 g of the indicated antibodies: anti-YB purified rabbit polyclonal (23); anti-p300 (Santa Cruz, sc-585x); anti-PCAF and anti-GCN5 (a kind gift of Y. Nakatani, Harvard University, Boston, MA); anti-E2F4 (Santa Cruz, sc-1082x); anti-E2F1 and E2F6 (Active Motif, 39313 and 39509); anti-HDAC1 (Sigma, H3284); anti-HDAC 3 and HDAC4 (Active Motif, 40968 and 40969); and anti-acetylated H3 and H4 (Upstate, 06-599 and 06-866). The chromatin solution was precleared by adding ProtG-Sepharose for 2 h at 4°C and was aliquoted and incubated with the antibodies overnight at 4°C on a rotating wheel. ProtG-Sepharose was blocked with 1 g/l salmon sperm DNA (Sigma) and 1 g/l bovine serum albumin overnight at 4°C and then incubated with chromatin and antibody for 2 h. Immunoprecipitated material was washed 9 times with wash buffer as described in Ref

In Vivo Binding of NF-Y and HATs to CCAAT Cell-cycle
Promoters-To assess the in vivo recruitment of NF-Y and HATs on cell-cycle promoters, we used the ChIP assay with anti-YB and anti-HATs (p300, PCAF, and GCN5) antibodies in the NIH3T3 mouse fibroblasts system (3). Chromatin from cells starved by serum withdrawal (G 0 ) and after restimulation with 10% fetal calf serum for 6, 12, 18, and 24 h was prepared and immunoprecipitated with anti-NF-Y, p300, PCAF, GCN5, and control anti-leader-binding protein 1/CCAAT protein 2 antibodies. Fluorescence-activated cell sorter analysis of the cell populations at the different time points was run to verify cell-cycle progression, which was indeed as expected and shown in Fig. 1. Note that the 24 h time point was more heterogeneous with 20/30% of the cells that re-entered G 1 . We first amplified a control fragment in the nucleolin CCAAT-less intronic enhancer ( Fig. 1, upper panel, and Ref. 33). The anti-NF-Y, p300 antibodies were negative, whereas CCAAT protein 2 was apparently bound between 6 -18 h (33). This factor has been implicated in the activation of the CCAAT-less TS promoter in S phase (34). PCAF/GCN5 were weakly positive at 18 h, and this correlates with the binding of MYC and TRRAP, although with a slower kinetic (33).
The CCAAT promoters chosen for analysis are active in G 0 (PDGF␤-R) shortly after the G 0 -G 1 transition (c-Jun and JunB), late G 1 /early S (E2F1, RNR-R1, PCNA, and Cyclin A), late S/G 2 (TopoII␣, CDC25C, and PLK), and G 2 /M (Cdc2 and Cyclin B2). Fig. 1 shows a comprehensive analysis of the results obtained with the target promoters in the different time points.
It is clear that NF-Y binding to these promoters is not constitutive. In G 0 cells, it is present only on PDGF␤-R, c-Jun, and JunB. At 6 h, these promoters are negative, whereas association is weakly positive to E2F1. In Late G 1 , NF-Y binding is maximal on E2F1, PCNA, and RNR-R1 promoters, and detectable on TopoII␣, Cyclin A, PLK, and CDC25C. Finally, Cdc2 and Cyclin B2 become positive only at 18 h. It should also be noted that promoters that have been transcriptionally switched off, such as PDGF␤-R, cJun, and JunB at 6 h and PCNA at 24 h, are unloaded of NF-Y, whereas the occupancy of others (RNR-R1 and E2F1) is considerably decreased. We conclude that the binding of NF-Y to cell-cycle promoters is highly dynamic.
The pattern of p300 binding closely matches promoter activation. In G 0 , the only promoter that is clearly positive is the transcribing PDGF-␤R, but weak association is observed with c-Jun and JunB, whereas all other promoters are negative. These data correlate well with histone H4 acetylation on PDGF-␤R (Fig. 3), and the low but detectable levels observed in previous studies on JunB (35). These promoters are inactive at 6 h (14, 35). G 1 /S promoters are positive at 12 h, later becoming negative, and Cdc2 and Cyclin B2 are bound at 18 h. Binding to Cdc2, but not to Cyclin B2, is substantially reduced at 24 h. S/G 2 promoters are somewhat intermediate, being already associated at 12 h. It is important to note that p300 is never found on promoters in the absence of NF-Y.
The binding of PCAF to all promoters is, in general, more pronounced than that of hGCN5, with the exception of RNR-R1 at early G 1 time points. This most likely reflects the far higher (5/10-fold) amounts of PCAF in NIH3T3 cells. The patterns of binding of these HATs appear to precede promoter activation, often well before NF-Y and p300 binding. In G 0 cells, only the transcribing PDGF␤-R is bound, whereas at 6 h all three G 1 /S promoters, TopoII␣, PLK, and CDC25C, are significantly bound in conditions in which neither NF-Y nor p300 are associated. Cyclin B2 starts to be loaded at 12 h. Interestingly, PCNA, RNR-R1, E2F1, Cyclin A, and Cdc2 become rapidly unloaded of PCAF after the promoter has been switched off. In the case of PDGF␤-R, PCAF is on at 6 h before reloading of NF-Y and p300 at 12 h. As for c-Jun and JunB, which are rapidly (0.5 h) activated following serum addition, PCAF/GCN5 are not bound in G 0 nor after 6 h. Most likely, we have missed the early PCAF/hGCN5 loading because of the kinetic used in our assays; however, the two HATs are on at 12 h. Note that maximal NF-Y and p300 reloading on JunB occurs at 18 h.
In Vivo Binding of E2Fs and HDACs to CCAAT Cell-cycle Promoters-The experiments shown above prompted us to control the association of the E2F factors to CCAAT promoters. Essentially, the same protocol was used with anti-E2F1/4/6 antibodies on a representative set of promoters. Fig. 2 shows the results of such ChIP analysis. The PDGF␤-R promoter is apparently devoid of E2Fs, at least within the time points considered. The E2F1 promoter is bound by E2F4 and, to a lesser extent, by E2F1 in G 0 cells. At 12 h, when the gene is being transcribed, E2F1 clearly becomes prominent, then declines rapidly. Cdc2 and Cyclin A show a similar behavior, except that association of E2F1 in S is not observed and E2F4 is clearly bound at later time points (24 h) when a proportion of the cell population (20/30%) is re-entering G 1 . Furthermore, the repressive E2F6 is bound to Cyclin A (weakly) and to Cdc2 (more strongly) in G 0 cells, rapidly becoming unloaded at 6 h. Because E2F4/1 have been detected on several S/G 2 promoters in genome-wide experiments (6, 7), we also tested Cyclin B2 in these ChIPs. It is clear that E2F4 and, to a lesser degree, E2F1, are associated in G 0 , decreasing at 6 h and desappearing at 12 h. E2F6, which is not bound in G 0 , is the prominent E2F bound at 6 h. Overall, this analysis confirms and extends the data previously reported in the NIH3T3 system (3) and in human fibroblasts (2, 4, 5), with the exception of Cyclin B2, previously undetected as an E2F target.
HDACs are associated with repressive E2Fs in many of the

FIG. 1. In vivo binding of NF-Y and HATs to CCAAT cell-cycle promoters.
Chromatin was prepared from cells at G 0 after starvation or 6, 12, 18, and 24 h after serum addition, as indicated, and immunoprecipitated with anti-NF-YB, anti-p300, anti PCAF, anti-GCN5, and anti-CCAAT protein 2 control antibodies.
E2F-regulated promoters and that HDAC binding correlates with low levels of histones H3 and H4 acetylation (2, 3). Fig. 3 shows ChIP analysis with antibodies against HDAC1/3/4 and acetyl-K of histones H3 and H4. The PDGF␤-R promoter is apparently devoid of HDACs at G 0 , and indeed acetylation of H4 is observed, as expected for an actively transcribing promoter. At 6 h, histones are deacetylated and HDAC1/4 are bound. At later time points histones (expecially H3) are acetylated, with little binding of HDACs. The E2F1 promoter is bound by HDAC1/3/4 and the histones deacetylated in G 0 . HDACs association is no longer visible beyond 6 h, whereas H3 and, subsequently, H4 acetylation is observed during activation. Cdc2 and Cyclin A show binding of HDAC1 in G 0 , which is released during activation at 12/18 h. On Cyclin A, HDAC3/4 become reassociated at the late 24 h time point. The levels of histone acetylation correlate with promoter activation, indeed preceding it and being specular to that of HDAC binding. On Cyclin B2, HDAC1/4 are associated at 6 h, consistent with the binding of E2F6 at this time point, but not with the binding of E2F4 at G 0 . Acetylation of H3 is maximal at 12 h, following HDACs release and before transcriptional activation. The data presented lead us to conclude that the acetylation of histones H3-H4 tails faithfully reflects the HATs/HDACs dynamic behavior on these promoters, further validating the data on HATs and NF-Y binding.
NF-Y Is Differentially Bound in Cells Blocked at the G 1 /S Transition-The results obtained above with a protocol of serum starvation-restimulation might be peculiar to a situation of G 0 -G 1 re-entry and not reflect changes of promoter occupancy in normally growing cells. We sought to verify the behavior of NF-Y in a different experimental setting, that is in cells arrested in their normal cycles with the cyclin-dependent kinase 2 inhibitor R-roscovitine (32). Treatment of cycling cells with pharmacological doses of the drug indeed leads to a complete block in G 1 (not shown). Fig. 4, left panel, shows that the promoters of all genes are bound by NF-Y in cycling cells, a result well expected, given the heterogeneous nature of this cell population. After the R-roscovitine block, NF-Y is found only on G 1 /S promoters (Fig. 4, right panel). In full support of the data of Fig. 1, however, in none of the G 2 /M promoters NF-Y is loaded, a clear indication that the late G 1 block has prevented its association with promoters active in subsequent phases. Thus, we conclude that the cell-cycle-dependent association of NF-Y to its targets is not a peculiar phenomenon of serum restimulated fibroblasts, but rather a theme common to normally growing cells. DISCUSSION Through the use of the chromatin immunoprecipitation technique, we report three important results. (i) Contrary to predictions based upon in vitro binding activity in EMSAs and ChIP data on other promoters, the in vivo binding of NF-Y to cell-cycle genes is strictly regulated during the different phases of the cell-cycle. (ii) We detail for the first time the association of p300 and PCAF/GCN5 to cell-cycle promoters in vivo. The distinct timing of binding of these two HAT families suggests different functions. (iii) Although confirming and extending previous ChIP data on E2Fs, HDACs, and histones H3-H4 acetylation, we further detect E2F4/6 on the G 2 /M Cyclin B2 promoter, thus establishing that promoters active in all phases of the cycle are regulated by E2Fs.
The bulk of the data presented here on E2F1/4/6 in Fig. 2 are in large agreement with previous experiments (2-5) performed on the Cdc2, Cyclin A, and E2F1 promoters in human systems and on Cdc2 in the mouse NIH3T3 cells. Here we find that E2F4 is also bound to Cyclin B2 in G 0 and early G 1 , with the repressive E2F6 becoming the predominant E2F in this latter phase, concomitant with the binding of HDACs, histones H3-H4 deacetylation, and promoter repression. This switch among repressive E2Fs has not been described before and might be peculiar to G 2 /M genes. Interestingly, E2F binding fits well with the activity of the CDE-CHR promoter elements (16 and 19 and references therein). These cis-acting sites, found in Cyclin B1/B2, Cdc2, CDC25 and Cyclin A, have been detailed with genetic experiments as required to keep promoters at bay until S phase. We propose that E2Fs are part of the elusive CDE-CHR binding and regulating activities and that S/G 2 promoters are under negative regulation in G 0 -G 1 through E2F4/6 binding, with activators binding after E2Fs/HDACs repression is relieved in late G 1 . In Cyclin B2, a gap in late G 1 is observed between the release of repressors and NF-Y binding, during which histones, particularly H3, are hyperacetylated (Fig. 3). Fig. 2.

FIG. 4. In vivo analysis of NF-Y binding to cell-cycle promoters in cycling NIH3T3 (left Panel) or G 1 /S arrested with R-roscovitine (right panel).
This is in agreement with the association of PCAF (Fig. 1), but it is unclear at the moment which factor(s) is bound to DNA in this phase and recruits the HAT activity. It might be one of the activating E2Fs that have not been analyzed here (E2F2/3) or a protein such as MYC, which binds in vivo to an E box in the related Cyclin B1 promoter (36).
The highly dynamic role of NF-Y described in this set of promoters does not match previous experiments on inducible systems, which are constitutively bound by NF-Y (22,23). Our data on the rapidly inducible JunB and c-Jun reinforce the notion that NF-Y binding is a prerequisite for promoter activation and precedes stable p300 binding. NF-Y is generally required to set the chromatin stage in the near proximity of transcriptional start site(s), thanks to its peculiar histone fold nature, and allow further activators/coactivators buildup (20,29). Because E2Fs regulate most, if not all, cell-cycle promoters and CCAAT boxes are often adjacent, the two proteins are likely to interplay. In the E2F1 promoter, for example, NF-Y might still be required for E2F1 association or function, as suggested by transfection studies (9), whereas E2F4 might actually prevent binding of NF-Y until late G 1 . In G 2 /M promoters, which contain at least two and sometimes more NF-Y binding sites, whose distance is highly conserved and constant (10 -12, 15, 17), two or three NF-Y might be self-sufficient, without the need of an activating E2F. Alternatively, one of the activating E2Fs not analyzed here might partake in activation through the CDE-CHR.
One of the fundamental questions raised by our results is how NF-Y discriminates between the different targets in vivo in any given phase of the cell-cycle. Two important facts should be remembered: (i) the overall CCAAT-binding activity of the trimer, as assayed in vitro in EMSAs, is nearly equally abundant in all phases of the cycle (10), and (ii) all the binding sites analyzed here are optimal in terms of NF-Y affinity in vitro. What then directs NF-Y to PDGF␤-R in G 0 , E2F1 in late G 1 , and Cyclin B2 in G 2 ? A first mechanism might be promoterdriven, that is, the specific positioning of the CCAAT boxes with respect to nucleosomes influence NF-Y binding. This is suggested by in vitro experiments in which recombinant NF-Y easily accesses a well positioned nucleosome on the tissuespecific Major Histocompatibility Complex class II Ea promoter (37), but it is incapable of doing so on a similar nucleosome reconstituted on the Cyclin B2 promoter, despite the presence of three CCAAT boxes. 3 A second mechanism might point to specific post-translational modifications, such as the acetylation of NF-YB detected in Xenopus (29), which might alter promoter penetrance of the trimer. Finally, the overall acetylation status of nucleosomes could locally influence association, as suggested by our finding here that Cyclin B2 nucleosomes are maximally acetylated before NF-Y binding. These three hypotheses are most likely not mutually exclusive and pose a formidable task at dissecting the mechanistic details.
Another important aspect of our data concerns the association of HATs to cell-cycle promoters. The role of p300/CBP in the control of cell growth is exemplified by the finding that (i) E1A mutants that cannot bind to p300 exhibit defective cellular transformation; (ii) p300 Ϫ/Ϫ and cbp Ϫ/Ϫ knockout mice have impairments in cell-cycle regulation and differentiation; and (iii) the HAT activity is required for G 1 /S transition, by regulating E2Fs directly through acetylation and phosphorylation (38,39, and reviewed in Ref. 40). The proteins mediating p300 recruitment might vary in a promoter-specific way. On E2F1, a combination of E2F1 and NF-Y might be required, whereas in Cyclin B2 the multiple NF-Ys might be sufficient, as suggested by in vitro experiments (17).
The association of the p300-related CBP and of PCAF/GCN5 was detailed in studies on the inducible IFN-␤ and on the tissue-specific ␣ 1 -AT promoters (41)(42)(43). On IFN-␤, GCN5 association precedes that of CBP, which is concomitant with that of PolII and is followed by TBP and TBP-associated factors (TAFs) recruitment (41), on ␣ 1 -AT, TBP, TAFs, and PolII are on before PCAF and CBP association (43). Our findings are more reminiscent of the IFN-␤ system because the recruitment of PCAF/GCN5 is manifestly antecedent to that of p300. Ample evidence exists of the PCAF/GCN5 involvement in cell-cycle control (reviewed in Ref. 44). Because of the early timing of their association, it is tempting to speculate that these HATs actually help recruit NF-Y on some promoters, and not viceversa. Unlike for p300, NF-Y has a high intrinsic affinity for PCAF/GCN5, as these HATs are found in NF-Y immunopurified fractions; the interactions are mediated by the NF-YB-NF-YC HFM subunits (26). What then recruits PCAF/GCN5? In some promoters, such as E2F1 and Cdc2, it might be one of the E2Fs, but the picture is likely to be more complex, as these HATs exist as a multi-subunit complex (also known as TFTC or STAGA) formed by 15 or more proteins, including many HFMcontaining TAFs (45)(46)(47). Functional interactions between E2Fs, GCN5, and TRRAP (the largest subunit of the complex) have been described (48). It is currently unknown whether the composition of the complex actually changes in vivo during cell-cycle progression and whether different subcomplexes specifically target different classes of promoters. In addition to PCAF, NF-Y is able to bind in vitro to some of the components of the complex, such as the HFM-containing TAF12, possibly a partner of the H2A-like STAF42 (47). Other H3/H4-likes (the PAF65␣-TAF9 presumed dimer) might interact with NF-Y as TAF6-TAF9 do (49). In the IFN-␤ system cited above, NFkB and ATF2 serve as a pre-bound platform for HATs recruitment (41), whereas in the ␣ 1 -AT promoter only Hepstatic Nuclear Factor 1 is constitutively bound, and the DNA-binding protein Hepstatic Nuclear Factor 4 becomes associated later (43). In this respect NF-Y behaves like this latter factor being associated shortly before activation. Thus, we favor a scenario in which transitory recruitment of the PCAF complex through E2Fs, presumably subsequent to the release of inhibitory components of the E2F/retinoblastoma/HDAC complexes, followed by a burst of histone acetylation, favors efficient NF-Y binding, which in turn is required for p300 association, the latter phenomenon coinciding with transcriptional activation. Finally, the displacement of NF-Y might be instrumental in turning off transcription (see PDGF␤-R, cJun, JunB, E2F1). Given the stunning stability of NF-Y-CCAAT complexes in vitro, we suspect that an active mechanism must be operational to remove it in vivo. This might involve modifications of the protein of nearby nucleosomes or the activity of chromatin remodellers. It will now be interesting to assess the association of other subunits of the PCAF complex, of chromatin remodelling machines, as well as of general transcription factors in this well characterized experimental setting.