BRD4 Phosphorylation Regulates HPV E2-Mediated Viral Transcription, Origin Replication and Cellular MMP-9 Expression

SUMMARY

Post-translational modification can modulate protein conformation and alter binding partner recruitment within promoter regulatory regions. Here, we find that bromodomain-containing protein 4 (BRD4), a transcription cofactor and chromatin regulator, uses a phosphorylation-induced switch mechanism to recruit E2 protein encoded by cancer-associated human papillomavirus (HPV) to viral early gene and cellular matrix metalloproteinase-9 (MMP-9) promoters. Enhanced MMP-9 expression, induced upon keratinocyte differentiation, occurs via BRD4-dependent recruitment of active AP-1 and NFκB to their target sequences. This is triggered by replacement of AP-1 family members JunB and JunD by c-Jun and also by relocalization of NFκB from the cytoplasm to the nucleus. In addition, BRD4 phosphorylation is also critical for E2- and origin-dependent HPV DNA replication. A class of phospho-BRD4-targeting compounds, distinct from the BET bromodomain inhibitors, effectively blocks BRD4 phosphorylation-specific functions in transcription and cofactor recruitment.

Graphical Abstract

INTRODUCTION

Bromodomain and extraterminal (BET) family proteins are epigenetic readers that bind to select acetylated lysine residues in histone and nonhistone proteins to facilitate the action of DNA-binding factors and associated co-regulators in modulating transcription (Wu et al., 2013; Roe et al., 2015). In humans there are four BET family proteins, including widely-expressed bromodomain-containing protein 4 (BRD4), BRD3, BRD2, and testes/germ cell-specific BRDT (Wu and Chiang, 2007). In addition, three alternative splicing-generated protein isoforms and a chromosomal translocation-created fusion protein were described for BRD4 (Floyd et al., 2013; Chiang, 2014). These distinct BET proteins regulate transcription at the template commitment, initiation, or elongation steps through recruitment of gene-specific transcription factors, the Mediator complex, or the positive transcription elongation factor b (P-TEFb) to RNA polymerase II (Pol II)-driven promoters (Chiang, 2009).

Recent identification of BRD4 as an anti-inflammatory and cancer therapeutic target has been facilitated by compounds that mimic acetyl-lysine such as JQ1 (Filippakopoulos et al., 2010), I-BET (Nicodeme et al., 2010) and MS417 (Zhang et al., 2012). These compounds specifically target BET bromodomains and have stimulated broad interest in understanding BRD4-regulated cancer-specific and drug-resistant pathways (Fong et al., 2015; Rathert et al., 2015). There is also strong interest in the normal cellular functions of BRD4, including intestinal stem cell renewal and T-cell development (Bolden et al., 2014), memory formation and learning (Korb et al., 2015), and X-chromosome inactivation (Wu et al., 2015). Although significant progress has been made in understanding BRD4-activated pathways, very little is known about BRD4-inhibited transcription. It is unclear whether BRD4 entails the same or distinct domains and partner binding proteins to effect gene activation versus repression.

To further define the molecular mechanisms underlying BRD4-mediated gene activation and repression, we used the human papillomavirus (HPV)-encoded E2 protein. E2 is a sequence-specific transcription/replication factor (Chiang et al., 1992; McBride, 2013) that inhibits activator protein-1 (AP-1)-driven transcription from HPV chromatin, wherein BRD4 is required for repression (Wu et al., 2006). Depending on the sequence context, E2 can activate transcription in a BRD4-dependent manner (Lee and Chiang, 2009).

We now report that the phosphorylation-dependent interaction domain (PDID) of BRD4, previously shown to be crucial for chromatin targeting and gene activation by p53 tumor suppressor protein (Wu et al., 2013), is also important for E2-regulated transcription and chromatin contact. Remarkably, the PDID is recognized only by E2 proteins encoded by the oncogenic HPV-16 and HPV-18 strains that are associated with cervical, head-and-neck, and anal cancers, but not by non-oncogenic low-risk HPV-11 and bovine papillomavirus type 1 (BPV-1). PDID-dependent inhibition of the viral promoter requires E2 binding to its target site, and the same region of BRD4 is also needed for E2-mediated HPV origin replication and E2-enhanced NFκB-induced matrix metalloproteinase-9 (MMP-9) transcription. Our findings indicate that BRD4 phosphorylation is essential for viral and cellular transcription factors to reprogram gene expression pathways implicated in cell fate decisions and differentiation control.

RESULTS

Identification of Two Phospho Switch-Regulated E2-Interacting Domains in BRD4

Besides the C-terminal motif (CTM) of BRD4, which is known to recruit P-TEFb (Bisgrove et al., 2007) and different types of E2 protein (Muller et al., 2012), we found an internal region of mouse Brd4 (aa 280–580) able to interact directly with HPV-11 E2 (Wu et al., 2006). To see if this region represents another universal E2-interacting domain likewise used by human BRD4, we performed a GST pull-down experiment by incubating the bacterially (Bac)-expressed FLAG-tagged human BRD4 (f:BRD4) region (aa 279–579) with immobilized GST-tagged BPV-1 E2 (BE2), HPV-11 E2 (11E2), HPV-16 E2 (16E2), or HPV-18 E2 (18E2; Figure S1A). This internal BRD4 region indeed interacted with different types of E2 (Figure 1A, row a). Further mapping pinpointed this second universal E2-interacting region to a basic residue-enriched interaction domain (BID) spanning aa 524–579 (Figure 1A, rows a–c; and Figure S1B and S1C). Intriguingly, the region preceding the BID (aa 287–530, fragment b), when expressed in and purified from insect Sf9 cells (fragment 5), showed selective interaction only with high-risk 16E2 and 18E2 but not low-risk 11E2 and BE2 (lanes 5–6 vs. 3–4). In contrast, Sf9-BID (fragment 6) and Sf9-CTM (fragment 10) both interacted with every type of E2 (Figure 1A). This aa 287–530 fragment b region was previously defined as a phosphorylation-dependent interaction domain (PDID) for casein kinase II (CK2)-mediated BRD4 association with p53 (Wu et al., 2013). To test the importance of phosphorylation, we treated Sf9-purified FLAG-tagged PDID (f:PDID) with calf intestinal alkaline phosphatase (CIP) and Bac-purified f:PDID with CK2 that respectively abolished (Figure 1B, top panel) or restored (bottom panel) PDID interaction with 16E2 and 18E2. These data show phosphorylation of PDID is crucial for selective BRD4 association with high-risk E2.

Figure 1. BRD4 Uses BID for Common E2 Interaction and Phospho-PDID for Select High-Risk E2 Contact.

(A) Domain mapping of BRD4 for E2 interaction by GST pulldown.

(B) CIP treatment of Sf9-purified phospho-PDID abolishes its interaction with 16E2/18E2, whereas CK2 treatment of Bac-purified PDID confers its interaction with 16E2/18E2.

(C) BID and CTM each interact with different types of E2, and phospho-NPS (pNPS) interacts specifically with high-risk 16E2/18E2.

(D) Substitution of S488, S492, S494, S498 and S499, individually, to alanine reduces respective Sf9-PDID interaction with 16E2, 18E2, p53 and BID, with T500/S503 additionally required for BID interaction.

(E) Model for phosphorylation-regulated E2 contact with PDID and BID.

(F) Deletion of CTM abolishes BRD4 interaction completely with 11E2 but partially with 16E2/18E2 in transfected 293 cells. * indicates a cross-reacting species.

(G) BiFC live-cell imaging in transfected HeLa cells.

(H) Counting of BiFC-positive cell numbers in transfected C-33A.

To determine whether phosphorylation at the N-terminal cluster of CK2 phosphorylation sites (NPS) in PDID confers selective E2 association, we deleted bromodomain II (ΔBD2) from the PDID and used CK2-phosphorylated Bac-PDIDΔBD2 (named pNPS) for the GST pulldown. While GST-CTM and GST-BID interact with every type of E2, as expected, pNPS only interacted with 16E2 and 18E2 (Figure 1C). Alanine substitution of serine (or threonine) at each of the eight CK2 phosphorylation sites in Sf9-PDID showed that S488, S492, S494, S498 and S499 in NPS are important for phospho-PDID interaction with 16E2, 18E2, p53 and BID, with T500 and S503 also required for BID interaction (Figure 1D). Since NPS phosphorylation is necessary for its intra-molecular contact with BID (Wu et al., 2013), accessibility of BID for universal E2 interaction is likely regulated by the phosphorylation status of NPS (Figure 1E). Indeed, a protein fragment containing both PDID and BID with unphosphorylated NPS could interact with every type of E2, whereas the same fragment when phosphorylated showed select contact only with high-risk 16E2 and 18E2 (Figure 1A, fragment a vs. fragment 4). This was further validated in vivo by GST pulldown in transfected human 293 cells (Figure S1D). Phosphorylation at the C-terminal cluster of CK2 phosphorylation sites (CPS) residing in aa 598–785 of BRD4 could not confer E2 association (Figure 1A, fragment 7; Figure S1E). Furthermore, the phospho-mimic 8E mutant of Bac-f:PDID also failed to interact with 16E2, 18E2, p53 and BID (Figure S1F). These results strongly indicate that context-dependent CK2 phosphorylation, not purely charge distribution, is critical for select E2 contact with BRD4.

The existence of a CTM-independent E2-interacting region in BRD4 was examined in vivo by co-IP in 293 cells expressing FLAG-tagged full-length (FL) or CTM-deleted (ΔCTM) BRD4 and GST-tagged 11E2, 16E2, or 18E2. While endogenous p53 interacts equally well with FL and ΔCTM BRD4 in a CTM-independent manner (Wu et al., 2013) and ΔCTM is unable to interact with 11E2 due to masking of the BID in vivo by phospho-PDID, ΔCTM retains a significant level of 16E2 and 18E2 interaction (Figure 1F). Unmasking of BID for 11E2 interaction could be demonstrated in vitro by using CIP-treated ΔCTM protein, with or without additional deletion of other non-E2-interacting domains (Figure S1G, lanes 3 vs. 7). The CTM-dependent 11E2 association and ΔCTM-permissible 16E2 and 18E2 interaction was further illustrated by live cell imaging using a bimolecular fluorescence complementation (BiFC) assay (Wu et al., 2013). Co-expression of Venus N-terminus (VN)-linked E2 and Venus C-terminus (VC)-conjugated BRD4 in HPV-positive HeLa cells (Figure 1G) and HPV-negative C-33A cells (Figure 1H) restored fluorescence from the split Venus fluorophore.

Phospho-PDID Contact Residues in E2 Residing at the C-Terminal End of Its DNA Recognition Helix

To identify the region of E2 contacting the identified BID and phospho-PDID in BRD4, we performed GST pull-down experiments by incubating phosphorylated Sf9-f:PDID, non-phosphorylated Bac-f:PDID (negative control), Bac-f:BID, and Bac-f:CTM (positive control) each with the N-terminal (N), C-terminal (C), or hinge plus C (HC) region of GST-11E2 or -18E2 (Figure S2A). While f:CTM interacts with the E2 N-terminal regulatory domain as previously reported (Lee and Chiang, 2009), f:BID interacts specifically with the C-terminal DNA-binding/dimerization (DBD) domain of 11E2 and 18E2, and only Sf9-f:PDID interacts with 18E2-C (Figure 2A). This bipartite interaction between E2 and BRD4 was further examined using Sf9-purified FLAG-tagged full-length (FL) BRD4 and ΔCTM (Figure 2B). 18E2 N and 11E2 N interacted with FL but not ΔCTM, with or without CIP treatment (lanes 2 and 4). 18E2 C, but not 11E2 C, interacted with both FL and ΔCTM (lanes 3 and 5, 1st and 3rd rows). Upon CIP treat, 11E2 C interacted with both FL and ΔCTM (lane 5). These in vitro interaction assays demonstrate that BRD4 uses the CTM as a common E2-N-interacting domain and phospho-PDID for direct contact only with high-risk E2 (Figure 2C), but not low-risk E2 (Figure 2D). Normally masked BID in FL and ΔCTM BRD4 could be unmasked for universal E2 interaction when NPS is dephosphorylated, highlighting the discovery of a selective and a common E2-interaction domain in internal BRD4 regulated by phosphorylation/dephosphorylation.

Figure 2. Identification of Phospho-PDID Contact Residues in E2 to the C-Terminal End of Its DNA Recognition Helix.

(A) GST pulldown shows BRD4 CTM interacts with 11E2 N and 18E2 N, whereas BID and phospho-PDID both interact with 18E2 C.

(B) CIP treatment releases phospho-PDID-masked BID in FL and ΔCTM for 11E2 C interaction but does not alter BRD4 interaction with 18E2.

(C) Model for phosphorylation-switchable bipartite BRD4 interaction with high-risk E2.

(D) Model for phosphorylation-regulated mono- or bipartite BRD4 interaction with low-risk E2.

(E) Amino acid sequence alignment of 16E2, 18E2 and 11E2 DBDs with important residues for phospho-PDID binding underlined or arrowed in red.

(F) Switch of two amino acids at the C-terminal end of DNA recognition helix between high-risk and low-risk E2 alters their interaction specificity with phospho-PDID of BRD4.

(G) BiFC imaging in HeLa cells showing substitutions of two contact selectivity residues in FL E2 change their binding specificity with FL and ΔCTM BRD4.

(H) 16E2 interaction with phospho-PDID is reduced significantly by mutation at K306 and partially by substitution of K307 and R302 residues.

(I) Model for 18E2 DBD binding to DNA (E2BS) showing surface-exposed R303, R307 and K308 critical for BRD4 interaction.

To unravel the molecular basis for phospho-PDID contact with select E2, we inspected the DBD sequences of 16E2, 18E2 and 11E2, in alignment with other phospho-PDID-interacting domains, such as BID and p53 C-terminal regulatory domain (Wu et al., 2013; Figure S2B). These alignments revealed two conserved basic residues in 16E2 (K306 and K307) and 18E2 (R307 and K308) that are replaced by non-basic residues in 11E2 (N304 and D305; Figure 2E, two red arrows). Mutation of these two basic residues in 16E2 and 18E2 to ND, as found in 11E2, completely abolished phospho-PDID interaction, whereas substituting ND in 11E2 with KK, as found in 16E2, led to 11E2 interaction with phospho-PDID at a level comparable to that seen with wild-type 16E2 and 18E2 (Figure 2F). These two amino acid substitutions in full-length E2, with levels of expression similar to the wild-type proteins (Figure S2C), were also examined for their interaction in vivo with FL and ΔCTM BRD4 by BiFC live cell imaging. 16E2-ND and 18E2-ND mutants showed the expected interaction with FL but not ΔCTM, reflecting a strict CTM-dependent interaction due to the loss of their internal phospho-PDID contact residues (Figure 2G). In contrast, the 11E2-KK mutant acquires a unique interaction with ΔCTM. These data demonstrate that the two basic residues present in high-risk 16E2 and 18E2 are crucial determinants for select contact with BRD4 phospho-PDID.

These two residues are situated in the C-terminal end of the DNA recognition helix in the E2 DBD, which features a dimeric barrel possessing eight anti-parallel β strands and four α helices (Figure S2D) with α1 of each monomer contacting the major groove (Figure S2E) at specific bases and phosphates (Figure S2F). We mutated each of the three conserved basic residues (K299, R302, and R304) in the 16E2 α1 DNA recognition helix (Figure 2E) to leucine (L). We then examined, along with the K306N and K307D selectivity-determining mutants, their ability to interact with phospho-PDID by GST pulldown. Clearly, K306 is more important than K307 for interacting with phospho-PDID. R302, but not the base-contacting K299 and R304 residues, is also crucial for phospho-PDID interaction (Figure 2H). Interestingly, these three phospho-PDID-interacting residues in 16E2 (K306, K307, and R302) are adjacent within a solvent-accessible area on the opposite side of the DNA-contacting surface based on the locations of the corresponding residues (R307, K308, and R303) in the available 18E2 DBD/DNA structure (Figures 2I and S2G; Kim et al., 2000; McBride, 2013).

Phospho-PDID-Targeting Compounds Disrupt Select E2-BRD4 Interaction

Given the importance of phosphorylation in regulating intra-molecular PDID-BID contact that, in turn, regulates inter-molecular BRD4-E2 interaction, we tested if our previously isolated phospho-PDID-targeting peptoid compounds, DC-1 and DC-2 (Cai et al., 2011; Figure S3, A–C), could disrupt phospho-PDID association with 16E2, 18E2, and BID. We found that DC-1 and DC-2 both effectively blocked Sf9-f:PDID interaction with 16E2, 18E2 and BID in a dose-dependent manner. In contrast, a randomized control peptoid (Figure S3D) and the BET bromodomain-specific inhibitor JQ1(+) had no effect on these binary interactions (Figure 3A). The disruption by DC-1 and DC-2 was specific to phospho-PDID-mediated interaction, as CTM-driven association with 16E2 and 18E2 was not affected by the addition of these two compounds (Figure 3B). Consistent with the observation that phospho-PDID bound 16E2 most strongly, followed by BID and then 18E2 (Figure S3E), higher concentrations of DC-1 and DC-2 were needed to disrupt phospho-PDID interaction with 16E2 than with 18E2 (Figure 3A and 3C).

Figure 3. Phospho-PDID-Targeting Compounds Disrupt Select E2-BRD4 Interaction.

(A) Inhibition of phospho-PDID interaction with 16E2, 18E2 and BID by increasing doses of DC-1 and DC-2, but not DC-control peptoid and JQ1(+).

(B) DC-1 and DC-2 cannot disrupt CTM interaction with 16E2/18E2.

(C) The dissociation constant (Kd) and half maximal inhibitory concentration (IC50) of DC-1/DC-2 for phospho-PDID interaction with 16E2, 18E2 and BID.

(D) Titration of DC-1 for disrupting CTM-lacking BRD4-S interaction with 16E2/18E2 or unmasking BID in BRD4-S for 11E2 interaction.

(E) Titration of DC-1 for disrupting CTM-containing FL BRD4 interaction with 16E2/18E2 or unmasking BID in BRD4 for 11E2 interaction.

(F) Comparison of DC-1 and JQ1(+) concentrations needed for blocking BRD4 binding to acetylated (Ac-) chromatin.

(G) Model for low-concentration DC-1 blocking of 18E2 recruitment and high-concentration DC-1-induced conformational change masking BD2 binding to Ac-chromatin.

To examine whether compounds that target phospho-PDID could disrupt E2 interaction with intact BRD4 protein, we analyzed the ability of the more effective DC-1 compound to inhibit the CTM-independent interaction between the Sf9-purified short (S) isoform of BRD4 (aa 1–722; Wu et al., 2013) and 16E2, 18E2, and 11E2. We observed that 20 μM of DC-1 (lane 4) inhibited 18E2 but not 16E2 interaction with BRD4-S (Figure 3D), consistent with the inhibitory dosage seen with phospho-PDID and 18E2 interaction (Figure 3A, lane 4). No inhibition was seen when the DC-1 concentration was greater than 20 μM (lanes 5 and 6). This result is likely due to a switch of E2 interaction with BRD4-S from phospho-PDID to BID, thereby disguising the loss of phospho-PDID and 18E2 contact, as independently revealed by unmasking of the BID upon DC-1-disrupted phospho-PDID/BID interaction now available for 11E2 interaction (Figure 3D, last row).

When inhibition by DC-1 was repeated with the long (full-length) isoform of BRD4 (aa 1–1362), similar patterns of dose-dependent inhibition were observed, except that CTM dependence was clearly seen in 11E2 interaction with BRD4 (Figure 3E, last row, vs. Figure 3D, last row, lanes 1–4). Evidently, CTM-dependent BRD4 interaction with the N-terminal domain of E2 could be strengthened by a second domain interaction between phospho-PDID and the C-terminal domain of high-risk E2 or between DC-1-unmasked BID and the C-terminal domain of E2, particularly 11E2 (Figure 3E, lanes 1–4 vs. 5–7). To investigate the direct involvement of unmasked BID with E2, we removed BID from full-length BRD4 and observed no enhanced E2 interaction with BRD4ΔBID at high DC-1 concentrations (Figure S3F).

Disruption of phospho-PDID and BID interaction also results in an intra-molecular contact switch from NPS-BID to NPS-BD2 (Wu et al., 2013). Therefore, we predicted that DC-1 would block BRD4 binding to acetylated chromatin once BD2, which is capable of binding to acetyl-lysine in nucleosomal histones, is blocked by the unleashed NPS from its BID anchor. Indeed, DC-1, but not the control peptoid, blocked BRD4 binding to acetylated chromatin at 330 and 500 μM (Figure 3F, lanes 1–8), concentrations greater than the 27 μM IC50 measured for DC-1-mediated disruption of phospho-PDID-BID interaction (Figure 3C). This inhibition, however, was ~1000-fold less effective as that achieved by active JQ1(+) relative to its inactive stereoisomer JQ1(−) (Figure 3F). These data indicate that inhibition of BRD4 binding to acetylated chromatin by an indirect mechanism via DC-1-induced NPS masking of BD2 is significantly weaker than direct blocking of bromodomain access to chromatin by JQ1. Higher concentrations of DC-1 are probably needed to block not only BD2-mediated chromatin association, but also BD1-synergized BD2 binding to acetylated chromatin (Wu et al., 2013). Nevertheless, phospho-PDID-targeting compounds provide a unique window of opportunity at low concentrations to block factor (e.g., 18E2) association with phospho-PDID without inhibiting BRD4 binding to its chromatin targets, even though at high concentrations they can also inhibit bromodomain binding to acetylated chromatin (Figure 3G).

Bipartite Contact Crucial for BRD4-Mediated E2 Repression of HPV Early Promoter

To define the role of the three BRD4 E2-interacting domains in E2-regulated promoter function, we analyzed integrated HPV-18 long control region (LCR)-driven luciferase reporter activity in three human cervical cancer-derived C-33A cell lines stably expressing FLAG-HA-tagged 16E2, BE2, or vector alone (Figure 4A; Smith et al., 2010). Expression of either E2 protein significantly suppresses HPV early promoter-containing LCR activity (Figure 4B, solid bars), consistent with E2 acting primarily as a transcriptional repressor inhibiting HPV early promoter activity (Hou et al., 2000; Wu et al., 2006; Smith et al., 2010).

Figure 4. NPS of BRD4 Is Specifically Required for High-Risk But Not Low-Risk E2-Mediated Inhibition of HPV Early Promoter Activity.

(A) Three C-33A-derived HPV-18 LCR-driven luciferase reporter lines expressing 16E2, BE2, or vector alone.

(B) Inhibition of 18LCR-containing HPV early promoter activity in 16E2-/BE2-expressing cells is partially alleviated by JQ1(+), but not JQ1(−), treatment.

(C) Knockdown of BRD4 alleviates 16E2/BE2 repression of HPV early promoter activity in E2-containing reporter lines.

(D) Generation of 15 stable luciferase reporter cell lines expressing ectopic or a domain-specific deletion of BRD4 in three parental reporter E2 lines.

(E) Levels of ectopic BRD4 expression in 15 reporter lines.

(F) Only ectopically expressed FL but not domain-deleted BRD4 rescues 16E2 repression in endogenous BRD4-knockdown reporter lines.

(G) Ectopically expressed FL and ΔNPS rescue BE2 repression in endogenous BRD4-knockdown reporter lines.

(H and I) ChIP assay showing E2 and BRD4 binding to promoter-proximal E2BS in different reporter lines.

Error bars, SD (n = 3–6 in B, C, F and G; n = 3 in H and I).

Alleviation of E2 repression could be partly achieved by adding JQ1(+), but not JQ1(−), to the culture medium (Figures 4B and S4A) or by siRNA knockdown of endogenous BRD4 (Figure 4C). These data further confirm that BRD4 is a cellular corepressor for E2-inhibited HPV transcription (Wu et al., 2006; Smith et al., 2010). Involvement of the three E2-interacting domains in BRD4-mediated E2 repression was then addressed by creating 15 C-33A/18LCR-derived stable cell lines expressing triple FLAG-tagged BRD4 (3f:BRD4) FL, domain-specific deletion of each E2-interacting region (ΔNPS, ΔBID, and ΔCTM), or control vector (−) in each of the three reporter lines concurrently expressing 16E2, BE2, or vector alone (Figure 4D). Expression of exogenous BRD4 derivatives is comparable in these lines (Figure 4E, α-FLAG antibody). The total amount of BRD4 in each line was equivalent to the level of endogenous BRD4 found in vector control cells except in BE2-expressing cells where exogenous BRD4 derivatives, particularly ΔBID, are higher than the endogenous protein (Figure S4B, α-BRD4 N antibody).

Knockdown of endogenous BRD4 in 16E2-expressing cells led to derepression of LCR-driven reporter activity, which was rescued only by ectopic expression of FL BRD4 but not by any of the domain-specific deletion derivatives (Figure 4F). These data show that both NPS and CTM of BRD4 are required for 16E2-mediated repression of HPV early promoter. Since ΔBID is a chromatin binding-defective mutant (Wu et al., 2013), ectopic expression of ΔBID was unable to compensate for the loss of endogenous BRD4 function. In contrast, NPS is dispensable for ectopic BRD4 to confer BE2 repression in endogenous BRD4-knockdown BE2 cells (Figure 4G), providing functional validation of unique NPS contact with high-risk but not low-risk E2.

We also monitored E2 binding by chromatin immunoprecipitation (ChIP) to the promoter-proximal #3 and #4 E2-binding sites (E2BS) in the LCR, which are critical for E2-mediated repression of HPV early promoter (Hou et al., 2000; Wu et al., 2006). These experiments revealed that E2 binding to these two cognate sites accounted for both the ability of ectopic FL BRD4 to rescue E2-mediated repression in endogenous BRD4-knockdown 16E2 cells (Figure 4H, left panel) and to restore E2 repression by ectopic FL and ΔNPS in siBRD4-treated BE2 cells (Figure 4I, left panel). Occupancy of BRD4 derivatives at #3/#4 E2BS is consistent with the described chromatin-binding property of these BRD4 mutants (Wu et al., 2013) in that ΔNPS exhibits enhanced chromatin binding due to deletion of the bromodomain-masking NPS, ΔBID is defective in chromatin binding, and ΔCTM has no effect on BRD4 binding to chromatin (Figure 4H and 4I, right panels). These data highlight the universal BRD4 domain-implicated regulation applicable not only to transcriptional activation as previously described for p53 (Wu et al., 2013), but also to transcriptional repression as illustrated here for HPV E2.

Phosphorylation of BRD4 NPS Is Critical for High-Risk E2-Mediated HPV Promoter Repression and Viral Origin Replication

To demonstrate involvement of phosphorylation at the NPS of BRD4 in E2-inhibited HPV promoter activity, we established another eight C-33A/18LCR-derived stable cell lines expressing different alanine-substituted serine/threonine mutants (a, b, c and 7A) of BRD4. The levels of total BRD4 were comparable in reporter lines concurrently expressing 16E2 or BE2 (Figure S4C, see α-BRD4 N antibody) that showed impaired phosphorylation at specific residues as described (Wu et al., 2013; Figure 5A). None of these mutants were able to substitute for wild-type BRD4 in conferring repression in endogenous BRD4-knockdown 16E2 cells (Figure 5B). In contrast, all phosphorylation-defective BRD4 mutants retained corepressor activity as effective as wild-type BRD4 in endogenous BRD4-knockdown BE2 cells (Figure 5C), confirming that NPS phosphorylation is critical for high-risk but not low-risk E2-mediated repression of HPV promoter activity.

Figure 5. Phosphorylation of BRD4 Is Critical for High-Risk E2-Mediated HPV Promoter Repression and Viral Origin Replication.

(A) Immunoblotting of BRD4 wild type (WT), ΔNPS and alanine-substituted NPS-phosphorylation-site mutants in clusters (a, b and c) or in combination (7A).

(B) Ectopically expressed WT BRD4, but not NPS-phosphorylation-site mutants, rescues 16E2-mediated repression of HPV-18 promoter activity in endogenous BRD4-knockdown reporter lines.

(C) All of the ectopically expressed BRD4 NPS-phosphorylation-site mutants rescue BE2-repressed HPV-18 promoter activity in endogenous BRD4-knockdown reporter lines.

(D) Knockdown of endogenous BRD4 alleviates 16E2-repressed endogenous HPV-18 promoter activity, as shown by enhanced 18E6 RNA (left) and protein (right) levels, in 16E2-expressing HeLa/16E6/16E7 cells.

(E) PP2A activator PTZ (phenothiazine) elevates 18E6 RNA level (left) and reduces BRD4 phosphorylation (right) in 16E2-containing but not in E2-absent (−) cells.

(F) Ectopic expression of WT BRD4 but not 7A mutant promotes E1/E2-dependent ori replication (left) and rescues (right) loss of ori replication in endogenous BRD4-knockdown C-33A cells. HPV-16 E1, Myc-tagged HPV-16 E2 (Myc:E2), and an HPV-16 ori plasmid were co-transfected into C-33A cells with or without BRD4 knockdown. The insert shows comparable expression levels of triple FLAG-tagged (3f:) WT BRD4 and 7A mutant expressed in transfected C-33A cells.

(G) Confocal cell imaging showing ectopic expression of WT BRD4, but not 7A mutant, enhances ori-dependent Myc:E2/BRD4 foci formation in transfected C-33A cells. Cells with >10 Myc:E2/3f:BRD4-costained nuclear foci were quantitated with percentage of positive staining shown in a bar graph. Scale bar = 5 μm.

Error bars, SD (n = 4–6 in B–E, and n = 3 in F–G).

To investigate if the phosphorylation requirement of BRD4 could also be observed in HPV-positive cervical cancer cells, we analyzed endogenous HPV-18 promoter activity upon BRD4 knockdown in HeLa cells that also express exogenous HPV-16 E6 and E7 oncoproteins to bypass stable 16E2 expression-induced cell cycle arrest and senescence (Smith et al., 2010). Knockdown of BRD4 alleviated 16E2-mediated repression of endogenous HPV-18 E6 RNA and protein (Figure 5D). Loss of 16E2 inhibition in BRD4-knockdown cells was rescued by exogenous expression of wild-type BRD4, but not by the 7A phosphorylation-defective mutant (Figure S4D). Alleviation of 16E2-mediated inhibition of HPV-18 promoter activity could also be achieved by adding a pharmacological stimulator (PTZ) of protein phosphatase 2A (PP2A) that dephosphorylates BRD4 in cells (Figure 5E; Shu et al., 2016).

Since E2 is also important for HPV origin of replication (ori) function (Chiang et al., 1992) and BRD4-E2 interaction was shown to be critical for HPV-16 ori replication in C-33A cells (Wang et al., 2013), we wondered whether phosphorylation of BRD4 plays an essential role in HPV-16 ori replication. Using an E1/E2-dependent transient replication assay (Chiang et al., 1992) performed in C-33A cells transfected with HPV-16 E1, Myc-tagged 16E2 (Myc:E2), and an HPV-16 ori-containing plasmid with or without exogenous 3f:BRD4 expression, we found E1/E2-potentiated ori replication could be further enhanced by wild-type BRD4, but not the 7A mutant (Figure 5F, left). In BRD4-knockdown cells, loss of E1/E2-dependent ori replication could be effectively rescued by wild-type BRD4, but not the 7A mutant (Figure 5F, right). Consistent with this finding, cell imaging showed that formation of ori-dependent Myc:E2/3f:BRD4 foci in transfected C-33A cells was significantly enhanced by wild-type, but not the 7A mutant (Figure 5G). These data demonstrate a critical role of BRD4 phosphorylation in both HPV transcription and ori-dependent replication control.

E2-Upregulated MMP-9 Expression in Differentiating Keratinocytes Depends on a Specific BRD4 Isoform, But Not BRD2 and BRD3

Since keratinocytes are the natural host for HPV infection, we used immortalized but non-transformed human Ker-CT keratinocytes to establish stable cell lines expressing low-risk 11E2, high-risk 18E2, or vector alone for analyzing cellular promoters co-regulated by E2 and phospho-BRD4. When differentiation was induced in these keratinocyte lines by addition of serum and calcium to serum-free keratinocyte growth medium, as shown by enhanced expression of Involucrin (Figure 6A), MMP-9 RNA was upregulated weakly by 11E2 and significantly by 18E2 in a differentiation-dependent manner (Figure 6B). In contrast, proliferation-related genes, such as ITGB4, K14, c-Myc, KLF4, RPL32, and RPL13A, were downregulated upon differentiation. 18E2-stimulated MMP-9 expression could be detected in the secreted/conditioned medium, which showed active MMP-9 gelatinase activity (and that of serum-carried MMP-2) in degrading the extracellular matrix component gelatin in an in-gel enzymatic activity assay (Figure 6C) and in stimulating HT1080 human fibrosarcoma cell migration (Figure 6D).

Figure 6. E2-Upregulated MMP-9 Expression in Differentiating Keratinocytes Depends Specifically on BRD4-L, But Not BRD4-S, BRD2 and BRD3.

(A) Expression of 11E2 or 18E2 in Ker-CT-derived stable lines does not change the level of Involucrin upon serum/calcium-induced differentiation. Asterisk indicates a non-specific band detected by α-FLAG antibody.

(B) Expression of differentiation-induced MMP-9, but not Involucrin, is further enhanced by 18E2.

(C) In-gel zymography assay showing gelatinase activity of secreted MMP-9 in 18E2-expressing Ker-CT cells is higher than that secreted from 11E2 and vector cells.

(D) Cell migration assay showing conditioned medium collected from 18E2-expressing Ker-CT cells enhances HT1080 cell migration.

(E) Proliferation assay showing JQ1(+) inhibits cell growth of the three Ker-CT lines with or without E2 expression. UNT, untreated.

(F) Scheme of JQ1 treatment.

(G) JQ1(+) suppresses differentiation-induced MMP-9 RNA level enhanced by E2 expression.

(H) Knockdown of BRD4-L, but not BRD4-S, BRD2 and BRD3, suppresses differentiation-induced MMP-9 RNA level in different Ker-CT-derived lines.

(I) JQ1(+) suppresses differentiation-induced Involucrin RNA level independent of E2 expression.

(J) Knockdown of BRD4-L, BRD2 or BRD3, but not BRD4-S, suppresses differentiation-induced Involucrin RNA level independent of E2 expression.

The involvement of BRD4 in keratinocyte proliferation is not E2-dependent, as growth of these three Ker-CT lines was similarly inhibited by addition of JQ1(+) (Figure 6E). This result is consistent with a lack of significant changes in proliferation-related gene expression (Figure 6B, first three columns). To determine whether E2-upregulated MMP-9 expression in differentiation-induced Ker-CT cells is likewise controlled by BRD4 as had been found in the E2-regulated HPV promoter, we added JQ1 (100 nM) 24 h prior to serum/calcium-induced differentiation with replenished JQ1 (Figure 6F). JQ1(+) significantly inhibited differentiation-induced MMP-9 gene expression in both basal (white bar) and E2-stimulated (grey and black bars) pathways (Figure 6G). Choosing a low JQ1 concentration with prolonged treatment allowed us to follow the extent of keratinocyte differentiation by monitoring protein markers, such as Involucrin, for differentiation, which could take up to 96 h depending on the culturing condition. Treating Ker-CT cells with JQ1 for only 2 h, instead of 24 h, prior to differentiation showed no difference in inhibiting MMP-9 RNA expression (Figure S4E).

To define which BET protein is involved in E2-dependent MMP-9 activation, we performed siRNA knockdown for long (L) form-specific, short (S) form-specific, and both L and S (Pan) of BRD4, as well as BRD2 and BRD3. Only knockdown of BRD4-L, as seen in Pan and L siRNA, suppressed 18E2-dependent activation of MMP-9 (Figure 6H). This phenocopied JQ1-inhibited MMP-9 expression (Figure 6G) and indicated that full-length BRD4, rather than other BET family proteins, was involved in high-risk E2-upregulated MMP-9 gene transcription. In contrast, differentiation-induced Involucrin expression inhibited by JQ1(+) was E2-independent (Figure 6I) and redundantly controlled by BRD4-L, BRD3 and BRD2 (Figure 6J), indicating that differentiation-induced genes could be either uniquely (e.g., MMP-9) or commonly (e.g., Involucrin) regulated by different BET family proteins. Interestingly, only BRD4-L, but not BRD4-S, BRD3 or BRD2, was also uniquely required for E2-inhibited HPV early promoter activity when each BET protein was knocked down by siRNA (Figure S4F). Enhancement of MMP-9 RNA level upon knockdown of BRD2 or BRD3, unlike that seen with Involucrin RNA, indicates that BET proteins have regulatory properties that are gene-specific and context-dependent.

MMP-9 Promoter Activity Is Upregulated by BRD4 Phospho-NPS-Potentiated E2 Function and NFκB Recruitment, Along with A Switch of AP-1 Family Members

To define the molecular mechanism of BRD4 and E2 in MMP-9 gene activation and the role of phospho-NPS in high-risk E2-specific regulation, we analyzed endogenous MMP-9 RNA level in C-33A cells co-transfected with expression plasmids for different types of E2 and either 3f:BRD4 FL or ΔNPS. While ectopic FL and ΔNPS 3f:BRD4 stimulated BE2- and 11E2-enhanced MMP-9 expression equally well, only FL and not ΔNPS potentiated 16E2- and 18E2-upregulated MMP-9 RNA level (Figure 7A). These data indicate that NPS in BRD4 is essential for high-risk E2-enhanced MMP-9 gene transcription. When 7A mutant which shows comparable E2 interaction as wild-type BRD4 (Figure S5A) was used, potentiation of 16E2, 18E2 and 11E2-KK mutant that bestows 11E2 the ability to interact with phospho-NPS was no longer observed (Figure S5B), substantiating the importance of NPS phosphorylation residues in BRD4-potentiated E2 activation.

Figure 7. MMP-9 Promoter Activity Is Upregulated by BRD4 Phospho-NPS-Potentiated E2 Function and NFκB Recruitment, Together with Switch of AP-1 Family Members.

(A) ΔNPS in BRD4 is essential for MMP-9 gene transcription upregulated by high-risk 16E2 and 18E2 but not low-risk 11E2 and BE2 in transfected C-33A cells. Both 18E2 and 11E2 used here are codon-optimized.

(B) Phospho-NPS-targeting DC-1 but not its control peptoid abolishes 18E2-potentiated MMP-9 RNA level in differentiating Ker-CT cells.

(C) CK2 inhibitor TBB abolishes 18E2-potentiated MMP-9 RNA level in differentiating Ker-CT-derived cells by reducing BRD4 phosphorylation.

(D) ChIP assay monitoring E2 and BRD4 occupancy of the MMP-9 promoter-proximal region in proliferating and differentiating Ker-CT-derived keratinocytes.

(E and F) ChIP assays monitoring E2 binding to regions 2 (E2BS) and 4 (proximal AP-1 site) and BRD4 binding to regions 1–6 of the MMP-9 promoter-proximal region in DC-1- or control peptoid-treated Ker-CT-derived keratinocytes expressing 11E2 or 18E2 in proliferation or differentiation states.

(G) ChIP assay monitoring JunB, JunD, c-Jun and p65 binding to regions 1–4 of the MMP-9 promoter-proximal region in proliferating and differentiating Ker-CT-derived keratinocytes.

(H) ChIP assay monitoring p65 binding to regions 3 of the MMP-9 promoter-proximal region in DC-1- or control peptoid-treated Ker-CT-derived keratinocytes in the differentiation-induced state.

(I) MMP-9 RNA level in different Ker-CT-derived cells untreated or treated with BMS-345541 or an AKT inhibitor.

(J) Model for combinatorial MMP-9 promoter regulation by BRD4-dependent E2, AP-1, and NFκB recruitment in proliferation and differentiation states.

Error bars, SD (n = 3 in A–I).

The requirement of NPS for endogenous BRD4-potentiated high-risk E2 activation was then examined in three stable Ker-CT-derived lines treated with phospho-NPS-targeting DC-1 or control peptoid. Indeed, enhanced MMP-9 expression by 18E2 in differentiating keratinocytes was reduced by DC-1 (Figure 7B) without affecting E2-independent Involucrin RNA expression (Figure S5C). This was independently confirmed by treating differentiating cells with the CK2 inhibitor, TBB, which abolished 18E2 activation upon reduction of BRD4 phosphorylation (Figure 7C). Thus, phospho-NPS of BRD4 is specifically required for high-risk E2-potentiated MMP-9 activation.

Since an E2BS situated in the MMP-9 promoter-proximal region also containing one NFκB and three AP-1 sites (Figure 7D, top drawing) is known to regulate MMP-9 expression (Akgül et al., 2011), we performed ChIP in these three Ker-CT lines to analyze whether BRD4 is implicated in the recruitment of E2, NFκB, and AP-1 to their cognate sites in regulating MMP-9 gene transcription. In proliferating keratinocytes, binding of E2 was detected at its cognate site in region 2, but also in E2BS-devoid regions 1, 3 and 4 (Figure 7D, middle). In differentiating keratinocytes, E2 binding was confined to regions 2 and 4 with some 18E2 occupancy additionally found in region 3. This binding profile suggests that E2-regulated MMP-9 expression involves both direct binding to its cognate site and indirect association with AP-1 and NFκB. Differentiation-induced factor switches of AP-1 and NFκB relocalization (see below) likely elicit changes of E2 recruitment to non-E2BS regions. For comparison, binding of BRD4 in proliferating keratinocytes, initially limited to regions 1, 3 and 4 in the absence of E2, was further detected at region 2 in the presence of 11E2. Strikingly, BRD4 occupancy was dramatically enhanced by 18E2 in all of the upstream regions, in contrast to the gradually diminishing binding pattern in differentiating keratinocytes starting at region 3 in the absence of E2 or region 2 in the presence of E2 and extending to the downstream coding region when MMP-9 gene transcription was induced by differentiation (Figure 7D, bottom).

To examine whether differential binding of E2 and BRD4 in proliferating and differentiating states was attributed to phospho-NPS-specific recruitment of high-risk E2, we added DC-1 and found that enhanced recruitment of 18E2 to its cognate E2BS in region 2 in proliferating but not differentiating keratinocytes was reduced by DC-1 to the same level occupied by 11E2 (Figure 7E, left). E2 recruitment to the proximal AP-1 site in region 4 in both cell states was unaffected by DC-1 (Figure 7E, right), suggesting that phospho-NPS of BRD4 is critical for enhancing the additional level of 18E2 (versus 11E2) occupancy to its target DNA sequence, but nonessential for indirect recruitment of E2 to other transcription factor-binding sites. Likewise, the extra level of 18E2-enhanced BRD4 recruitment was reduced by DC-1, which does not affect 11E2-recruited BRD4, in differentiating keratinocytes (Figure 7F) and proliferating cells (Figure S5D). Addition of JQ1(+), but not JQ1(−), completely abolished BRD4 and also E2 binding to all of the MMP-9 promoter-proximal and coding regions irrespective of the cell status (Figure S5E and S5F). Our data suggest a general role of BRD4 in recruiting/stabilizing E2 binding to the MMP-9 gene locus and a unique role of phospho-NPS in potentiating BRD4-mediated high-risk E2 function.

When recruitment of AP-1 and NFκB to MMP-9 was analyzed respectively by promoter occupancy of Jun family proteins and p65, we found that robust binding of JunB and JunD to AP-1 sites in regions 1, 3 and 4 in proliferating keratinocytes was significantly reduced upon differentiation in control and 11E2-expressing cells (Figure 7G). Binding was completely abolished in keratinocytes expressing 18E2. In contrast, occupancy of c-Jun and p65, both undetectable in proliferating cells, was dramatically increased for c-Jun at regions 3 and 4 and for p65 at region 3. Thus, JunB and JunD likely function as repressors or support only a low level of MMP-9 expression in the proliferating state and upon keratinocyte differentiation are then replaced by c-Jun activator acting specifically through promoter-proximal AP-1 sites centering at -76 and -533, correlating with differentiation-enhanced MMP-9 RNA level (Figure 6B and 6G).

Differentiation-induced p65 binding to the NFκB site in region 3 was further stimulated by 18E2, but not 11E2, which could be abolished by DC-1 treatment (Figure 7H) without altering p65 protein level (Figure S6A). Since c-Jun binding to AP-1 sites in regions 3 and 4 in the differentiation state was unaltered by E2 (see Figure 7G), it is not surprising that DC-1 did not affect c-Jun binding to these two cognate AP-1 sites in MMP-9 (Figure S6B), confirming the specificity and selectivity of DC-1 in targeting phospho-NPS-dependent function of BRD4.

Combinatorial Regulation of MMP-9 Promoter Activity by BRD4-Regulated AP-1 and NFκB Recruitment and 18E2-Enhanced NFκB Activity

To examine if differentiation-induced p65 binding to its cognate NFκB site in MMP-9 is triggered by nuclear translocation, we performed immunofluorescence in the three Ker-CT lines in proliferating or differentiating states. Indeed, p65 was found in the cytoplasm of all three proliferating cell lines and rapidly moved to the nucleus upon differentiation (Figure S6C). This movement was prevented by BMS-345541, a compound that blocks IKK-mediated phosphorylation and degradation of IκB (Burke et al., 2003) without affecting p65 protein level and the phosphorylation and protein level of BRD4 (Figure S6D). In contrast, differentiation-induced c-Jun binding to MMP-9 is caused primarily by phosphorylation of S73 and S63 in c-Jun and enhanced protein stability (Loesch et al., 2010), which also occurs in an E2-independent manner (Figure S6E) and is unaltered by BMS-345541 treatment (Figure S6D, second row). This finding highlights two distinct mechanisms for converting latent transcription factors to their active forms in regulating MMP-9 promoter activity. Although the effect of E2 is not on p65 translocation and c-Jun phosphorylation, 18E2 clearly potentiates NFκB-upregulated MMP-9 promoter activity as treatment of differentiating keratinocytes with BMS-345541, but not an AKT inhibitor, significantly reduced 18E2-stimulated MMP-9 RNA level to that sustained by 11E2 (Figure 7I).

BRD4 has no effect on differentiation-induced c-Jun activation by phosphorylation (Figure S7A) and p65 activation by nuclear translocation (Figure S7B). However, we do observe that BRD4 is essential for stable promoter-proximal occupancy of p65 and Jun family proteins to MMP-9 as treatment of JQ1(+) significantly reduced binding of these transcription factors to their cognate DNA sites (Figure S7C), regardless of cells in a transcription-poised or -activated state.

The requirement of BRD4 for AP-1 binding to the promoter region is mediated by direct interaction between BRD4 and each component of AP-1 family proteins, including c-Jun, JunB, c-Fos, FosB, Fra-1 and Fra-2, except JunD (Figure S7D). When dimeric AP-1 complexes were formed between Jun and Fos family members (Wang et al., 2011), all analyzed recombinant AP-1 dimers, including JunD-containing complexes, interact efficiently with purified BRD4 protein (Figure S7E). Most AP-1 members, except FosB, did not interact individually with CK2-phosphorylated NPS, distinct from the p50 and p65 members of NFκB that could interact independently with phospho-NPS (Figure S7F).

Together with the ability of BRD4 to bind directly to p65 (Huang et al., 2009) and to different types of E2 proteins (Muller et al., 2012), it becomes clear that BRD4, being an interacting core, facilitates viral E2 and cellular AP-1 and NFκB proteins to combinatorially regulate MMP-9 gene transcription via direct DNA contact, as well as indirect protein anchoring in both proliferating and differentiating keratinocytes (Figure 7J). The switch from phospho-NPS-dependent 18E2 recruitment in proliferating cells to 18E2- and phospho-NPS-dependent NFkB recruitment in differentiating cells highlights the elegance of phospho-BRD4-mediated protein partner switch and transcription reprogramming from initial assembly of poised factors to formation of a transcriptionally active enhanceosome complex occurring during cell identity shift.

DISCUSSION

BRD4 is a promising therapeutic target for a wide range of hematopoietic cancers and solid tumors. Several small molecules can block BRD4 binding to key regulatory regions controlling cancer-driver gene expression. These BET bromodomain-targeting compounds, although effective and highly selective for BET family proteins, nevertheless pose potential issues for long-term treatments that inevitably elicit side effects owing to inhibition of normal gene function critical for cell growth and differentiation.

In this study, we unravel a crucial role of BRD4 in modulating HPV E2-regulated viral gene transcription and cellular MMP-9 expression. Our findings reveal that phosphorylation triggers a switch within BRD4 that controls recruitment and functional selectivity between E2 proteins encoded by cancer-associated high-risk HPVs and non-cancer-related low-risk HPVs. This phospho-BRD4-modulated activity also supports HPV ori-dependent replication activity, potentiates high-risk E2-regulated MMP-9 expression, and regulates combinatorial E2-NFκB action in controlling MMP-9 gene transcription in differentiating keratinocytes. Compounds targeting this phospho-BRD4 region, such as DC-1, prove effective and highly selective in inhibiting high-risk E2-specific function without affecting general activity common to every type of E2, providing a proof-of-concept for future development of non-bromodomain-based BET inhibitors that are more selective in target gene inhibition.

In mapping E2-interacting domains in BRD4, we identified two internal regions (PDID and BID) that exhibit target selectivity in E2 interaction. PDID spans the second bromodomain (BD2) and a cluster of CK2 phosphorylation sites (NPS). These CK2 sites had been shown previously to control the “open” or “closed” configuration of BD2 in binding to acetylated chromatin and also masking or unmasking of the positively charged BID for interacting with p53 (Wu et al., 2013) and every type of E2 as reported here. It is important to note that protein-protein contact may not necessarily lead to active transcription, as only phospho-PDID-recruited p53 is active in target gene regulation whereas BID-associated p53 is inactive in DNA binding (Wu et al., 2013). This is similarly reflected in “poised” binding of 11E2 and 18E2 to their direct E2 target site in region 2 and in expanded recruitment, via BRD4 association with AP-1 family members, to the AP-1 site-containing regions 1, 3 and 4 of MMP-9 gene in proliferating keratinocytes (Figure 7D), which does not actually contribute to the transcriptional readout of MMP-9 (Figure 6B and 6G). Robust transcription occurs upon cytokine-induced differentiation when active NFκB becomes available in the nucleus for binding to its cognate site in region 3, potentiated by high-risk 18E2, and accompanied by dissociation of JunB/JunD binding to regions 1, 3 and 4 and re-association of c-Jun binding to regions 3 and 4 (Figure 7G). The dynamics of factor switches and reorganization of viral and cellular transcription factors in regulating MMP-9 gene transcription during conversion from proliferating to differentiation states elegantly illustrate the intricacy of combinatorial regulation of eukaryotic transcription coordinated by the epigenetic regulator BRD4 via a phosphorylation-controlled mechanism that likely extends to several hundreds of other DC-1-regulated differentiation-specific genes identified by exome sequencing (S.-Y. Wu, Y.J. Kim, T.H. Kim, and C.-M. Chiang, unpublished data) as well as to learning and memory (Korb et al., 2015) and cancer cell progression seen in triple negative breast cancer (Shu et al., 2016).

The existence of a second universal E2-interacting domain (BID) in BRD4 typically masked in vivo by phosphorylation of an adjacent regulatory domain (PDID), together with the previously characterized CTM that is surface-exposed and uniquely present in BRD4 among the ubiquitously expressed BET family proteins, highlights a specific requirement of full-length (long-form) BRD4 (but not CTM-lacking BRD2, BRD3 and BRD4-S) for E2-regulated HPV and cellular gene transcription. It further highlights a two-arm chelating mechanism unleashed upon dephosphorylation or factor-induced conformational change for high-risk E2 to switch contact with BRD4 from phospho-PDID to BID and for low-risk E2 to capture a newly exposed surface. This surface engages in a distinct type of functional regulation presumably via context-dependent partner switch. Our findings indicate that phosphorylation and combinatorial regulation play a key role in the control of cellular and viral transcription by BRD4.

EXPERIMENTAL PROCEDURES

Plasmids and Protein Purification

Procedures for constructing protein expression plasmids (Table S1) and protein purification are described (Thomas and Chiang, 2005; Wu and Chiang, 2009; Wang et al., 2011) and detailed in Supplemental Information (SI).

Peptoid Synthesis

Synthesis of DC-1, DC-2, and control peptoids is detailed in SI.

Cell Lines and Lentivirus

Stable lines derived from HeLa/16E6/16E7 and C-33A/18LCR (Smith et al., 2010) for expressing 16E2, BE2 or vector alone, and from Ker-CT (Ramirez et al., 2003) for expressing 11E2, 18E2 or vector alone were generated as described in SI, along with details for medium, transfection, and production of E2 lentivirus.

GST Pulldown and Co-Immunoprecipitation (IP)

Protein-protein interactions conducted with GST pulldown or co-IP and analyzed by immunoblotting are described in SI.

Bimolecular Fluorescence Complementation (BiFC) Assay

Detection of E2-BRD4 interaction in vivo by BiFC live-cell imaging (Wu et al., 2013) with HeLa and C-33A cells is described in SI.

Kinase/Chemical Inhibition and Phosphorylation Altering Assay

CIP and CK2 treatment of purified BRD4 proteins was described (Wu et al., 2013) with compound addition and determination of Kd and IC50 for BRD4 binding to E2 or acetylated chromatin described in SI.

Kinase Inhibitor Assay

Pharmacological inhibitors against CK2 (TBB), PP2A (PTZ), NFκB (BMS-345541) and AKT (1L6-hydroxymethyl-chiro-inositol-2-(R)-2-O-methyl-3-O-octadecyl-sn-glycerocarbonate) were added respectively to cells and analyzed for localization of p65 by immunofluorescence staining, the phosphorylation status of c-Jun and BRD4 by immunoblotting, and MMP-9 RNA levels by qRT-PCR as described in SI.

Transient Replication and Replication Foci Assay

Transient HPV replication assay was performed in C-33A cells by transfection with HPV-16 E1, HPV-16 E2, and HPV-16 ori-containing plasmids, together with wild-type BRD4, 7A, or vector construct, and analyzed by qPCR. Confocal imaging of E2-BRD4 foci in similarly transfected C-33A cells is also detailed in SI.

Cell Proliferation and Transwell Migration Assay

Proliferation of Ker-CT-derived cells analyzed with Calcein AM fluorescent dye and migration of HT1080 cells monitored by Transwell with subsequent Calcein AM labeling are detailed in SI.

Gel Zymography

Analysis of secreted MMP-9 gelatinase activity in Ker-CT-derived cells was conducted by in-gel zymography following SDS-PAGE (see SI).

Chromatin Immunoprecipitation (ChIP), qRT-PCR and siRNA Knockdown

ChIP assay, qRT-PCR, and siRNA knockdown were performed as described (Wu et al., 2013) using primer sequences for specific gene loci (Table S2) and for BET siRNA knockdown (Table S3), with sources of antibodies (Table S4) and procedures detailed in SI.

Structure Analysis and Sequence Alignment of E2 DNA-Binding Domains

Structures of 16E2 DBD (1ZZF, RCSB PDB) and 18E2 DBD in complex with DNA (1JJ4; Kim et al., 2000) were analyzed by PyMOL as detailed in SI.

In Brief.

BET bromodomain inhibitors effectively reverse cancer phenotypes but also alter normal cellular activity. Wu et al. describe a phosphorylated region of BRD4 that is critical for HPV origin replication, and interacts with the HPV E2 protein. Compounds targeting phospho-BRD4 block E2-regulated viral and cellular gene transcription.

Highlights.

• BRD4 phospho-NPS interacts selectively with high-risk HPV E2 protein

• Phospho-NPS-targeting compounds block phosphorylation-dependent BRD4 function

• Phosphorylated NPS residues are critical for BRD4-mediated HPV origin replication

• BRD4 regulates MMP-9 transcription jointly with NFkB and select AP-1 members