Histone Cross-talk Connects Protein Phosphatase 1α (PP1α) and Histone Deacetylase (HDAC) Pathways to Regulate the Functional Transition of Bromodomain-containing 4 (BRD4) for Inducible Gene Expression*

Background: Transcription elongation is a rate-limiting step for inducible gene expression. BRD4 must be released from chromatin to regulate transcription elongation. Results: Protein phosphatase 1α (PP1α) and histone deacetylases (HDAC) signaling pathways are required for this process. Conclusion: Histone cross-talk in trans between H3S10ph and H4K5ac/K8ac connects PP1α and HDACs signaling pathways to control functional transition of BRD4. Significance: BRD4 is regulated epigenetically for controlling stress-induced gene expression. Transcription elongation has been recognized as a rate-limiting step for the expression of signal-inducible genes. Through recruitment of positive transcription elongation factor P-TEFb, the bromodomain-containing protein BRD4 plays critical roles in regulating the transcription elongation of a vast array of inducible genes that are important for multiple cellular processes. The diverse biological roles of BRD4 have been proposed to rely on its functional transition between chromatin targeting and transcription regulation. The signaling pathways and the molecular mechanism for regulating this transition process, however, are largely unknown. Here, we report a novel role of phosphorylated Ser10 of histone H3 (H3S10ph) in governing the functional transition of BRD4. We identified that the acetylated lysines 5 and 8 of nucleosomal histone H4 (H4K5ac/K8ac) is the BRD4 binding site, and the protein phosphatase PP1α and class I histone deacetylase (HDAC1/2/3) signaling pathways are essential for the stress-induced BRD4 release from chromatin. In the unstressed state, phosphorylated H3S10 prevents the deacetylation of nucleosomal H4K5ac/K8ac by HDAC1/2/3, thereby locking up the majority of BRD4 onto chromatin. Upon stress, PP1α-mediated dephosphorylation of H3S10ph allows the deacetylation of nucleosomal H4K5ac/K8ac by HDAC1/2/3, thereby leading to the release of chromatin-bound BRD4 for subsequent recruitment of P-TEFb to enhance the expression of inducible genes. Therefore, our study revealed a novel mechanism that the histone cross-talk between H3S10ph and H4K5ac/K8ac connects PP1α and HDACs to govern the functional transition of BRD4. Combined with previous studies on the regulation of P-TEFb activation, the intricate signaling network for the tight control of transcription elongation is established.

Transcription elongation has been recognized as a rate-limiting step for the expression of signal-inducible genes. Through recruitment of positive transcription elongation factor P-TEFb, the bromodomain-containing protein BRD4 plays critical roles in regulating the transcription elongation of a vast array of inducible genes that are important for multiple cellular processes. The diverse biological roles of BRD4 have been proposed to rely on its functional transition between chromatin targeting and transcription regulation. The signaling pathways and the molecular mechanism for regulating this transition process, however, are largely unknown. Here, we report a novel role of phosphorylated Ser 10 of histone H3 (H3S10ph) in governing the functional transition of BRD4. We identified that the acetylated lysines 5 and 8 of nucleosomal histone H4 (H4K5ac/K8ac) is the BRD4 binding site, and the protein phosphatase PP1␣ and class I histone deacetylase (HDAC1/2/3) signaling pathways are essential for the stress-induced BRD4 release from chromatin. In the unstressed state, phosphorylated H3S10 prevents the deacetylation of nucleosomal H4K5ac/K8ac by HDAC1/2/3, thereby locking up the majority of BRD4 onto chromatin. Upon stress, PP1␣-mediated dephosphorylation of H3S10ph allows the deacetylation of nucleosomal H4K5ac/K8ac by HDAC1/2/3, thereby leading to the release of chromatin-bound BRD4 for subsequent recruitment of P-TEFb to enhance the expression of inducible genes. Therefore, our study revealed a novel mecha-nism that the histone cross-talk between H3S10ph and H4K5ac/ K8ac connects PP1␣ and HDACs to govern the functional transition of BRD4. Combined with previous studies on the regulation of P-TEFb activation, the intricate signaling network for the tight control of transcription elongation is established.
In eukaryotic cells, the transcription of protein-coding genes is performed by RNA polymerase II (Pol II), 4 which is tightly regulated in multiple stages. Although control of the promoter recruitment of RNA Pol II for transcription initiation has been a long held paradigm for transcription regulation, recent studies indicate that transcription elongation is another rate-limiting step for the rapid expression of signal-inducible genes in metazoans (1)(2)(3)(4). Genome-wide surveys revealed that for a vast array of inducible genes, transcription initiation has already been accomplished even at the uninduced state, yet RNA Pol II is stalled at promoter-proximal regions (5)(6)(7)(8)(9). Upon stimulation, P-TEFb (positive transcription elongation factor b), a heterodimeric kinase consisting of CDK9 and mostly Cyclin T1, is recruited to these regions to convert the poised Pol II into a productive elongation mode, thereby triggering the synthesis of full-length mRNA (2,10,11). Hence, the recruitment of active P-TEFb emerges as a key point for modulating inducible gene expression. In cells, the activity of P-TEFb is tightly regulated, with the majority of P-TEFb being sequestrated in an inactive complex (termed 7SK snRNP) that also contains 7SK snRNA and nuclear proteins HEXIM1, MePCE, and LARP7 (12,13). During stress response, P-TEFb is liberated from 7SK snRNP through the concerted actions of signal-activated protein phos-phatase PP2B and PP1␣, thereby making P-TEFb available for promoter recruitment (14).
Bromodomain-containing protein BRD4, a ubiquitously expressed nuclear protein, is a key recruitment factor for P-TEFb (15,16). BRD4 belongs to the BET protein family, the members of which have a consensus structure of two N-terminal tandem bromodomains (BDI and BDII) followed by an extra terminal domain (17). Early studies indicated that the bromodomains of BRD4 are essential for its association with acetylated chromatin (18,19). A recent study using systematic peptide array screening combined with crystallography revealed that the BDI of BRD4, which simultaneously binds to diverse diacetylated histones, is the major contributor for its interaction with chromatin (20). Consistent with its chromatin-targeting nature, BRD4 was found to be persistently associated with acetylated chromosomes during mitosis in many cell lines (18,19). Later studies demonstrated that this association is crucial for the rapid expression of early G 1 genes upon exiting mitosis (21)(22)(23)(24). Although the mechanism is still unclear, BRD4 has been proposed to function in transmitting the epigenetic memory across cell division by marking the transcriptionally active genes (23,(25)(26)(27). Moreover, the association of BRD4 with mitotic chromosome was shown to be important for maintaining chromatin compaction (28), and facilitating the even segregation of viral DNAs into daughter cells during cell division (29).
The diverse biological roles of BRD4 have been proposed to rely on its functional transition between chromatin targeting and transcriptional regulation (17,25). In line with this notion, our recent study showed that, besides P-TEFb, the availability of BRD4 is also highly regulated (31). The survey of a variety of cell lines showed that almost all BRD4 is associated with chromatin in interphase cells and there is limited chromatin-free BRD4 available in the unstimulated state. Upon stress, BRD4 is released from chromatin, and subsequently it recruits P-TEFb to enhance expression of the HIV-1 gene (31). The signaling pathways and molecular mechanism for this process, however, was unknown.
Cell lines, Transfection, and Infection-HEK293T, HeLa, and HeLa-based F1C2(CDK9-f) cells stably expressing FLAGtagged CDK9 subunit of P-TEFb, and HeLa cells with an integrated HIV-LTR-luciferase reporter gene (HIV-LTR-Luc) were maintained as previously described (14,31,48,49). Cells at ϳ80% confluence were transfected with various cDNA constructs using a PEI transfection protocol as described previously (14). For puromycin selection, the constructs were cotransfected at a ratio of 5:1 with pBabe-puro vector that harbors a puromycin-resistant gene. Two days after transfection, the cells were selected in medium containing 1 g/ml of puromycin for 36 -48 h. The lentiviral infection was performed as described previously (31). For silencing PP1␣, two shRNA lentiviral constructs were used at a 1:1 ratio.
Treatment of Cells with UV or Pharmacological Compounds-Cells at ϳ50% confluence were preincubated with solvent, or the inhibitor trichostatin A (400 nM, 2 h), MS-275 (5 M, 2 h), cyclosporin A (5 M, 1 h), or microcystin LR (15 nM, 15 min), followed by treatment with 5 g/ml of DOX (1 h), 10 mM HMBA (2 h). For UV irradiation, cells were exposed to 254 nm UV at 80 J/m 2 without culture medium, and then incubated in the original medium for 1 h. Afterward, the cells were subjected to stepwise extraction with modified nuclear fractionation (MONF) protocol (31).
Preparation of LSF, HSF, NLF, and LSEN with MONF Protocol-The low salt fraction (LSF), high salt fraction (HSF), nuclear lysate (NLF), and low salt-extracted nuclei (LSEN) were prepared with MONF (31). Briefly, the buffer A-swollen cells were subjected to low salt extraction by incubating with 2ϫ packed cell volume of buffer A (with 1% Nonidet P-40) to break the plasma membrane. After centrifugation, the supernatant was saved. The nuclei pellet was further extracted twice by resuspending in 1ϫ packed cell volume of low salt buffer (75 mM), followed by centrifugation. The supernatants of the three extractions were combined and saved as the LSF. The LSEN were subsequently incubated with 6ϫ packed cell volume of high salt buffer (0.3 M). After centrifugation, the supernatant was saved as the HSF. The high salt-extracted nuclei were lysed with SDS-loading buffer to yield NLF.
Immunopurification-FLAG-tagged proteins and their associated factors were isolated by anti-FLAG immunoprecipitation from LSF or HSF of F1C2 (CDK9-f) cells or transfected HeLa cells as described previously (31,48). The levels of proteins were determined by Western blotting. To obtain stressactivated HDAC1, -2, or -3 enzymes, HEK293T cells transfected with FLAG-or HA-tagged HDAC expression vectors were treated with UV, DOX, or HMBA, followed by anti-FLAG or anti-HA affinity purification (48). To isolate intact nucleosomes, HeLa cells were infected with lentivirus expressing FLAG-tagged histone H3 or H4 for at least 96 h, and the LSEN prepared from ϳ10 7 cells were digested with micrococcal nuclease (10 units in a total of 300 l of low salt buffer) at 37°C for 20 min to release mononucleosomes into supernatant. After centrifugation at 10,000 ϫ g, 4°C for 5 min, the supernatant was subjected to anti-FLAG affinity isolation of nucleosome. To obtain H3/H4 dimers, the LSEN were digested with 20 units of micrococcal nuclease at 37°C for 1 h. The supernatant was adjusted to a salt concentration of 0.8 M NaCl to disrupt nucleosomes before anti-FLAG affinity isolation of H3/H4 dimers. To purify histone proteins, cells were lysed with high salt cell lysis buffer (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.5 mM EDTA, 0.5 M NaCl, 1% Triton X-100, 1 mM DTT, 0.5 mM PMSF, 1ϫ protease inhibitor mixture), and sonicated before affinity isolation. The purity of isolated H3 was analyzed by Western blotting to verify that the H3 preparation is free of H4, and vice versa.
In Vitro Assay for BRD4 Release-To determine the effect of PP1␣ and/or HDAC enzymatic treatments on BRD4 release from chromatin or nucleosomes, the LSEN or nucleosomes were immobilized on anti-FLAG resin via H3-f or H4-f, incubated with 1.0 unit of recombinant PP1␣ and/or stress-activated f-HDAC1, -2, or -3 enzymes in 40 l of 1ϫ PP1 reaction buffer at 30°C for 1 h. After centrifugation at 10,000 ϫ g for 2 min at 4°C, the supernatant and nuclear pellet or immobilized nucleosomes were saved for Western blotting. For two-round sequential incubation of LSEN, extensive washes with 1ϫ PP1 reaction buffer was performed between the two incubations.
In Vitro Pulldown Assay with H3/H4 Dimers-The H3/H4 dimers were isolated as described above, and immobilized on anti-FLAG beads via H3-f or H4-f. After extensive washes with D0.8 M buffer and then D0.1 M buffer, the immobilized H3/H4 dimers were preincubated with 2 g/l of BSA (15 l of beads in a 20-l volume) for 15 min, followed by incubation with affinity-purified HA-tagged BRD4 at 30°C for 15 min. The unbound proteins were washed away with D0.1 M buffer and the immobilized complex was eluted with FLAG peptide for Western blotting to determine the H3/H4 dimer-associated BRD4.
Luciferase Assay-HeLa cells with an integrated HIV-LTRluciferase reporter gene (HIV-LTR-Luc) were infected with lentivirus expressing the desired proteins for 48 -96 h, followed by incubation with 5 mM HMBA for 3-4 h as indicated. Cell lysates were prepared for measuring the luciferase activity (48). Data were represented as fold-enhancement compared with untreated cells. The error bars were standard deviations based on three independent experiments.
Chromatin Immunoprecipitation (ChIP)-ChIP was performed using HeLa cells containing an integrated HIV-LTR-Luc as described previously (31).  GCA TGA TCA GTA ACA, and gene body: forward, 5Ј-GCA  CAA ATA ATG GCT GAT CGT and reverse, 5Ј-TCA GAG  TCA TGT TGA CTT CTC C), c-jun (promoter region: forward, 5Ј-GCC AAC TCA TGC TAA CGC AG and reverse  5Ј-TTC TCT CCG TCG CAA CTT GT, and gene body: forward, 5Ј-GCC AAC TCA TGC TAA CGC AG and reverse,  5Ј-TTC TCT CCG TCG CAA CTT GT), and IL-8 (promoter region: forward, 5Ј-TGT CAT TGC CAG CTG TGT TG and reverse, 5Ј-AAC AAG TTT CAA CCA GCA AG). The PCR amplification was performed on Eppendorf Mastercycler ep realplex 2 with the following program: 95°C for 7 min followed by 40 cycles of 10 s at 95°C, 30 s at 60°C for annealing and extension. The results from two independent experiments were averaged and plotted as percentage of input.
Flow Cytometry Analysis-Cells were trypsinized, washed three times with PBS, and 1-2 ϫ 10 6 cells were resuspended in 0.5 ml of PBS. The cells were dropped slowly into 2 ml of cold 87.5% ethanol with gentle mixing, and kept at 4°C overnight. Before flow cytometry analysis, cells were centrifuged at 300 ϫ g for 5 min, and the cell pellet was resuspended in 400 l of PBS, digested with 100 g/ml of RNaseA at 37°C for 30 min, and incubated with 50 g/ml of propidium iodine for 30 min in the dark. To synchronize cells to metaphase, the cells were incubated with 50 ng/ml of nocodazole for 16 h. For serum starvation, the cells were cultured without serum for 40 h.

Stress Induces H4K5ac/K8ac Deacetylation and H3S10ph
Dephosphorylation-By using a MONF protocol (31) (outlined in Fig. 1A), our previous study showed that stress-induced release of chromatin-bound BRD4 is essential for the induction of HIV-1 expression (31). To identify the signal pathway(s) required for BRD4 release from chromatin, we first surveyed the changes in histone modifications after UV, DOX, or HMBA treatment (Fig. 1B). Intriguingly, concomitant with the release of chromatin-bound BRD4 as detected in the LSF (containing chromatin-free proteins), both the levels of acetylated lysine 5 and 8 of nucleosomal histone H4 (H4K5ac/K8ac) and phosphorylated serine 10 of nucleosomal histone H3 (H3S10ph) were reduced upon stress treatments (Fig. 1B). Consistently, the ChIP assay showed that HMBA treatment induced the reduction of H3S10ph and H4K5ac/K8ac at the promoter region or gene body of several endogenous genes as well as HIV-LTR-Luc reporter gene (Fig. 1C).
As the level of H3S10ph was known to be greatly increased during metaphase (50), we examined whether stress treatment reduced the H3S10ph level by blocking cell cycle progression into metaphase, or by promoting cells to exit from metaphase. As shown in Fig. 1D, 1 h after UV treatment, or DOX or HMBA treatments for 2 h, did not change the cell cycle distribution of HeLa cells. When cells were synchronized to metaphase by nocodazole, the levels of nucleosomal H3S10ph or H4K5ac/ K8ac were not affected by the stress treatments (Fig. 1E). In contrast, stress induced the H3S10ph dephosphorylation and H4K5ac/K8ac deacetylation in HCT116 cells that had been synchronized to the G 0 phase by 40 h of serum starvation (Fig. 1F). The stress-induced H3S10ph dephosphorylation was also observed in both starved and non-starved HeLa cells (data not shown). Taken together, these observations suggest that stressinduced H3S10ph dephosphorylation and H4K5ac/K8ac deacetylation occur only in interphase and are not due to changes in cell cycle distribution.
H4K5ac/K8ac Are the Sites for BRD4 Binding to Nucleosome in Vivo-BRD4 has been shown to associate with H4K5ac/K8ac by an in vitro peptide binding assay (20). Here we tested whether these two sites are required for the chromatin association of BRD4 in vivo. To this end, we generated mutant histone H4 by substituting lysine 5 and/or 8 with arginine (K5R/K8R). HeLa cells were infected with lentivirus expressing FLAGtagged wild type (wt-) or K5R/K8R double mutant (2M-) H4, followed by treatment with trichostatin A, a broad spectrum HDAC inhibitor, to ensure the maximal association of BRD4 with acetylated chromatin (31). The nucleosomes were immunoaffinity purified and the level of associated BRD4 was examined by Western blot (outlined in Fig. 2A). Notably, for the nucleosomes containing about half the amount of 2M-H4, the level of associated BRD4 was also reduced to half (Fig. 2B), suggesting that H4K5ac/K8ac are necessary for the binding of BRD4 to nucleosome. Moreover, we performed an in vitro pulldown assay to test the association of affinity purified HA-BRD4 with f-H4/H3 dimers that were immobilized via FLAG-tagged H4 (outlined in Fig. 2C). H4/H3 dimers containing wt, K5R, or K8R H4 retained a large amount of HA-BRD4, but those containing 2M-H4 retained little HA-BRD4 (Fig. 2D). These data indicate that H4K5ac/K8ac is essential for the in vivo binding of BRD4 to nucleosome.
HDAC1/2/3 Are Required for Stress-induced BRD4 Release and Inducible Gene Expression-Next, we identified which HDAC was responsible for H4K5ac/K8ac deacetylation and BRD4 release during stress treatments. Humans have 11 HDACs belonging to four classes (51). We first focused on the ubiquitously expressed class I HDAC that includes HDAC1, -2, -3, and -8. Pre-treating cells with MS-275, which preferentially inhibits HDAC1/2/3 but not HDAC8 or other classes of HDACs (52), blocked stress-induced BRD4 release from chromatin with comparable efficiency to the wide spectrum HDAC inhibitor trichostatin A (Fig. 3A), suggesting that HDAC1/2/3 are required for this process. Unexpectedly, knocking down any one (data not shown) or any two of the three HDACs by shRNA (Fig. 3B) failed to block stress-induced release of chromatinbound BRD4. Only when all three HDACs were knocked down was the stress-induced BRD4 release impaired (Fig. 3C), suggesting that HDAC1/2/3 are functionally redundant for stressinduced BRD4 release.
In line with the HDAC knockdown experiments, ectopically expressing any one of the enzymatically inactive mutants of HDAC1/2/3 (mt-HDAC1/2/3) failed to block stress-induced BRD4 release (data not shown), whereas co-expressing all three mt-HDACs blocked both of the stress-induced deacetylation of H4K5ac/K8ac and the release of BRD4 (Fig. 3D), as well as the HMBA-induced expression of the HIV-LTR-luciferase reporter gene (HIV-LTR-Luc) (Fig. 3E) and endogenous genes (Fig. 3F) whose transcriptions are BRD4-dependent (Fig. 3G). Notably, without stress treatment, co-expressing wt-HDACs induced neither the release of BRD4, nor the deacetylation of H4K5ac/K8ac (Fig. 3D, lane 1), indicating that the HDAC pathway is necessary but not sufficient for mediating BRD4 release.

PP1␣ Pathway Is Involved in Stress-induced BRD4 Release and Inducible Gene Expression-The observation that
H3S10ph was dephosphorylated upon stress treatment (Fig.  1B) prompted us to study its functional relevance to BRD4 release. Remarkably, when stress-induced H3S10ph dephosphorylation was blocked by pre-treatment with the PP1/ PP2A inhibitor microcystin LR (14), BRD4 release was also abolished, whereas the PP2B inhibitor cyclosporin A did not affect either process (Fig. 4A), suggesting that the PP1 or PP2A signaling pathways may be involved in this process. There are three isoforms of PP1 and two isoforms of PP2A in mammals (53). To identify which is responsible for H3S10ph dephosphorylation and BRD4 release, we examined whether ectopically expressing individual PP1 or PP2A isoforms could induce H3S10ph dephosphorylation. Significantly, overexpressing PP1␣, but not the other phosphatases or the inactive form of H66N-mutant (mt-) PP1␣, reduced the level of H3S10ph (Fig. 4B).
Next, we tested the role of PP1␣ in stress-induced BRD4 release. When PP1␣ was silenced with shPP1␣, both the stressinduced H3S10ph dephosphorylation and BRD4 release were blocked (Fig. 4C). Moreover, ectopically expressing H66N mt-PP1␣ also abolished stress-induced H3S10ph dephosphorylation and BRD4 release (Fig. 4D), as well as HMBA-induced expression of HIV-LTR-Luc (Fig. 4E) and endogenous genes (Fig. 4F). Interestingly, impairing PP1␣ activity also abolished the stress-induced deacetylation of H4K5ac/K8ac (Fig. 4, C and  D), suggesting that the PP1␣ pathway has an effect on the function of HDACs (see below). In line with these observations, overexpression of mt-PP1␣ enhanced the association of BRD4 with chromatin, as indicated by the high level of BRD4 remaining in nuclei after high salt extraction (Fig. 4G). Of note, neither knocking down PP1␣ nor overexpressing H66N mt-PP1␣ affected the expression level of BRD4 (data not shown). These data indicate that the PP1␣ pathway is also required for stressinduced BRD4 release.
PP1␣ and HDAC1/2/3 Act Sequentially to Induce BRD4 Release-Because both HDAC1/2/3 and PP1␣ pathways are involved in stress-induced BRD4 release from chromatin, we tested whether they are sufficient for this process. To this end, we employed an in vitro assay (outlined in Fig. 5A, left) by incubating the LSEN, which contains a significant amount of endogenous HDAC1/2/3 (data not shown), with recombinant PP1␣ enzyme. Interestingly, whereas PP1␣ failed to induce the release of chromatin-bound BRD4 in the LSEN that were prepared from unstressed HeLa cells (Fig. 5A, lane 2), it successfully induced BRD4 release in the LSEN prepared from stressed cells (lanes 4, 7, and 9). Remarkably, this release could be blocked by preincubating LSEN with HDAC1/2/3 inhibitor MS-275 (Fig. 5A, lanes 5, 8, and 10), suggesting that HDAC1/ 2/3 in the unstressed LSEN are functionally inactive, and only the stress-activated HDACs could induce BRD4 release. Hence, we affinity-purified HDACs from stress-treated HEK293T cells. When LSEN from unstressed cells were incubated with PP1␣ enzyme plus any one of the HDAC1/2/3 purified from stressed cells (Fig. 5B), BRD4 was released, whereas the HDACs purified from unstressed cells failed to do so (Fig. 5C). These data indicate that the stress-activated HDAC1/2/3 and PP1␣ pathways are sufficient for inducing BRD4 release from chromatin.
As impairing PP1␣ blocked HDAC-mediated H4K5ac/K8ac deacetylation in vivo (see Fig. 4, C and D), we examined whether PP1␣ acts prior to HDACs when inducing BRD4 release in vitro. LSEN from unstressed cells were subjected to two-round incubations with stress-activated HDAC3 and recombinant PP1␣ enzyme, with extensive washes between the two incubations. Remarkably, BRD4 release and H4K5ac/K8ac deacetylation were much more efficient when PP1␣ was added before HDAC3 (Fig. 5D, lane 3), indicating that PP1␣ and HDACs act sequentially to induce BRD4 release.

H3S10ph Dephosphorylation Is a Prerequisite for BRD4
Release and Inducible Gene Expression-To test whether the PP1␣-mediated H3S10ph dephosphorylation or the other PP1␣-dependent events was essential for BRD4 release, we generated a mutant histone H3 by substituting serine 10 with alanine (S10A mt-H3) to mimic the H3S10-dephosphorylated H3. If the dephosphorylation of H3S10ph is crucial for BRD4 release, it is expected that chromatin containing S10A mt-H3 might bypass requirement of the PP1␣ enzyme. Indeed, in the absence of PP1␣, HDAC1 alone induced the release of BRD4 in the LSEN containing mt-H3, but not in that containing wt-H3 (Fig. 6A). Similarly, when affinity-purified nucleosomes were used as substrate, the deacetylation of H4K5ac/K8ac and the release of BRD4 were observed only from the nucleosomes containing mt-H3, but not those containing wt-H3 (Fig. 6B).
In vivo, H66N mt-PP1␣ inhibited HMBA-induced BRD4 release in cells expressing wt-H3 (Fig. 6C, lane 3), but not in cells expressing mt-H3 (lane 4). Accordingly, in the presence of H66N mt-PP1␣, efficient induction of the expression of HIV-LTR-Luc (Fig. 6D) and endogenous genes (Fig. 6E) by HMBA were observed in cells expressing S10A mt-H3 but not wt-H3. Because the endogenous PP1␣ activity had been quenched either by dominant-negative H66N mt-PP1␣ (compare Fig. 6, A and C to E) or by pre-treatment with the PP1 inhibitor microcystin LR (Fig. 6B), these data ruled out other PP1␣-dependent events, but indicated that H3S10ph dephosphorylation per se is a prerequisite for HDAC-mediated H4K5ac/K8ac deacetylation and BRD4 release for inducible gene expression.
Coordinated BRD4 Release and P-TEFb Liberation for Transcription Elongation-Our previous study showed that PP1␣mediated dephosphorylation of T186ph at the CDK9 T-loop is critical for stress-induced liberation of P-TEFb from the inactive 7SK complex (7SK snRNP) (14). Identification of the PP1␣ pathway in BRD4 release prompted us to test whether the two processes are concerted. F1C2(CDK9-f) cells, a cell line stably expressing FLAG-CDK9 (49), were stressed and extracted stepwise into LSF and HSF using a modified nuclear fractionation protocol (31) (outlined Fig. 1A). BRD4 release was examined by Western blot assay of the level of BRD4 in an aliquot of LSF (Fig.  7A, top 2 rows), the P-TEFb liberation was determined by analyzing the loss of P-TEFb-bound HEXIM1 in anti-FLAG immunoprecipitates of the same LSF (bottom 2 rows), and interaction of P-TEFb with BRD4 and recruitment of P-TEFb⅐BRD4 complex onto chromatin was assessed by measuring the level of P-TEFb-bound BRD4 in anti-FLAG immunoprecipitates of HSF (middle 2 rows). Interestingly, stress-triggered BRD4 release, P-TEFb liberation, as well as the recruitment of the BRD4⅐P-TEFb complex in parallel (Fig. 7A), without affecting the expressing levels of P-TEFb and its associated factors (data not shown). Consistently, the ChIP assay showed that HMBA treatment induced the enrichment of BRD4 and P-TEFb at the promoter region of inducible genes (Fig. 7B).
Next, we tested the role of the PP1␣ pathway in these events by examining LSF and HSF, as shown in Fig. 4, C and D. When PP1␣ was knocked down by shRNA, in addition to the release of BRD4 (Fig. 7C, top), recruitment of the BRD4⅐P-TEFb complex (middle) and liberation of P-TEFb (bottom) were also blocked.  (WB). B, in vitro assay of BRD4 release in LSEN induced by PP1␣ and HDAC co-incubation. LSEN prepared from unstressed HeLa cells was co-incubated with the indicated enzymes. FLAG-tagged HDAC1, -2, and -3 were affinity purified from HMBA-treated HEK293T cells that had been transfected with the corresponding cDNA constructs, and analyzed by Western blot (right). C, co-incubation with PP1␣ and non-activated HDAC1, -2, or -3 is unable to induce BRD4 release from the LSEN of the unstressed HeLa cells. For lanes 1-3, the FLAG-tagged wt-HDAC1, -2, and -3 were isolated from unstressed HEK293T cells with corresponding transfections. For lanes 4 and 5, FLAG-tagged mt-HDAC3 and wt-HDAC3 were isolated from stressed HEK293T cells transfected with the corresponding constructs. D, two-round enzymatic incubation assay of LSEN. LSEN of unstressed cells was sequentially incubated with HDAC3 and PP1␣ enzymes as indicated, with extensive washes between the two incubations. The changes in histone modifications were analyzed by Western blot.
However, overexpressing H66N mt-PP1␣ did not block P-TEFb liberation (Fig. 7D, bottom), but only abolished the stress-induced BRD4 release (top) and thereby BRD4⅐P-TEFb recruitment (middle), suggesting that H66N mt-PP1␣ only has a dominant-negative effect on BRD4 release. Significantly, overexpressing mt-PP1␣ attenuated HMBA-induced transcription elongation, but not initiation, as indicated by qRT-PCR analysis of the HIV-LTR-Luc transcripts representing initiation and elongation (Fig. 7E). Moreover, impairing cellular HDACs by either HDAC inhibitor (data not shown) or shRNA knockdown (Fig. 7F) blocked the release of BRD4 (top) and recruitment of the BRD4⅐P-TEFb complex (middle), but not P-TEFb liberation (bottom). Similar to the inhibitory effect of H66N mt-PP1␣ on HMBA-induced elongation of HIV-1 (Fig.  7E), overexpressing the dominant-negative mutant HDAC1/ 2/3 attenuated the HMBA-induced transcription elongation of HIV-LTR-Luc, but not transcription initiation (data not shown). Taken together, these observations support the notion that the coordinated BRD4 release and P-TEFb liberation are essential for the recruitment of the BRD4⅐P-TEFb complex and enhancement of transcription elongation.

DISCUSSION
Combined with previous studies (14,31), the data presented here are consistent with a model for the concerted regulation of BRD4 and P-TEFb during stress-induced gene expression (Fig.  8). In the unstressed state, HDAC1/2/3 are inactive, and the majority of BRD4 is associated with chromatin (Fig. 8, top).
Meanwhile, most P-TEFb is sequestrated in the inactive 7SK complex (7SK snRNP, Fig. 8, bottom) (14). Upon stress treatment, HDAC1/2/3 are activated yet still not able to deacetylate nucleosomal H4K5ac/K8ac. PP1␣-mediated dephosphorylation of H3S10ph is required to potentiate the nucleosomes to be deacetylated by HDAC1/2/3 at H4K5ac/K8ac, thereby releasing BRD4 from the nucleosome (Fig. 8, top). Concomitantly, stress-activated PP2B cooperates with PP1␣ to induce the dephosphorylation of CDK9 T-loop at Thr 186 , thereby leading to the liberation of P-TEFb from 7SK snRNP (14) (Fig. 8, bottom). Through selective association with P-TEFb that has its CDK9 re-phosphorylated at Thr 186 by an unknown kinase (31) (Fig. 8, bottom), the released BRD4 recruits the active form of P-TEFb onto the promoter, where it modulates the processivity of RNA Pol II for productive transcription elongation. In this context, the dephosphorylation of H3S10ph controls the role of BRD4 switching from chromosomal targeting to transcription regulation of the expression of inducible genes.
To form a functional BRD4⅐P-TEFb complex, BRD4 and P-TEFb must be available at the same time. Analysis of the release of chromatin-bound BRD4 and the liberation of sequestered P-TEFb using LSF and HSF prepared from stressed cells demonstrated that the two processes occur in parallel (Fig. 7A). How a cell coordinates these processes is intriguing. A recent report showed that overexpression of the PID domain of BRD4 could free P-TEFb from 7SK snRNP (54), suggesting that these processes may be coordinated by BRD4 itself. Our data showed, FIGURE 6. Dephosphorylation of H3S10ph is a prerequisite for stress-induced BRD4 release for inducible gene expression. A, in vitro assay of HDAC1induced BRD4 release in the LSEN containing either wt-or S10A mt-H3. LSEN of HeLa cells with the indicated lentiviral infection were incubated with HDAC1 enzyme, and the BRD4 released into supernatant (sup) and remained in nuclei (nuc) was analyzed by Western blot (WB). B, in vitro assay of HDAC1-induced BRD4 release from nucleosomes containing either wt-or S10A mt-H3. Nucleosome from HeLa cells with the indicated lentiviral infection and treatment was immobilized on anti-FLAG resin, and incubated with HA-HDAC1 enzyme. The levels of BRD4 released in the supernatant (sup) and remained in nucleosome, and the changes in histone modifications, were analyzed by Western blot. C, HMBA-induced BRD4 release in the cells expressing either wt-or S10A mt-H3. LSF prepared from HeLa cells with the indicated lentiviral infection and HMBA treatment was analyzed by Western blot. D and E, S10A mt-H3 allowing cells to bypass the requirement for PP1␣ during HMBA-induced expression of inducible genes. HIV-LTR-Luc cells with the indicated lentiviral infection and HMBA treatment were subjected to luciferase assay (D) and qRT-PCR analysis (E).
however, that when BRD4 release was blocked by either overexpressing H66N mt-PP1␣ (Fig. 7D) or impairing the function of HDAC (Fig. 7F), the stress-induced P-TEFb liberation was unaffected, indicating that BRD4 is not required for P-TEFb liberation. Interestingly, silencing of PP1␣ blocked both BRD4 release and P-TEFb liberation (Fig. 7C), indicating that the PP1␣ pathway serves as the coordinator of these processes. Of note, because H66N mt-PP1␣ inhibited only the release of BRD4, but not the liberation of P-TEFb (Fig. 7D), the functional PP1␣ complexes must be different for the two processes. In mammals, PP1 is estimated to have ϳ650 interacting partners for diverse functions (55). Most likely, H66N mt-PP1␣ is able to associate with the partner (oval A in Fig. 8) that is essential for H3S10ph dephosphorylation and therefore has a dominantnegative effect on this process, but not with the partner (oval B) for P-TEFb liberation. In this way, the same core PP1␣ pathway could coordinate the regulation of BRD4 and the availability of P-TEFb for formation of the BRD4⅐P-TEFb complex.
Formation of the BRD4⅐P-TEFb complex relies on the liberation of P-TEFb and the release of BRD4, both of which are tightly controlled by more than one signaling pathway. For P-TEFb liberation, as shown previously (14), both PP2B and PP1␣ signaling pathways are required, with PP2B acting prior to PP1␣ (Fig. 8, bottom). For the release of BRD4 from nucleosomes, the PP1␣ and HDAC1/2/3 signaling pathways are employed (Fig. 8, top). In this case, PP1␣ must act prior to HDACs (Figs. 5D and 6). Of note, this order of actions is connected by trans-histone cross-talk between H3S10ph and H4K5ac/K8ac, in which PP1␣-induced H3S10ph dephosphorylation enables HDAC-mediated H4K5ac/K8ac deacetylation. This was confirmed by the nucleosomes containing S10A mt-H3 that no longer require PP1␣ for HDAC-mediated FIGURE 7. PP1␣ coordinates BRD4 release and P-TEFb liberation for regulating transcription elongation. A, effect of stress on BRD4 release, P-TEFb liberation, and BRD4⅐P-TEFb recruitment. F1C2(CDK9-f) cells were treated as indicated, followed by stepwise fractionation to yield the LSF and HSF. BRD4 released into LSF (top), BRD4 in association with P-TEFb in the anti-FLAG immunoprecipitates of HSF (middle), and HEXIM1 in association with P-TEFb in the immunoprecipitates from the rest of LSF (bottom) were examined by Western blotting (WB). B, ChIP-qPCR analysis of BRD4 and CDK9/P-TEFb levels at promoter regions of the indicated genes after HMBA treatments of HIV-LTR-luciferase cells. The levels in non-treated cells were set as 1. C and D, effect of silencing PP1␣ (B) or expressing H66N mt-PP1␣ (C) on stress-induced BRD4 release, P-TEFb liberation, and BRD4/P-TEFb recruitment. shScr, scrambled shRNA control. LSF and HSF were prepared as described in the legend to Fig. 3, B and C, and assayed as in A. E, inhibitory effect of H66N mt-PP1␣ on HMBA-induced transcription elongation of the HIV-LTR-luciferase gene. HIV-LTR-Luc cells with the indicated lentiviral infection and HMBA treatment were subjected to qRT-PCR analysis of the levels of initiation and elongation transcripts as illustrated in top panel. Data were averaged and presented as fold-enhancement compared with untreated sample. F, knocking down HDAC1/2/3 blocks BRD4 release, but not P-TEFb liberation. LSF and HSF were prepared from F1C2(CDK9-f) cells with the indicated lentiviral co-infection and treatments, and assayed as in A.
deacetylation of H4K5ac/K8ac (Fig. 6). The molecular mechanism for the cross-talk, however, is unclear. Recently it was reported that HDAC1/2 was able to bind peptides corresponding to the unmodified H3 tail, but not those harboring H3S10ph (56). It is possible that the unmodified H3 peptide had some affinity for HDACs to facilitate the binding of HDACs to nucleosome. Alternatively, H3S10ph may pose steric hindrance in the context of a nucleosome for binding of HDAC.
For more than two decades, the signal-induced phosphorylation of H3S10 has been regarded as a positive epigenetic mark for transcriptional activation of several inducible genes (50,(57)(58)(59)(60)(61). Surprisingly, we found that dephosphorylation of H3S10ph is crucial for BRD4-mediated expression of inducible genes. Although apparently contradictory, these events are likely to occur at different regions of the genome. During heat shock of Drosophila cells, H3S10ph decreased globally, but increased at the promoters of the heat shock genes (62). In mammals, the signal-induced phosphorylation of H3S10 was detected only at some promoters (59), or at an intronic enhancer region (32). Moreover, the signal-induced phosphorylation of H3S10 usually correlates with the chromatin remod-eling of promoter (50), which occurs prior to the transcription initiation, whereas the stress-induced dephosphorylation of H3S10ph enables BRD4 to augment transcription elongation ( Fig. 7E and data not shown), an event that happened after initiation. Hence, one may envision that for a small fraction of nucleosomes located at promoters or enhancer regions, signalinduced H3S10 phosphorylation facilitates chromatin remodeling for transcription initiation, whereas the global decline in H3S10ph may have a distinct function in governing the release of chromatin-bound BRD4 for transcription elongation.
During stress response, the BRD4⅐P-TEFb complex must be recruited onto promoters to augment the expression of inducible genes ( Fig. 7E and data not shown). Because H4K5ac/K8ac is essential for the association of BRD4 with nucleosomes but become deacetylated during stress treatment (Fig. 1, A and B), it is perplexing how the BRD4⅐P-TEFb complex is recruited to target genes. Several studies suggested that recruitment of the BRD4⅐P-TEFb complex to promoters may occur by its association with gene-specific transcription factors. For example, upon TNF-␣ stimulation, BRD4 was found to interact with acetylated NF-B/RelA via its bromodomains, thereby bringing H3S10ph prevents HDAC from deacetylating H4K5ac/K8ac, thereby prohibiting the inappropriate activation of BRD4 (2). Upon stimulation, the external signals (left) trigger the activation of PP1␣, which in association with partner A, dephosphorylates H3S10ph (3). Subsequently, the signal-activated HDAC1/2/3 deacetylate H4K5ac/K8ac, which leads to BRD4 release from chromatin. Bottom: (1) meanwhile, the signal-activated phosphatase PP2B induces a conformational change in the inactive 7SK complex to expose CDK9 T186ph (2). Consequently, PP1␣ with partner B is able to dephosphorylate CDK9 at Thr 186 , thereby liberating P-TEFb from the inactive 7SK complex (14) (3). Through selective association with the transcriptionally active form of P-TEFb that has its CDK9 T-loop re-phosphorylated by unknown kinase(s), (4) the released BRD4 mediates the recruitment of P-TEFb to the promoter-proximal region to enhance the processivity of RNA Pol II for productive elongation.
BRD4/P-TEFb to the promoters of NF-B-dependent genes (37,63). Moreover, P-TEFb could also be recruited by c-Myc to specific promoters (64). Furthermore, the acetylated K16 of nucleosomal histone H4 (H4K16ac) at the FOSL1 intronic enhancer was shown to play a role in tethering the BRD4⅐P-TEFb complex onto its promoter region after stimulating 293 cells with 50% serum (32). It is possible that different genes or regulatory elements could use diverse means for recruiting the BRD4⅐P-TEFb complex to promoter regions.
In summary, this study revealed the intricate network in which signal-induced histone cross-talk revises H3S10ph and H4K5ac/K8ac modifications, and thereby leads to BRD4 activation for augmenting inducible gene expression. The mechanism for the activation of PP1␣ and HDACs during stress response, as well as the molecular mechanism for the blocking effect of H3S10ph on the function of HDAC, await future investigation.