Regulation of the p53 transcriptional response by structurally diverse core promoters

Abstract

p53 target promoters are structurally diverse and display pronounced differences in RNA polymerase II (RNAP II) occupancy even in unstressed cells, with higher levels observed on cell cycle arrest genes (p21) compared with apoptotic genes (Fas/APO1). This occupancy correlates well with their ability to undergo rapid or delayed stress induction. To understand the basis for such distinct temporal assembly of transcription complexes, we examined the role of core promoter structures in this process. We find that the p21 core promoter directs rapid, TATA box-dependent assembly of RNAP II preinitiation complexes (PICs), but permits few rounds of RNAP II reinitiation. In contrast, PIC formation at the Fas/APO1 core promoter is very inefficient but supports multiple rounds of transcription. We define a downstream element within the Fas/APO1 core promoter that is essential for its activation, and identify nuclear transcription factor Y (NF-Y) as its binding partner. NF-Y acts as a bifunctional transcription factor that regulates basal expression of Fas/APO1 in vivo. Thus, two critical parameters of the stress-induced p53 transcriptional response are the kinetics of gene induction and duration of expression through frequent reinitiation. These features are intrinsic, DNA-encoded properties of diverse core promoters that may be fundamental to anticipatory programming of p53 response genes upon stress.

The ability of cells to undergo cell cycle arrest or apoptosis after acquiring malignant alterations is of fundamental importance to our normal surveillance mechanisms that are designed to prevent tumor progression. The tumor suppressor protein p53 is a critical component of this anti-tumor response by regulating diverse gene pathways that control cell cycle arrest, angiogenesis, DNA repair, senescence, and apoptosis (Murray-Zmijewski et al. 2008; Vousden 2009; Vousden and Prives 2009). Induction of cell cycle arrest (at G1–S) by p53 results from transcriptional activation of CDKN1A (p21), leading to inhibition of cyclin-dependent kinase (CDK)–cyclin complexes and proliferating nuclear antigen (PCNA) (el-Deiry et al. 1993; Abbas and Dutta 2009). The molecular events that promote p53-dependent apoptosis are more complex and occur through activation of critical genes involved in the mitochondrial apoptotic pathway (PUMA) and the death receptor pathway (Fas/APO1). The kinetics of expression after p53 induction varies considerably among different target genes, with those involved in cell cycle control being expressed early and proapoptotic genes being expressed later (Zhao et al. 2000).

A critical issue that remains to be elucidated is how p53 chooses which of its multiple target genes to activate or repress in response to a given stress. In this regard, an important source of p53 functional diversity that could contribute to selective gene regulation and cell fate choice resides within the core promoters of p53 target genes. The core promoter is defined as the DNA sequence required to direct accurate transcriptional initiation by the RNA polymerase II (RNAP II) complex. It contains the region around the initiation site and usually one or more conserved sequence motifs such as the TATA box, initiator (Inr), TFIIB recognition element (BRE), downstream promoter element (DPE), and downstream core element (DCE) that impose different requirements for transcription initiation (Heintzman and Ren 2007; Sandelin et al. 2007; Juven-Gershon and Kadonaga 2009). The series of regulatory events that direct the activity of p53 target promoters must ultimately relay through the basal RNAP II machinery. Thus, it is important to understand not only the relationship of p53 to the RNAP II complex, but also how architectural diversity among its promoters affects this relationship and contributes to the overall stress-induced transcriptional program.

Previous studies have shown that different levels of RNAP II transcription preinitiation complexes (PICs) are assembled on endogenous p53 target promoters even before stress induction (Espinosa et al. 2003). These levels correlate with the timing of transcription activation during the stress response. The pro-cell cycle arrest gene p21 contains high levels of RNAP II and other initiation components in unstressed cells and is rapidly induced by DNA-damaging agents. This is achieved by conversion of RNAP II to an elongating form through recruitment of elongation factors to distinct regions of the p21 gene (Espinosa et al. 2003). In contrast, proapoptotic genes like Fas/APO1, PUMA, and APAF-1 contain very low levels of RNAP II and display delayed induction kinetics relative to p21. These genes are more likely to be controlled at the level of initiation. Interestingly, high levels of promoter-bound RNAP II complexes do not correlate with the duration of gene expression during the damage response, since mRNA synthesis from proapoptotic genes can equal or exceed that of p21. Prolonged transcription after damage may depend, in part, on the efficiency of RNAP II reinitiation from specific promoters. In addition, p53 is required to assemble the RNAP II complex on the endogenous p21 promoter before stress, and to differentially recruit and retain specific initiation and elongation factors after distinct types of DNA damage in vivo (Espinosa et al. 2003; Gomes et al. 2006). However, since p53 has been shown to interact with both p21 and proapoptotic genes before stress (Szak et al. 2001; Kaeser and Iggo 2002; Espinosa et al. 2003), the variation in promoter structure among these genes is likely to influence the composition and rate of assembly of promoter-specific RNAP II transcription complexes. This, in turn, affects whether activation of specific genes occurs with early or delayed kinetics, and how long expression is sustained during the stress response.

We investigated the role of core promoter architecture in directing the transcriptional kinetics of two functionally and structurally diverse p53 target genes, p21 and Fas/APO1, using biochemical and cell-based approaches. Surprisingly, we find that differences in RNAP II affinity and reinitiation kinetics observed between the endogenous p21 and Fas/APO1 genes can be recapitulated in vitro using DNA templates in a manner dependent on their respective core promoters. We find that the p21 core promoter has a high intrinsic affinity for RNAP II and rapidly assembles PICs in vitro, whereas the opposite is observed for Fas/APO1. Although Fas/APO1 has a low affinity for RNAP II, once a PIC is formed, this core promoter directs very efficient reinitiation of transcription, in marked contrast to p21, which reinitiates poorly. Consequently, affinity for RNAP II does not necessarily dictate prolonged gene expression through high reinitiation frequency, and these two important aspects of transcription are independently regulated. Upon mutagenesis of the p21 and Fas/APO1 core promoters, we defined the TATA box in p21 and a downstream element in Fas as key regulators of promoter-specific RNAP II activity. Additionally, we identified the bifunctional transcription factor NF-Y (nuclear transcription factor Y) as specifically interacting with the Fas downstream element and regulating basal expression of endogenous Fas/APO1 in vivo.

These studies support the notion that diverse core promoters among p53 target genes play a fundamental role in stress-induced cell fate decisions by regulating the kinetics of gene activation and the duration of expression through RNAP II reinitiation. Interestingly, these intrinsic features of distinct core promoter structures are “hard-wired” into their DNA without the need for chromatin, p53, or coactivators, which act at other levels of control. Thus, the programming of p53 response genes in unstressed cells to be “poised” by engaged RNAP II for rapid activation or “unpoised” for delayed, but sustained expression is, in part, genetically determined to anticipate how and when these genes will need to function in the stress response. This default, intrinsic programming can be modulated by epigenetic events and specific p53 and coactivator complexes that tailor gene activity and cell fate choices in a tissue- and stress-specific manner (Espinosa 2008). Thus, core promoter architecture provides another important mechanistic level by which p53 coordinates a physiologically appropriate response to diverse stress conditions.

Results

Intrinsic features of diverse p53 core promoters regulate differences in RNAP II-binding affinity and reinitiation kinetics

To understand the function of diverse core promoter structures that exist among p53 target genes, we examined which features may influence the differences in apparent RNAP II-binding affinities and expression rates observed among p53 target genes using an in vitro transcription system. Initially, we assumed that p53 interaction with chromatin-assembled recombinant genes would be required to recapitulate these aspects of regulation. We began by simply measuring the relative amounts of transcription obtained from cloned natural p21 and Fas/APO1 promoter–reporter plasmids as naked DNA in vitro using protein extracts from unstressed HeLa cells. These promoters have very dissimilar structures, but each directed efficient RNA synthesis in vitro, with p21 being the stronger template (Fig. 1A,B). Although the Fas/APO1 gene has been reported to have multiple start sites within a relatively close region of ∼200 base pairs (bp), only one major initiation site was used under in vitro transcription conditions. This mapped to ∼80 bp upstream of the ATG start codon and within 4 bp of previously reported start site predictions and RACE analysis (Behrmann et al. 1994; Cheng et al. 1995).

Figure 1.

Functional characterization of the p21 and Fas/APO1 core promoters. (A) Schematic diagram of p21 and Fas/APO1 promoters, including their respective p53-binding sites. (B) In vitro transcription of p21 or Fas/APO1 plasmids was performed and analyzed by primer extension. (C) Schematic diagram of single- and multiple-round transcription in the presence of sarkosyl. Addition of sarkosyl to form PICs prevents new PIC formation. Templates without sarkosyl undergo multiple rounds of transcription. (D) Rates of PIC formation were analyzed by adding sarkosyl just before (time 0), or various times after, combining DNA templates with HeLa nuclear extracts (HNEs). Transcription was initiated by adding NTPs. (E) Quantification and graphical representation of D. (F) Rounds of transcription were analyzed by allowing PICs to form for 1 h before initiating transcription (time 0). The initiated reactions then underwent multiple rounds of transcription before sarkosyl addition at intervals from 15 sec to 30 min. The first lane shows that sarkosyl addition to DNA templates before exposure to HeLa extract prevents formation of functional PICs. (G) Rounds of transcription were also measured using DNA templates immobilized on magnetic beads. Extract and immobilized DNA were mixed for 1 h to allow PIC formation similar to C. For single-round transcription, the DNA template was washed with buffer followed by addition of NTPs to allow elongation, whereas this wash step was not included for multiple rounds of transcription. (H) Rounds of transcription were quantified, and the signal at each time point was compared with time 0 to calculate the rounds of reinitiation.

To define the relative promoter strength and affinity of the transcriptional machinery, the p21 and Fas/APO1 DNA templates were transcribed after PIC formation in the presence of the detergent sarkosyl, which limits transcription to a single round by inhibiting RNAP II reinitiation (diagrammed in Fig. 1C; Cai and Luse 1987; Hawley and Roeder 1987; Kadonaga 1990). By performing kinetic analyses of transcription in the presence and absence of sarkosyl, we could measure two important parameters: the rate of PIC formation, and the number of rounds of reinitiation directed by each promoter. To analyze the rate of PIC formation, PICs were allowed to assemble on each promoter between 0 and 2 h, with further PIC formation or RNAP II reinitiation blocked by the addition of sarkosyl to 0.04%. Immediately afterward, transcription was initiated by addition of nucleotide triphosphates (NTPs) (Fig. 1D,E). This revealed that the p21 promoter was activated as early as 5 min after addition of NTPs, reaching a plateau at ∼1 h. In contrast, the Fas/APO1 promoter was much less efficient than p21 in assembling PICs, with negligible transcription detected under these single-round conditions even after 2 h (Fig. 1D,E). We also observe similar kinetics on the APAF-1 promoter (Supplemental Fig. 2). These results, on naked DNA templates, were surprisingly consistent with the chromatin immunoprecipitation (ChIP) analysis in unstressed cells, showing dramatic differences in levels of RNAP II binding to the endogenous p21 and Fas/APO1 promoters (Espinosa et al. 2003). This suggests that variations in RNAP II-binding affinities are directed by intrinsic features of p53 target promoters encoded in their DNA sequence, since neither chromatin nor p53 was required for this level of basic regulation, although both obviously modulate transcriptional activity at more complex stages.

Next, we analyzed the RNAP II reinitiation frequency by measuring the approximate rounds of transcription directed from each promoter in vitro. PIC formation was allowed to occur on each template for 1 h, followed by the addition of NTPs to initiate RNA synthesis. Once transcription was initiated, we conducted a time course of sarkosyl addition from 0 to 30 min to stop further reinitiation at specific time intervals (Fig. 1F). These results were quantified using a PhosphorImager by comparing the transcriptional output at each time point with that of a single round at time 0. Unexpectedly, this analysis revealed that, although the p21 promoter can assemble a PIC far more rapidly than Fas/APO1 (Fig. 1E), it is much less efficient in directing RNAP II reinitiation (four rounds) than Fas/APO1 (21 rounds). Previous in vitro assays using sarkosyl on more classical promoters have reported a maximal two to six rounds of transcription using mammalian or Drosophila protein extracts (Hawley and Roeder 1987; Kadonaga 1990). To confirm our results, we tested whether differences in the general sensitivity of sarkosyl toward transcription of p21 and Fas/APO1 existed, but found none (Supplemental Fig. 1). As a more rigorous verification of the large difference in reinitiation that we observed between p21 and Fas/APO1, we developed an assay to estimate rounds of transcription without using sarkosyl. First, the p21 or Fas/APO1 templates were immobilized on magnetic beads and incubated with HeLa extracts to form a PIC on each promoter. Then, instead of adding sarkosyl to eliminate reinitiation, unbound RNAP II and other proteins were washed away from each template with buffer before starting transcription by the addition of NTPs (Fig. 1G). By this means, only engaged RNAP II complexes at the time of washing were capable of initiating single-round transcription, whereas templates left with excess unbound RNAP II (unwashed) could reinitiate multiple rounds of RNA synthesis. The results of this experiment were quantified as before and closely matched that obtained under sarkosyl conditions, with p21 undergoing seven rounds of transcription and Fas/APO1 undergoing 18 rounds (Fig. 1H).

The preceding experiments were performed using full-length, natural p21 and Fas/APO1 recombinant promoters (Fig. 1A). We now asked whether the pronounced differences in RNAP II behavior observed between the two templates were regulated by DNA sequences within or outside of their respective core promoters. To this end, we generated templates containing core promoter sequences from −149 to +42 of p21 and from −50 to +78 of Fas/APO1 (Fig. 2A). Both the full-length and core promoters of each gene generated approximately equal amounts of RNA synthesis (Fig. 2B), consistent with the general observation that in vitro transcription of DNA templates mainly reflects their core promoter activity. A time course of sarkosyl addition during PIC formation followed by NTP addition, similar to that shown in Figure 1D, demonstrated that the core promoters of p21 and Fas/APO1 direct PIC formation and reinitiation efficiency in an identical manner as that observed with the full-length promoters (Fig. 2C). The activities were quantified, and a graphical representation of the results is presented (Supplemental Fig. 5). Taken together, these results support the notion that the differences in levels of RNAP II binding to p53 target genes in unstressed cells and the kinetics of gene expression upon stress induction are strongly influenced by intrinsic features of structurally diverse core promoters at the level of DNA.

Figure 2.

Analysis of PIC assembly kinetics on full-length or core p21 and Fas/APO1 promoters. (A) Diagram of the full-length promoters containing p53 response elements or core promoters. (B) In vitro transcription reactions using HeLa extracts to compare promoter strength between full-length and core promoter templates. (C) Rates of PIC formation on p21 and Fas/APO1 full-length or core promoters were measured by in vitro transcription under conditions similar to Figure 1D.

Mapping critical elements within the p21 and Fas/APO1 core promoters

To extend our observation that the p21 and Fas/APO1 core promoters were each sufficient to direct dissimilar RNAP II-binding kinetics (Fig. 2), we mapped the functional elements within each promoter by scanning mutagenesis of progressive 10-bp blocks to the transverse bases using a PCR-based strategy (Invitrogen) (Fig. 3A,B). All mutations were generated in the context of the full-length promoters and were assessed by in vitro transcription using HeLa extracts. An examination of the p21 promoter set revealed that the TATA box was the most critical sequence in the core region, since mutation of this element decreased transcription to a negligible extent (Fig. 3C, Scan C). Mutation of other sequences had far less deleterious effects, except to change the major transcriptional start site, such as Scan D (adjacent to the TATA box) and Scan F (impairs the Initiator), or create additional initiation sites like Scan B (disrupts an Sp1 site flanking the TATA box) and Scan E (disrupts sequence just upstream of the start site).

Figure 3.

Functional core promoter elements mapped by scanning mutagenesis. (A,B) Diagram of p21 (A) and Fas/APO1 (B) promoter sequences indicating the location of scanning mutations. Progressive 10-bp transversion mutations were generated in the context of the full-length promoters. (C) In vitro transcription reactions of p21 scanning mutations. The two left lanes labeled “G” and “A” contain DNA sequencing reactions used to map the transcription start site(s). (D) Same as C but using Fas/APO1 scanning mutations. (E) Transcriptional analysis of the scanning mutations “F” and “G” of Fas/APO1 that were further defined by creating four 5-bp block mutations between Scans F and G to create Fas scan mutants F.1, F.2, G.1, and G.2. (F) Luciferase expression analysis in HCT116 cells using transiently transfected p21 or Fas/APO-1 core promoters within a pGL3 luciferase plasmid.

A similar analysis of the Fas/APO1 promoter set revealed little change in transcriptional efficiency of templates containing scanning mutations of sequences from −52 to the Initiator region (Scans A–E) (Fig. 3B). Interestingly, mutation of sequences just 3′ of the Initiator from +7 (Scans F–G) completely abolished transcription, whereas adjacent 3′ sequences (Scans H–I) had no effect. To further define this essential region, transversion mutations in consecutive blocks of 5 bp were created within Scans F–G. After transcriptional analysis of these mutated templates, the essential sequences for Fas/APO1 expression in vitro were localized to a 15-bp element (Scans F.2, G.1, and G.2) residing between +12 to +26 (Fig. 3E). The reactions were quantified, and a graphical representation of the results is shown (Supplemental Fig. 6). To verify that this region was also responsible for basal transcription in vivo, we created a luciferase reporter system using the core promoters of p21 and Fas/APO1 and compared the wild-type activity with the mutated regions important for transcription (Fig. 3F). Similar to previous studies, the TATA-box mutation (p21 C) drastically reduced luciferase activity from the p21 promoter. The core promoter mutations within the Fas Scan F and Scan G mutations also decreased the basal activity of Fas/APO1 to varying degrees. Thus, our mutagenesis studies identified the most critical sequences for p21 and Fas/APO1 intrinsic core promoter function as the p21 TATA box and a Fas/APO1 downstream element.

Analysis of the TATA box and Fas/APO1 downstream elements using chimeric p21 and Fas/APO1 promoters

Next, we examined whether the distinct characteristics of RNAP II binding and reinitiation observed on the p21 and Fas/APO1 core promoters were regulated by the p21 TATA box or the Fas/APO1 downstream element. To address this, a chimeric promoter called Fas-TATA was constructed in which the p21 TATA sequence (ATATCAG) was inserted into the Fas/APO1 promoter to replace the region from −29 to −23 (GAGGCCA) and generate the sequence TATATCAGG beginning at −30 (Fig. 4A). A functional analysis of the Fas-TATA promoter template revealed that it was much more efficiently transcribed in vitro than templates containing the wild-type Fas/APO1 (Fas-wt) promoter (Fig. 4B). Moreover, a time course of sarkosyl addition under single-round transcription conditions, analogous to the experiment shown in Figure 1D, showed that the Fas-TATA promoter was capable of forming detectable functional PICs much faster, like p21 but unlike Fas/APO1 wild type, and could still direct multiple rounds of reinitiation (Fig. 4C; represented graphically in Supplemental Fig. 7). Thus, the chimeric promoter has the dual capacity of high-affinity PIC formation, conferred by the TATA box, and efficient rates of reinitiation, like Fas/APO1. Interestingly, mutation of the Fas downstream element (Scan F) within the Fas-TATA hybrid template (“F-TATA”) resulted in a severe reduction of transcription (Fig. 4D). This indicates that the TATA box can synergize with the Fas downstream element within the context of the Fas/APO1 full-length promoter to greatly stimulate expression, but that the downstream element is an essential feature that cannot be functionally replaced by the TATA box. No sequences homologous to the Fas downstream element have been found within the p21 core promoter, and none of our scanning p21 mutations significantly impaired TATA function, suggesting that the TFIID complex can act relatively independently in these in vitro assays (Fig. 3A).

Figure 4.

Analysis of critical core promoter elements using chimeric templates. (A) Diagram of the Fas/APO1 promoter sequence indicating the location in which the p21 ATATCAG sequence was inserted to replace −23 to −29 and create the Fas-TATA promoter. (B) In vitro transcription to examine the activity of Fas-TATA compared with the Fas/APO1 promoter. (C) In vitro transcription analyzing the rate of PIC formation of p21, Fas/APO1, and Fas-TATA. HeLa nuclear extract was allowed to bind to template for 0–2 h before addition of sarkosyl, similar to Figure 2D. To generate multiple rounds of transcription, sarkosyl was not added to the last lane (2 h*). (D) Transcriptional analysis of Fas-TATA compared with F-TATA (Fas-TATA with the Scan F mutation). (E) Analysis of the Fas downstream element within the p21 wild-type or mutated full-length promoter.

We further analyzed the function of the Fas downstream element (Scans F–G) by placing that sequence (+7 to +26 bp) in the p21 promoter at the corresponding location from +7 to +26 (Fig. 4E; represented graphically in Supplemental Fig. 7). We discovered that the Fas downstream element repressed p21 transcription, and this was mediated specifically by the Scan F sequence +7 to +16 bp (Fig. 4E, cf. lanes 2 and 3). When either the Scan F or G downstream element was inserted in p21 promoters lacking a functional TATA box, only negligible transcription was obtained. This indicates that, in the context of the p21 promoter, unlike the Fas/APO1 promoter, the Fas downstream element cannot synergize with the TATA box or functionally replace it (Fig. 4E, lanes 6–8).

The Fas/APO1 downstream core promoter element specifically interacts with NF-Y

We next wished to identify which proteins associated with the critical core promoter elements of p21 and Fas/APO1. A DNase protection analysis demonstrated that both native TFIID and recombinant TATA-binding protein (TBP) could interact with the p21 TATA box; however, no binding was observed on Fas/APO1. This substantiates the notion that significant differences in binding affinities for general transcriptional factors exist between these diverse core promoters, which may result in the assembly of compositionally distinct PICs (Supplemental Fig. 3). To determine whether the DPE that is essential for Fas/APO1 transcription specifically interacts with proteins present in our HeLa transcription extracts, we performed an electrophoretic mobility shift assay (EMSA) using oligonucleotides that span −1 to +35 bp, which includes the sequences in Scans F–G (Fig. 5A). We observed specific protein complex formation using wild-type oligos that was not generated with oligos containing mutations in the 20 bp corresponding to elements F–G (Fig. 5B). Although a different shift was observed using mutated oligos, this is likely to be a distinct protein, since mutations were generated across the functional sequences (middle 20 bp out of 36), and even a 10-fold excess of mutant oligo failed to compete for binding of the protein to the wild-type oligo (Fig. 5C).

Figure 5.

The NF-Y complex can bind to the critical Fas downstream element. (A) Sequences of the wild-type or mutant probes used for EMSA. (B) EMSA analysis of Fas downstream element (Fas “F–G”)-binding protein(s) from HeLa nuclear extracts. (C) EMSA competition assay using cold wild-type or mutated competitor F–G oligonucleotides. (First lane) Wild-type Fas “F–G” oligos were shifted with 4 μg of HNE in the absence or presence of cold wild-type or mutated competitor DNA. (D) Diagram indicating the steps to capture and characterize the complex binding the Fas downstream element. The DNA affinity pull-down was performed using immobilized multimers of the DNA sequence used for EMSA. (E) Proteins that bound to the wild-type Fas downstream element or the mutated sequence were step-eluted with buffer containing 0.1 M, 0.25 M, 0.5 M, or 1 M NaCl and tested for binding activity by EMSA. (F) Supershift analysis to test the specificity of NF-Y binding. In lanes 1–7, EMSA reactions were incubated for 30 min followed by addition of the specified antibody and incubated for 15 min. For lanes 8 and 9, antibodies were incubated with HNE for 10 min before mixing with DNA probe.

We next identified and characterized the protein(s) binding to this DNA region by oligonucleotide affinity chromatography, similar to what has been used previously to purify a corepressor of p53, SnoN, on the α-fetoprotein gene (Wilkinson et al. 2005). In this approach, DNA sequences comprising the Fas activator element (Scans F–G) that are capable of protein binding by EMSA were immobilized as multiple copies on streptavidin-coated magnetic beads. The multimers were produced using a self-primer PCR method (Hemat and McEntee 1994; Yaneva and Tempst 2006). Protein–DNA-binding reactions were conducted in the presence of the synthetic polymer dI–dC to reduce nonspecific interactions, and specific protein–DNA complexes were captured with a magnet to remove unbound material. We also incorporated a “preclearing” step in which HeLa extracts plus dI–dC were initially incubated with the immobilized, mutated Fas downstream element (which abolishes specific protein binding as determined by EMSA) to further minimize contaminating proteins in our subsequent analysis. Proteins that “flowed through” the precleared chromatographic steps were enriched for factors that specifically interact with the Fas downstream element rather than those that spuriously bind to mutated sequences. The immobilized templates were then incubated with HeLa extract containing the potential Fas downstream element-interacting protein(s), washed, and step-eluted with 0.1, 0.25, 0.5, and 1 M NaCl buffer (diagrammed in Fig. 5D). The eluates were analyzed by EMSA, which showed that the downstream element-binding activity eluted mainly between 0.1 and 0.25 M NaCl (Fig. 5E).

These fractions, eluted from immobilized wild-type and mutated downstream element oligos, were compared by electrophoresis on SDS-PAGE. Analysis of silver-stained proteins revealed a specific band between 49 and 64 kDa in the 0.1 and 0.25 M NaCl elutions that was not bound to the mutated template (Supplemental Fig. 3A). This protein band (along with the corresponding area of the mutant lane in Supplemental Fig. 3A) was excised from the gel, digested into peptides, and analyzed by MALDI-TOF mass spectrometry. One of the top candidates from the mass spectrometry analysis was NF-Y subunit C, which was not detected in the corresponding mutant oligo sample. Western blot analysis against the NF-Y subunit C confirmed the mass spectrometry results (Supplemental Fig. 3B). NF-Y is a trimeric complex that is comprised of subunits A, B, and C (Mantovani 1999). To verify that the trimeric complex actually binds to the downstream Fas promoter element, we performed a supershift EMSA assay using antibodies against two of the three NF-Y subunits (anti-NF-YA and anti-NF-YC), using anti-YY1 and anti-Bmi as controls. The specific band observed previously by EMSA was clearly supershifted using either anti-NF-YA or anti-NF-YC antibody, whereas no supershift was observed with the control antibodies (Fig. 5F). The reactions in lanes 8 and 9 of Figure 5F are similar to lanes 5 and 7, respectively, except that HeLa extract was incubated with the oligo probe before antibody addition.

NF-Y binds to the Fas/APO1 promoter in vivo and activates Fas/APO1, but not p21, transcription

NF-Y is a conserved transcription factor that binds with high specificity to CCAAT (or reverse: ATTGG) motifs in the promoter regions in a variety of genes. The C subunit forms a tight dimer with the B subunit, a prerequisite for association with subunit A. The resulting trimer binds to DNA with high specificity and affinity and is responsible for transcriptional regulation of numerous promoters (McNabb et al. 1995; Sinha et al. 1995; Caretti et al. 2003). NF-Y became an obvious candidate after its identification from the chromatographic elutions because the DNA sequence within the critical promoter region for Fas/APO1 transcription contains “…GGGTTGGTGG…” and is very similar, although not identical, to the reverse ATTGG NF-Y consensus sequence. To examine the physiological relevance of NF-Y to Fas/APO1 gene regulation, we used the human breast cancer MCF7 cell line, since previous studies have shown that Fas/APO1 plays a role in apoptosis in these cells (Ruiz-Ruiz et al. 2003; Hernandez-Vargas et al. 2006). Upon treatment of MCF7 cells with 5-fluorouracil (5-FU) for 14 h to induce DNA damage, Fas/APO1 mRNA synthesis was highly elevated (Fig. 6A). A ChIP analysis was performed in these cells, in the presence or absence of 5-FU, to determine whether NF-Y was associated with the Fas/APO1 promoter in vivo. For this purpose, we used an antibody to the NF-Y subunit A and mapped protein binding with primers near the transcriptional start site of Fas/APO1 and a control region far downstream. The p21 and CCNB1 promoters were used as negative and positive controls, respectively, for NF-YA interaction. These studies show that NF-YA is bound near the Fas/APO1 initiator region (Inr) in unstressed MCF7 cells, and this interaction increases upon 5-FU stress induction. In contrast, negligible NF-YA binding was observed on the p21 promoter in treated and untreated MCF7 cells. A strong ChIP signal to NF-YA was also generated on the known target promoter CCNB1 (Fig. 6B). These ChIP results are consistent with our biochemical data showing specific binding of NF-Y to the Fas downstream core promoter element by EMSA and recruitment to immobilized Fas/APO1 templates (Fig. 5), and thereby confirm the biological relevance of this protein–DNA interaction. We next analyzed whether NF-Y influenced Fas/APO1 transcription in vivo. To this end, the three subunits of NF-Y were each overexpressed to similar levels in MCF7 cells using CMV-driven cDNA constructs (Fig. 6C), and mRNA levels of the endogenous Fas/APO1 gene were subsequently measured by quantitative PCR (qPCR). We found that overexpression of the NF-Y trimeric complex activated Fas/APO1 mRNA nearly twofold, whereas transcription of endogenous p21 and a control gene, SDHA, remained unchanged (Fig. 6D). This suggests that NF-Y positively regulates Fas/APO1 expression and can function selectively on p53 target genes, consistent with our biochemical studies. These results also support the notion that the specialized roles of NF-Y and other components of the basal transcription machinery in regulating the p53 response are directed by diverse core promoter structures within p53 target genes.

Figure 6.

NF-Y binds to and regulates the FAS/APO-1 promoter in MCF7 cells. (A) mRNA analysis of Fas/APO1 activation upon 5-FU (50 ng/mL) treatment for 14 h. (B) ChIPs of the FAS/APO-1, p21, and CCNB1 genes were assayed for the presence of NF-Y in MCF7 cells. (C) Western blot analysis of NF-Y subunits overexpressed in MCF7 cells. CMV-driven NF-YA, NF-YB, and NF-YC plasmids were used for transfection and assayed with subunit-specific antibodies. (D) mRNA from MCF7 cells overexpressing NF-Y were analyzed by qPCR.

Discussion

In unstressed cells, certain p53 target promoters, like p21, are “preloaded” with paused RNAP II, whereas proapoptotic promoters, among others, have negligible RNAP II association. Such striking variation in levels of promoter-bound RNAP II may have direct bearing on the differential activation kinetics observed after stress induction of p53-responsive genes (Espinosa et al. 2003). The existence of such regulatory mechanisms acting before DNA damage to establish a default programmatic transcriptional response to stress is very intriguing. Our data suggest that the intrinsic properties of diverse p53 core promoters play a key role in regulating RNAP II affinity and dynamics to coordinate appropriate responses to different stress conditions. We find an unexpected level of transcriptional regulation governing RNAP II dynamics that is encoded within the DNA sequence of diverse core promoters that drives expression of p53-responsive genes (Fig. 7). The TATA box within the p21 promoter has a critical role in recruiting the transcriptional machinery by promoting rapid formation of a functional PIC that is poised for initiation. However, the p21 core promoter is intrinsically inefficient for reinitiation, which may be enhanced by signal-dependent components acting at other levels of regulation to facilitate PIC reformation and prolonged RNA synthesis. In contrast to p21, the Fas/APO1 promoter does not contain a TATA box or other well-characterized core motifs, and the rate of PIC formation is very slow. A Fas downstream element that binds to NF-Y is essential for core promoter activity in vitro, and may nucleate PIC assembly by direct interaction with the general transcription machinery. Surprisingly, once transcription is engaged, the Fas/APO1 promoter is capable of efficient RNAP II reinitiation events. Published reports have demonstrated that initiation and reinitiation can be experimentally uncoupled and, in one example, reinitiation is faster than initiation (Jiang and Gralla 1993), which we also observe with Fas/APO1.

Figure 7.

Schematic model of p21 and Fas/APO1 PIC formation and reinitiation kinetics. The p21 core promoter supports efficient PIC assembly through the TATA box for rapid transcriptional activation, but only poor reinitiation capability. In contrast to p21, the Fas/APO1 core promoter has low affinity for PIC recruitment but supports multiple reinitiation events. The downstream NF-Y-binding element is required for core promoter activity and may facilitate nucleation of the PIC.

Transcriptional activation of genes responsible for important cellular events such as cell cycle arrest and apoptosis must be highly regulated. For example, it is advantageous for the cell to have a poised RNAP II on the p21 promoter to ensure its rapid activation upon stress signaling in order to quickly halt cell cycle progression. In contrast, it is equally advantageous to program proapoptotic genes, like Fas/APO1, for slow transcriptional initiation followed by rapid reinitiation as a safeguard against inappropriately timed cell death. Thus, a genetic system that is “hard-wired” within the DNA sequence of core promoters may be evolutionarily beneficial to protect cells against unwarranted cell proliferation after damage or apoptosis when damage can be repaired. For Fas/APO1, and potentially other proapoptotic genes, having a fail-safe mechanism embedded in the promoter that prevents rapid transcription may be required to act as a buffer until critical signaling thresholds are surpassed. Once this point is reached, apoptotic genes can be actively transcribed and undergo multiple rounds of reinitiation to guarantee sufficient mRNA synthesis to drive cell suicide. Our data reveal that architecturally distinct core promoters, independent of chromatin or p53 binding, contain the information to impart these properties.

Role of the TATA box in p21 promoter regulation

The TATA box was the first core promoter element to be identified and was originally believed to be required for all promoters. However, recent studies indicate that as few as 10% of all core promoters actually have a TATA sequence (Suzuki et al. 2001; Yang et al. 2007). For TATA-containing genes, this motif has been shown to be important in transcriptional activity and PIC formation, with TFIID recruited before other general PIC components (Mathis and Chambon 1981; O'Shea-Greenfield and Smale 1992; Ranish et al. 1999). The core promoter of p21 contains a nearly perfect TATA sequence at about −30 bp that is capable of interacting with either TBP or the TFIID holocomplex (Supplemental Fig. 3). We show that the TATA box is not only critical for p21 transcription, but it enables the p21 promoter to efficiently assemble a poised PIC before transcriptional initiation. This intrinsic feature presumably facilitates the rapid stress-induced p21 expression kinetics observed in vivo. Interestingly, ChIP analyses of the p21 promoter in wild-type and p53-null cells show greatly reduced PIC formation in cells lacking p53 (Espinosa et al. 2003). In this case, p53 may function to relieve chromatin compaction through recruitment of cofactors, such as p300, to generate greater nucleosome accessibility for PIC formation before stress. In addition, specific post-translational modifications of p53 may also facilitate increased PIC interaction with the p21 promoter. In this regard, acetylation of Lys 373 and Lys 382 on p53 has been shown to directly promote TFIID binding to the p21 core promoter (Li et al. 2007).

Regulation of the FasAPO1 core promoter downstream element by NF-Y

One explanation for the ability of the Fas/APO1 promoter to undergo multiple rounds of transcription in vitro is through a particular multifunctional protein complex. Using immobilized DNA affinity assays, we captured and identified NF-Y as specifically binding to the Fas DPE. The physiological relevance was confirmed by demonstrating that NF-Y interacts with the endogenous Fas/APO1 promoter and that overexpression of the NF-Y complex can stimulate basal Fas/APO1 transcription in vivo. Interestingly, NF-Y is known to activate or repress numerous promoters and associate with general transcription factors such as RNAP II, TFIID, TBP, and PC4, as well as cofactors such as histone acetyl transferases (HATs) and histone deacetylases (HDACs) (Bellorini et al. 1997; Currie 1998; Caretti et al. 2003; Peng and Jahroudi 2003; Kabe et al. 2005). Previous studies have shown that certain promoters have faster reinitiation rates by using a “scaffold,” consisting of TFIID, TFIIA, TFIIH, TFIIE, and Mediator, which remains stably bound to the promoter after the first round of transcription (Zawel et al. 1995; Yudkovsky et al. 2000). This scaffold may facilitate RNAP II reinitiation by avoiding the requirement to reassemble a full PIC after each round of transcription. Our in vitro observations of slow initiation followed by efficient, multiple rounds of transcription from the Fas/APO1, but not p21, promoter are consistent with the existence of a scaffold. However, scaffold retention on Fas/APO1 may be achieved by an unconventional mechanism, since this promoter does not contain typical core elements, but instead uses a near-perfect NF-Y-binding site (an inverted CCAAT box centered at +20). Because NF-Y can associate with TFIID, TBP, and RNAP II, it may facilitate nucleation of PIC components and/or help the core promoter to retain a partial PIC “scaffold” to enhance reinitiation events. Another possibility is that NF-Y bends the DNA sequence in such a way as to increase transcriptional output from the core promoter. In this regard, the observation from chimeric promoters that the NF-Y-binding motif (Fas downstream element) works positively with the TATA box in the Fas/APO1 promoter but negatively when it is placed in the p21 promoter suggests that NF-Y has a unique activity and must function in particular promoter contexts, underscoring its importance.

As a gatekeeper of cell growth and division, p53 must orchestrate the activities of numerous factors, such as NF-Y, that regulate transcription to ensure orderly cell cycle progression. Several studies have implicated a relationship between p53 and NF-Y in which p53 interacts directly with NF-Y to repress various cell cycle genes (Imbriano et al. 2005). Interestingly, NF-Y knockdown by siRNA causes apoptosis while activating many p53 target genes. We analyzed RNA levels in NF-Y knockdown cells and observed activation of Fas/APO1, p21, and PUMA genes, possibly by indirect induction (Supplemental Fig. 8). This is consistent with previous studies, although Fas/APO1 expression was not measured (Benatti et al. 2008). Activation of p53 target genes is not entirely surprising, since Benatti et al. (2008) demonstrated that NF-Y depletion from HCT116 cells resulted in down-regulation of 478 genes and up-regulation of 803 genes, indicating that NF-Y knockdown affects multiple genes in a global manner. We hypothesize that NF-Y is a critical, functionally diverse transcription factor and, upon its cellular depletion, a crisis ensues that results in apoptosis. The multifunctional nature of NF-Y on its distinct target genes may be conferred by phosphorylation of NF-YA by CDK2 (Chae et al. 2004). Interestingly, p53-dependent activation of p21, an inhibitor of CDK2, may create a regulatory loop that ultimately affects NF-Y phosphorylation and activity (Yun et al. 2003). With regard to Fas/APO1, we speculate that stress-dependent gene activation requires a positive interaction between promoter-bound p53 and NF-Y, and the involvement of post-translational modifications as well as specific interactions with Mediator, TFIID-associated factors (TAFs), or chromatin-modifying enzymes to generate a fully regulated transcriptional response (An et al. 2004; Donner et al. 2007; Sullivan and Lu 2007).

Core promoter diversity

Our results provide insight into how TATA-less promoters are transcribed. Previous studies demonstrated that TATA-less promoters rely on other core elements that interact directly with various TAFs (Martinez et al. 1994; Burke and Kadonaga 1996; Smale and Kadonaga 2003). Here we show that Fas/APO1 can use an alternative promoter element that associates with NF-Y and potentially facilitates PIC assembly. In addition, our analysis of chimeric promoters demonstrates the importance of the relative context of core elements. An inserted TATA box within Fas/APO1 increases transcription but cannot functionally replace the Fas downstream element. However, the Fas downstream element placed within the p21 promoter negatively affects transcription and does not rescue a mutated TATA box. It is clear that the interplay between distinct core promoter elements confers another level of promoter regulation and can result in diverse transcriptional outputs. Recent studies have shed light on how these core regulatory circuits control RNAP II activity. One seminal report demonstrated cross-talk between the TATA box and DPE in which factors bound specifically to one element had an inhibitory effect on transcription initiated at the other (Hsu et al. 2008). Furthermore, TFIID has been shown to regulate promoters through two distinct motifs—the DPE and DCE—in a phosphorylation-dependent manner driven by CK2 (Lewis et al. 2005). This is significant because it demonstrates that extracellular signals can impact gene activity by determining which core element a given transcription factor will function through.

The diversity of core promoter structures within p53-regulated genes implies that each gene may be uniquely regulated. However, the fact that most apoptotic target genes have much lower levels of poised RNAP II compared with cell cycle arrest genes suggests that there are some general similarities. A comparison of core promoters from several p53 target genes reveals that most cell cycle arrest genes have focused promoters (single start sites and typical core elements), whereas most apoptotic genes including Fas/APO1, PUMA, and APAF-1 appear to resemble dispersed promoters (several start sites over 50–100 nucleotides and few typical core elements) (Juven-Gershon and Kadonaga 2009; JM Morachis, E Scheeff, and G Manning, unpubl.). We speculate that differences in core promoter structures exist to establish a default transcriptional state, with respect to PIC formation and dynamics, which insures an appropriate p53 programmatic response to diverse forms of stress.

Thus, two critical parameters of p53-dependent gene activation—the kinetics of induction and duration of expression through frequent reinitiation—are intrinsic, DNA-encoded features of diverse core promoters that may be fundamental to anticipatory programming of p53 response genes. The default mode, as seen at the p21 promoter, is to rapidly form a PIC but undergo few rounds of reinitiation, whereas that of Fas/APO1 is the opposite. Of course, sustained p21 expression requiring multiple rounds of RNAP II reinitiation and reduced Fas/APO1 expression by infrequent reinitiation can be achieved by overriding the default programming through sophisticated epigenetic processes that are tailored to specific stress environments. Considering the advantage of preserving flexibility to fine-tune cell fate decisions, having a default, genetic program embedded in core promoter DNA would safeguard against misregulation, particularly of apoptotic genes. This may reflect an evolutionary need to balance cell growth while limiting the ability to self-destruct. It is difficult to envision a cellular system that would evolve to activate cell cycle arrest and apoptotic genes identically. If this were true, apoptosis would likely override the cell cycle arrest program without allowing the cell to recover from stress or DNA damage. Further investigation into the default mechanisms used by structurally diverse p53 target genes may provide insight into how the programmatic response to stress is regulated and how it can be manipulated for targeted therapies.

Materials and methods

In vitro transcription assays

Nuclear protein extracts from HeLa cells were prepared as described (Dignam et al. 1983). Transcription reactions included 10 μL (∼5–6 μg/μL) of HeLa nuclear extract (HNE), 15 μL of HeLa Dialysis Buffer (HDB) (20 mM Hepes at pH 7.9, 50 mM KCl, 1 mM DTT, 0.2 mM EDTA, 10% glycerol), and 25 μL of transcription mix (0.4 mg/mL BSA, 20 mM HEPES at pH 7.9, 70 mM KCl, 3 mM DTT, 1.2 mM NTPs, 1–3 mM MgCl₂, 0.5 μL of RNase inhibitor per reaction), and 500 ng of DNA templates. For the PIC kinetics analyses, NTPs were omitted from the transcription mix, and the PIC was allowed to form for 0 min to 2 h at 25°C before adding 2 μL of NTP mix to start the reaction (final volumes were adjusted with HDB). Sarkosyl (Sigma-Aldrich) was introduced in each reaction by adding 2 μL of a 1% stock to 0.04% final concentration unless otherwise noted. Transcription reactions were incubated in a 30°C water bath or at room temperature. Reactions were stopped and processed using reagents from Zymo Research (RNA Clean-up Kit-5) by adding 200 μL of RNA-binding buffer, applying the mixture to columns, washing two times with wash buffer, and then eluting with 8 μL of RNase-free water. Primer extension was performed by adding 3 μL of primer annealing mix (10 mM Tris, 1 mM EDTA, 1.25 M KCl) to each reaction and heating for 2–3 min at 75°C in heating blocks. Reactions were removed from the heat blocks and allowed to cool slowly to ∼37°C. This was followed by addition of 23 μL of reverse transcription mix (20 mM Tris-HCl at pH 8, 10 mM MgCl₂, 0.1 mg/mL Actinomycin D, 5 mM DTT, 0.33 mM dNTP) and 0.5 μL of M-MLV Reverse Transcriptase (Promega) per reaction, and incubation for 1 h at 37°C. Final reactions were precipitated and washed with ethanol and placed in a speedvac for 5 min. DNA pellets were each resuspended in 10 μL of formamide with 1 mM EDTA/0.1 mM NaOH (2:1) and heated for 2–3 min to 95°C followed by snap-cooling on ice. Samples were electrophoresed through 8% polyacrylamide/TBE gels (SequaGel-8, National Diagnostics). For transcription reactions using immobilized DNA templates, plasmids containing the p21 and Fas full promoters were first linearized by restriction enzyme cleavage with NotI, followed by cleavage with a second restriction enzyme, EcoRI, to generate sticky ends that were filled in by Klenow DNAP with biotinylated dATP and dUTP. After removal of excess nucleotides, the biotinylated fragments were incubated with streptavidin-coated magnetic beads (Dynal, Invitrogen) and purified from unbiotinylated DNA using a magnet. The immobilized p21 or Fas/APO1 DNA templates (250 fmol) were each incubated with HNE for in vitro transcription as described.

Plasmids and mutagenesis

The p21 (−2.4 kb to +42 bp) and Fas/APO1 (−1 kb to +700 bp) promoters were each cloned into pBSKII plasmids using EcoRI and XbaI and named p21DPE-A and FasMut2, respectively. For the “core” p21 and Fas promoters, PCR was used to generate the sequence of −149 to 42 bp for p21 and −50 to 78 bp for Fas and then subcloned back into pBSKII. All mutational analyses were performed by progressively generating 10-bp transversion mutations in the context of the full-length promoters using the GeneTailor site-directed mutagenesis system following the manufacturer's instructions (Invitrogen). Fas-TATA was created using Fas/APO1 promoter sequence and inserting the p21 ATATCAG sequence to replace −23 to −29 and create the Fas-TATA promoter. F-TATA was created using Fas-TATA and mutating the Scan F region as in Scan F of Figure 3B. p21 wild-type (TATA) or p21 Scan C (mutated TATA) promoters were mutated by inserting the F, G, or F + G elements of Fas/APO1 at the same position with respect to the transcription start site (from +7 to +26 bp). All of the templates used for in vitro transcription contained a 27-bp sequence from the luciferase gene (5′-GCGTCTTCCATTTTACCAACAGTACCG-3′) that was used for primer extension. For the luciferase reporter assays, the core promoters of p21 (−134 to +42 bp) and Fas (−130 to +46 bp) were generated by PCR, and each was subcloned into pGL3 reporter plasmids.

EMSA

Protein–DNA-binding reactions included 2 μg of HNE, 250 ng of dI–dC, and 5 μL of premix (1 μL of 5× shift buffer [100 mM HEPES, 350 mM KCl, 25 mM MgCl₂, 15 mM DTT]), 0.5 μL of 4 mg/mL BSA, and 1.5 μL 15% Ficoll adjusted to 10-μL final volume with HDB. The premix was incubated for 15 min at 4°C followed by the addition of 1 μL (15 fmol/μL) of ³²P-labeled oligonucleotide. Binding reactions were performed for 30 min at room temperature. For antibody “supershift” assays, 2 μg of anti-NF-YA (C-18), anti-NF-YC (H-120), anti-YY1 (H-414), and anti-Bmi-1 (C-20) from Santa Cruz Biotechnologies were incubated with binding reactions for 15 min before gel loading. Samples were then electrophoresed through 4% PAGE (39:1 acrylamide/bis) in TBE and scanned using a Fuji FLA-5100 PhosphorImager.

DNA affinity protein chromatography

5′-Biotin-labeled DNA primers (Supplemental Table 1) were used in an efficient PCR method (Hemat and McEntee 1994) to generate ∼500-bp multimers of the wild-type and mutated DNA sequences used in the EMSA reactions. The stock bead reaction contained 5 μg of DNA and 2 μg of Streptavidin-coated magnetic beads (Dynal, Invitrogen) in 400-μL final volume. For the recruitment assay, HeLa nuclear extracts were first precleared by mixing 100 μL (∼600 μg) of HNE with 100 μL of “premix” (similar to EMSA buffer) and passing the reaction through 30 μL (∼450 ng) of immobilized mutated DNA three times. The precleared extract was then separated from the beads, mixed with immobilized wild-type or mutated DNA, and allowed to bind for 30 min at room temperature. Reactions were washed three times with HDB containing 0.1 mg/mL of single- and double-stranded Escherichia coli DNA. Bound proteins were then eluted with 20 μL of HDB containing 100 mM, 250 mM, 500 mM, or 1 M NaCl. These fractions were analyzed for protein composition by SDS-PAGE (Western blot, silver staining, and mass spectrometry) and for specific DNA binding by EMSA.

Mass spectrometry

A detailed description of the mass spectrometry used is included in the Supplemental Material.

Transfection for dual luciferase assay and NF-Y overexpression

HCT116 cells were plated in a 12-well plate and grown at 70% confluency for transfection. For luciferase assays, 1.8 μg of pGL3, p21-pGL3, or Fas-pGL3 core promoter constructs and 200 ng of renilla (pRL-TK) were used per well for transfection with Fugene HD (Roche). After transfection, cells were maintained in 500 μL of DMEM growth media overnight. Cells were then lysed in 100 μL of 1× passive lysis buffer (PLB buffer, Promega) for 15 min at room temperature. For dual luciferase assays, 20 μL of lysate per well were aliquoted into a 96-well plate and analyzed using firefly luciferase (50 μL of LAR II) and renilla (50 μL of stop and glow buffer). All analyses were performed in duplicate, with each experiment performed at least twice.

For overexpression of NF-Y, MCF7 cells were plated in 12-well plates and grown to 80% confluency for transfection. We used NF-YA(SC112917), NF-YB(SC116285), and NF-YC(SC112622) human cDNA clones in pCMV6-XL5 plasmids (Origene) and pCMV-GPF as a control. Fugene HD (Roche) transfection reagent was used at an 8:2 DNA:reagent ratio, transfecting 20 μL of transfection mix and 800 ng of total DNA. Cells were collected 24 h after transfection and processed for mRNA purification (Qiagen) or lysed for Western blot analysis.

Real-time RT-PCR reactions

Cells transfected with NF-Y expression plasmids were collected 24 h later, and total RNA was prepared with the Qiagen RNAeasy Kit. RT reactions were performed with SuperScriptase III (Invitrogen) using the random hexamer protocol following the manufacturer's instructions. Primers for FAS/APO-1, p21, and SDHA (a control housekeeping gene) were used to analyze gene expression using SYBR Green, and values were normalized to β-actin.

ChIP assays

MCF7 cultures were grown to 50%–60% confluency and were treated with 50 ng/mL 5-FU (Sigma-Aldrich) for 14 h. After washing with PBS, cells were cross-linked with a 1% formaldehyde/PBS solution for 15 min at room temperature. Cross-linking was stopped by addition of glycine to 125 mM final concentration. Cells were washed twice with cold PBS and harvested in RIPA buffer (150 mM NaCl, 1% Igepal CA-630, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris-HCl at pH 8, 5 mM EDTA, 20 mM NaF, 0.2 mM sodium orthovanadate, 5 μM trichostatin A, 5 mM sodium butyrate, protease inhibitors). Samples were sonicated to generate DNA fragments <1000 bp. For immunoprecipitation, 1 mg of protein extract was precleared for 1 h with 40 μL of 50% A/G protein–Sepharose slurry before addition of indicated antibodies. The antibodies used were anti-NF-YA (H-209) or rabbit IgG (Santa Cruz Biotechnologies). Each antibody (2 μg) was added to the samples and incubated overnight at 4°C in the presence of 40 μL of protein G beads preblocked with 1 mg/mL BSA and 0.3 mg/mL salmon sperm DNA. Beads were washed once with RIPA buffer, three times with ChIP wash buffer (100 mM Tris-HCl at pH 8.5, 500 mM LiCL, 1% [v/v] NP-40, 1% [w/v] deoxycholic acid), once again with RIPA buffer, and twice with 1× TE. Immunocomplexes were eluted for 10 min at 65°C with 1% SDS, and cross-linking was reversed by adjusting to 200 mM NaCl and incubating 5 h at 65°C. DNA was purified, and a fraction was used as template in real-time PCR reactions.

Western blot analysis

Proteins were electrophoresed through 10% SDS-PAGE and transferred onto nitrocellulose membranes (Amersham Biosciences) before being probed with the following antibodies: anti-NF-YA (H-209), anti-NF-YB (FL-207), anti-NF-YC (H-120), anti-RNAP II (H224) (Santa Cruz Biotechnologies).