BRD4 binds the nucleosome via both histone and DNA interactions

Summary

BRD4, a bromodomain and extraterminal (BET) family transcriptional regulator, is believed to be recruited to chromatin via interactions between its tandem bromodomains (BD1 and BD2) and acetylated histone tails. Although extensive studies have explained how individual BRD4 bromodomains bind to acetylated peptides and how BET inhibitors interfere with such interactions, equivalent studies of full-length BRD4 protein with the nucleosome were lacking. Our cryo-EM structure of the BRD4 short (BRD4-S) isoform bound to a nucleosome diacetylated on histone H4 shows how BRD4 BD1 engages both the H4 tail and nucleosomal DNA. Unlike other chromatin reader domain/nucleosome structures, BRD4 BD1 presents the acetylated histone tail for potential interaction with additional chromatin proteins. Unexpectedly, our biochemical studies indicate that BRD4 uses basic regions outside of the bromodomains to bind nucleosomes tightly even in the absence of histone acetylation. Our results further show that histone H4 acetylation influences the conformation of the BRD4/nucleosome complex.

eTOC Blurp

Zhu et al characterize how the transcriptional regulator, BRD4, interacts with acetylated nucleosomes. Their cryo-EM structure shows the BRD4 first bromodomain binds both the acetylated histone H4 tail and nucleosome DNA. Unexpectedly, BRD4 does not need histone acetylation to bind nucleosomes with high affinity in vitro.

Graphical Abstract

Introduction

BRD4, a member of the bromodomain and extraterminal (BET) family of gene regulatory proteins, has fundamental roles in transcriptional regulation, DNA replication, DNA repair, cell cycle progression, cell differentiation and cancer development^1–6. Like other BET family members BRD2, BRD3 and BRDT, it contains tandem bromodomains (BD1 and BD2), which bind acetylated lysine residues of target proteins such as histones^7–11, and the extraterminal (ET) domain which recruits other chromatin effector proteins^12–15. BRD4 is usually thought to recruit the positive transcriptional elongation factor b (P-TEFb) kinase complex which enables release of RNA polymerase II (Pol II) from a transcriptional paused state to a transcriptional elongation state^16–18. However, recent data indicate that BRD4 may not be required for P-TEFb localization^19–22 and that BRD4 may function instead to assemble functional Pol II elongation complexes^19,21,23.

Three isoforms of BRD4, two short (S) and one long (L), are present in human cells: BRD4-S(a), BRD4-S(b) and BRD4-L^24,25. All three contain the same N-terminal 719 residues including the BD1, BD2 and ET domains. BRD4-S(a) (hereafter abbreviated as BRD4-S) and BRD4-L exhibit opposing functions in breast cancer, with BRD4-S acting as an oncogene and BRD4-L as a tumor suppressor²⁵. All three BRD4 isoforms are believed to be targeted to acetylated chromatin through interactions between BD1/BD2 and acetylated histones^1,25,26, but BRD4-S has been found to bind more tightly to acetylated histone H4 than BRD4-L²⁷. Extensive structural and biochemical studies have characterized how bromodomains use a hydrophobic cavity to bind acetyl-lysine peptides and how BET inhibitors such as JQ1 compete for binding to this hydrophobic cavity^28–35. BRD4’s role in oncogenesis and its overexpression in cancer cells make it an anti-cancer target, and JQ1 and other BET inhibitors have been shown to slow tumor growth^2,31,36,37.

The vast majority of the structural and biochemical studies between BRD4 and chromatin have focused on the interactions between individual BRD4 bromodomains and histone peptides. To address the question of how the BRD4 protein interacts with acetylated histones in the context of the nucleosome and not just histone peptides, we have performed structural studies of the BRD4-S protein in complex with acetylated nucleosomes. Our cryo-EM structure of the BRD4/nucleosome complex shows how BRD4 BD1 interacts with not only the histone H4 N-terminal tail but also with nucleosomal DNA. Unexpectedly, we find that the BRD4 protein does not need histone acetylation or even the H4 tail to bind with high affinity to nucleosomes. Instead, we find that histone H4 acetylation influences the conformation of the BRD4/nucleosome complex.

Results

BRD4/nucleosome structure

We reconstituted BRD4-S with nucleosomes containing acetylated histones at a variety of positions on histone H3 and/or H4. We then analyzed these BRD4/nucleosome complexes using cryoelectron microscopy. For most of the BRD4/nucleosome complexes we examined, we were able to observe extra density on the histone face of the nucleosome in the resulting image reconstructions, but that density was diffuse and difficult to interpret. The exception was the BRD4/nucleosome complex containing H4 K12acK16ac and H3 K18ac for which tubular density corresponding to the bromodomain four-helix bundle was clearly visible (Figure 1a). The nominal resolution of the BRD4/nucleosome cryo-EM reconstruction using 80,700 particles was 2.9 Å (Figure S1, S2), but this reflects the much better resolution for the nucleosome component of the complex compared to the local resolution of about 4.5 Å for the BRD4 bromodomain (Figure S1). Such a difference in local resolution is often observed in nucleosome complexes presumably due to motion of the chromatin factor with respect to the nucleosome^38–41.

Figure 1: Overview of BRD4/H4 K12acK16ac nucleosome structure.

(a) Cryo-EM map and (b) cartoon representation of the BRD4-S/H4 K12acK16ac nucleosome complex showing how the BRD4 BD1 (pink) interacts with the nucleosome. The modeled histone H4 tail residues 11–16 are shown as a thicker gold line. The locations of the BRD4-S structured domains are shown below together with the color key for the figures.

Our structural model contains the nucleosome and the crystal structure of BRD4 BD1 built into the extranucleosomal tubular density corresponding to the bromodomain α-helices. (Figure 1b). We believe the extranucleosomal tubular density corresponds to BRD4 BD1 because the structural model positions the bromodomain’s peptide binding pocket to interact with histone H4 K12acK16ac N-terminal tail, one of the preferred peptides for BRD4 BD1⁴². We used nucleosomes also containing H3 K18ac in the hope that BRD4 BD2 would engage that modified histone tail. However, we only observe density for one bromodomain in our reconstruction. This suggests the rest of the BRD4-S protein, which includes BD2 and the ET domain, is conformationally flexible on the nucleosome. The structure of the nucleosome itself is essentially unchanged from other cryo-EM structures of the nucleosome with similar weak density for the DNA ends.

In our structural model, BRD4 BD1 is located on the histone face of the nucleosome above the histone H4 N-terminal tail. This allows the histone H4 tail to enter the BD1 peptide binding pocket, and in our cryo-EM map we observe density for the H4 tail approaching the peptide binding pocket (Figure S3a, b). In addition, BD1 is positioned to interact with the phosphate backbone across the minor groove of nucleosomal DNA around superhelical position (SHL) +1.5 via helix αZ residues R68, K72 and K76 (Figure S3b). At the local resolution of 4.5 Å, we do not observe density for BRD4 BD1 side chains, but we can deduce the location of side chains based on the BRD4 BD1 crystal structure built into the tubular density.

BRD4 does not require acetylated histones to bind with high affinity to nucleosomes To further understand how the BRD4 protein interacts with the nucleosome, we performed binding studies using insights from our structural studies. To quantitatively measure how BRD4-S interacts with nucleosomes, we employed the time-resolved fluorescence resonance energy transfer (TR-FRET) nucleosome binding assay recently developed by the McGinty laboratory⁴³. We initially performed these experiments at a NaCl concentration of 70 mM. Since the BRD4 bromodomains specifically bind to acetylated histone tails, we expected that BRD4-S would bind tighter to acetylated nucleosomes compared to unmodified nucleosomes. Our results show that BRD4-S binds tightly (K_d = 8.0 nM) to H4 K12acK16ac nucleosomes (Figure 2a, Figure S4a). However, we find that BRD4-S binds with similar or even slightly higher affinity to unmodified nucleosomes (6.6 nM) as it does to the acetylated nucleosomes. Even more unexpected is the result that BRD4-S binds equally well to H4 tailless nucleosomes, which lack the H4 residues thought to bind to BRD4, as it does to unmodified nucleosomes (Figure 2a). This indicates that BRD4-S’s affinity to the nucleosome is dictated by more than its bromodomains’ interactions with acetylated histone tails. Consistent with this conclusion is our finding that binding of individual BRD4 BD1 and BD2 bromodomains to H4 K12acK16ac nucleosomes could not be detected in our assay (Figure S4a).

Figure 2: BRD4-S binds to nucleosomes with nanomolar affinity at 70 mM NaCl independent of histone acetylation.

(a) Time-resolved FRET binding assay results for BRD4-S binding to unmodified (blue), H4 tailless = H4(24–102) (green) and H4 K12acK16ac nucleosomes (pink) in 70 mM NaCl. (b) Effect of NaCl concentration on BRD4-S binding to unmodified (blue) or H4 K12acK16ac nucleosomes (pink) as assayed by TR-FRET. P-values comparing the K_d for the acetylated vs unmodified nucleosomes indicated as follows: ns for not significant, * for p-value < 0.05, * for p-value < 0.01.

Role of BRD4 basic patches in high affinity binding to nucleosomes

Given our findings that BRD4-S binds tightly to both unmodified and H4 K12acK16ac nucleosomes and that individual BRD4 bromodomains are not sufficient to bind H4 K12acK16ac nucleosomes, we asked which regions of BRD4 mediated binding to nucleosomes. When we examined the BRD4-S protein sequence, we noticed five patches enriched with Lys and Arg basic residues (Figure 3a). The first three basic patches are contained within the 180 amino acid region between the tandem BRD4 bromodomains previously suggested to interact with nucleosomal DNA⁴⁴. Basic patch 1 (residues 177–184) follows almost immediately after BRD4 BD1, while basic patches 2 (residues 283–290) and 3 (residues 314–333) are positioned between basic patch 1 and BD2. Basic patch 2 is contained within the conserved BET family motif A⁴⁵. Basic patches 4 (residues 535–554) and 5 (residues 561–575) follow the phosphorylation-rich region after BD2 and are contained with the previously identified BID basic interaction domain⁴⁶. Since a major feature of the nucleosome is the negatively charged nucleosomal DNA, we mutated the Lys and Arg residues in these BRD4-S basic patches to either Ala to remove and neutralize the basic side chain or to Glu to reverse the charge. Basic patches 4 and 5 also contain acidic residues so we only mutated the basic residues to Ala to avoid dramatically changing the charge distribution (the 20-residue basic patch 4 contains 14 basic residues and 4 acidic residues). We find that mutating individual basic patches with Ala or Glu substitutions had little effect on BRD4-S’s affinity to H4 K12acK16ac nucleosomes at 70 mM NaCl (Figure 3b, Figure S5a, Table S3). Mutating pairs of basic patches (2 and 3, 4 and 5) or three patches (1, 2 and 3) similarly had minor effects on nucleosome binding. However, when all five BRD4-S basic patches were mutated, we could not detect nucleosome binding. These results suggest that the five BRD4-S basic patches may have redundant roles in mediating BRD4-S’s binding to the nucleosome.

Figure 3: BRD4 basic patches mediate redundant interactions with H4 K12acK16ac nucleosomes.

(a) BRD4-S domains and basic patches highlighted in cartoon and primary sequence (left) and identity of BRD4-S basic patch mutations studied (right), (b) TR-FRET dissociation constants for BRD4-S basic patch mutants binding to H4 K12acK16ac nucleosomes in 70 mM NaCl, (c) TR-FRET dissociation constants for BRD4-S basic patch mutants binding to H4 K12acK16ac nucleosomes in 150 mM NaCl, (d) effect of salt concentration on select BRD4-S basic patch mutations on binding to H4 K12acK16ac nucleosomes.

These results still did not explain how BRD4-S specifically recognizes acetylated nucleosomes. We reasoned that if BRD4-S uses basic residues to bind to nucleosomal DNA, these ionic interactions should depend on ionic strength. We therefore performed salt titrations to assess the ionic strength dependence of nucleosome binding by BRD4-S. Our results show that a clear dependence on ionic strength of BRD4’s affinity to H4 K12acK16ac nucleosomes compared to unmodified nucleosomes with similar affinities at 70 and 100 mM NaCl but increasing differences at 125 and 150 mM NaCl (Figure 2b, Figure S4b). At 150 mM NaCl, generally considered to be physiological ionic strength^47,48, BRD4-S binds to unmodified nucleosomes with a binding affinity of 43 nM versus 12 nM for H4 K12acK16ac nucleosomes for a 3.6x preference (Table S3). Thus, the bromodomain-acetylated histone tail interaction in the context of the nucleosome can be detected at higher ionic strength presumably by decreasing the contributions of the charge-dependent BRD4-S basic residues-nucleosomal DNA interactions. We therefore performed subsequent nucleosome binding affinity experiments in 150 mM NaCl.

We considered the possibility that the preference of BRD4-S for acetylated versus unmodified nucleosomes would be increased in the presence of competitor nucleosomes. We therefore performed the TR-FRET nucleosome assay under competitive conditions where the 1 nM of biotinylated nucleosomes (FRET donor when bound by europium labeled streptavidin) is titrated by increasing concentrations of the HIS-tagged BRD4-S (FRET acceptor when bound by the ULight anti-HIS antibody) in the presence of 5 or 10 nM unmodified competitor nucleosomes (Figure S4c). The results are similar to binding in the absence of competitor nucleosomes: at 70 mM NaCl, BRD4-S binds slightly better to unmodified nucleosomes compared to H4 K12acK16ac nucleosomes (K_d = 37 nM vs 42 nM). At 150 mM NaCl with 5 nM unmodified competitor nucleosomes, BRD4-S binds H4 K12acK16ac nucleosomes 3.9x tighter than unmodified nucleosomes (K_d = 18 nM vs 71 nM). At 150 mM NaCl with 10 nM unmodified competitor nucleosomes, BRD4-S also binds H4 K12acK16ac nucleosomes 3.9x tighter than unmodified nucleosomes (K_d = 26 nM vs 100 nM). Thus, the presence of competitor nucleosomes did not significantly increase the preference of BRD4-S for acetylated over unmodified nucleosomes.

We also examined the effect of using nucleosome arrays instead of mononucleosomes on BRD4-S binding. We employed the TR-FRET in competition mode⁴³ where mononuclesomes saturated with an excess of BRD-S were titrated with a 12x nucleosome array containing unmodified or H4 K12acK16ac histones (Figure S4d). We find that unmodified nucleosome arrays are slightly more efficient than H4 K12acK16ac at competing for binding to BRD4-S (IC₅₀ of 1.3 nM vs 2.4 nM) in 70 mM NaCl, analogous to what was observed for BRD-S binding to mononucleosomes. In 150 mM NaCl, the H4 K12acK16ac nucleosome array was more efficient than the unmodified nucleosome array at competing for binding to BRD4-S (IC₅₀ of 1.0 nM vs 3.3 nM, preference ratio of 3.3). After accounting for the 12 nucleosomes per array, the IC₅₀ values parallel the Kd’s using mononucleosome substrates (IC₅₀ of 12 nM vs Kd of 12 nM for the H4 K12acK16ac substrate, IC₅₀ of 40 nM vs Kd of 43 nM for the unmodified substrate), although it should be noted that the IC₅₀ values are estimates since we were unable to use high enough concentrations of nucleosome arrays to fully compete with BRD4-S/mononucleosome complexes. These results suggest that BRD4-S binds with similar affinity to H4 K12acK16ac nucleosome arrays as it does to mononucleosomes.

In addition to comparing binding to unmodified vs H4 K12acK16ac mononucleosomes, we examined the effect of BRD4 binding at 150 mM NaCl to acetylated H4 tails in the context of the nucleosome by removing the H4 tail, by adding the JQ1 competitive inhibitor³¹ and by mutating the BRD4 bromodomain peptide binding pockets (Figure 4a, b, Figure S6a, Table S3). We observe a similar 3- to 4-fold effect on nucleosome binding in each case. BRD4-S binds to H4 tailless nucleosomes 3.9-fold less tightly than to the H4 K12acK16ac nucleosomes. Similarly, incubating BRD4-S in the presence of JQ1, which competitively binds to the bromodomain acetylated peptide binding pocket, decreased binding affinity 3.1-fold. It is worth noting that despite this reduction in binding in the presence of the JQ1 competitive inhibitor, BRD4-S still binds acetylated nucleosomes tightly with a dissociation constant of 37 nM. The BRD4-S Y97F,N140A mutations in the BD1 acetylated peptide binding pocket, which severely diminished binding to acetyl-lysine peptides^42,49, decreased binding to H4 K12acK16ac nucleosomes 4.1-fold. This was almost twice the effect of the equivalent BD2 mutations, which decreased binding to H4 K12acK16ac nucleosomes only 2.2-fold. As a negative control, we mutated two basic residues on BD1 facing away from the nucleosome in our cryo-EM reconstruction. This BRD4-S K99E,K102E mutant protein bound H4 K12acK16ac nucleosomes with a dissociation constant of 15.6 nM, similar to the 12.2 nM for the wild-type protein. These results indicate that acetylation of histone H4 K12 and K16 contributes only modestly to the binding affinity of BRD4-S to nucleosomes.

Figure 4: Role of BRD4 BD1 bromodomain and basic patch 1 residues in nucleosome binding and structural compactness.

(a) BRD4 BD1 interactions with the nucleosome with key regions highlighted, (b) TR-FRET dissociation constants for BRD4-S bromodomain mutants binding to H4 K12acK16ac nucleosomes, (c) model for how BRD4 basic region 1 could interact with nucleosome DNA minor groove with C_α positions of the 5 basic residues shown in blue spheres, (d) effect of BRD4-S BD1 mutations on BRD4-S/nucleosome complex mobility in gel mobility shift assay, (e) NMR structure of HMG-I(Y) AT-hook Arg-Gly-Arg region binding to DNA (PDB ID 2EZD), protein residues outside of the Arg-Gly-Arg region not shown, (f) TR-FRET dissociation constants for BRD4-S basic patch 1 mutants binding to H4 K12acK16ac nucleosomes, (g) the BRD4-S WT/H4 K12acK16 ac nucleosome complex elutes slightly later than the BRD4-S WT/WT nucleosome complex (peak elution of 9.8 vs 9.6 ml) in Superdex 200 increase size-exclusion chromatography, (h) the BRD4-S m12/H4 K12acK16ac nucleosome complex (blue) elutes earlier than the BRD4-S WT/H4 K12acK16ac nucleosome complex in Superdex 200 increase size-exclusion chromatography, (i) the BRD4-S m12 protein elutes at similar time as the BRD4 WT protein in Superdex 200 increase size-exclusion chromatography

Our binding experiments also validate the interactions observed or deduced from our cryo-EM reconstruction (Figure 4a, b, Figure S6a, Table S3). Mutating BRD4 BD1 helix αZ residues R68, K72 and K76 (in position to interact with nucleosomal DNA) to Glu had a 5-fold adverse effect on BRD4-S’s binding affinity to H4 K12acK16ac nucleosomes, a greater effect than mutating the BD1 peptide binding site. The BRD4 BD1 residue W75 appears to interact with histone H4 residue R23 in our cryo-EM reconstruction of the BRD4/nucleosome complex, and we detect a two-fold decrease in binding affinity of the BRD4-S(W75A) variant to H4 K12acK16ac nucleosomes. However, another BRD4 BD1 residue, A80, which might also interact with H4 R23, showed only a 1.4-fold decrease in binding affinity, an effect similar to the K99E,K102E negative control.

We next analyzed the role of the BRD4 basic patches for binding to H4 K12acK16ac nucleosomes in 150 mM NaCl (Figure 3c, Figure S5b, Table S3). At this higher salt concentration, we now detect differences in binding affinity when individual BRD4 basic patches are mutated. Mutating basic patches 4 and 5 had the largest effect: binding of these mutated BRD4-S proteins to H4 K12acK16ac nucleosomes could not be detected in our TR-FRET binding assay. Partial deletion of basic patch 5 had previously been observed to adversely affect the binding of a BRD4 truncation containing BD2 and the BID to naked DNA⁵⁰, consistent with a possible role for the BRD4 basic patch 5 to bind nucleosomal DNA. A role for basic patches 4 and 5 in nucleosome binding is also consistent with findings that BRD4(462–599) containing the BID was sufficient to bind to nucleosomes purified from a human cell line⁵¹. Mutating basic patch 1 residues to Glu reduced binding about 8-fold, while the equivalent mutations to basic patches 2 or 3 each reduced binding 4 to 5-fold. How much positive charges are reduced in the BRD4 mutant proteins did not always correlate with the effect on nucleosome binding.

Mutating basic patch 5 yielded a BRD4-S variant with a theoretical pI of 7.56 and for which binding to acetylated nucleosomes could not be detected. In contrast, the BRD4-S protein with basic patch 3 basic residues mutated to Glu has a more acidic theoretical pI of 6.60 and this protein bound acetylated nucleosomes with a less severe 4.4-fold reduction in binding affinity. The concentration of salt affected the binding affinity of different BRD4-S basic patch mutant proteins differently. As described previously, at 70 mM NaCl all basic patch mutants examined bound with similar affinity to H4 K12acK16ac nucleosomes except when mutating all five BRD4 basic patches at the same time. At 100 mM NaCl, mutating basic patches 1 or 2+3 had very minor effects for BRD4-S binding to H4 K12acK16ac nucleosomes, whereas mutating basic patches 4+5 decreased the binding affinity 4-fold. These effects were further accentuated at 150 mM NaCl (Figure 3d, Figure S5c, Table S3).

We note that unlike many other nucleosome-binding proteins⁵², BRD4-S does not appear to engage the nucleosome acidic patch since mutations in the nucleosome acidic patch have no detectable effect on BRD4-S binding (Figure S6b). The same mutations adversely affected binding of RCC1, a protein known to use an arginine anchor to bind to the nucleosome acidic patch^52–55. Our results therefore suggest that BRD4-S uses its basic patches redundantly to interact with nucleosomal DNA and not with the nucleosomal histone acidic patch.

Role of BRD4 basic patch 1 in binding nucleosomal DNA

The deleterious effect of mutating the relatively small BRD4 basic patch 1 on nucleosome binding was particularly intriguing because we observe weak density at the C-terminal end of BRD4 BD1 in a subset of particles in our cryo-EM analysis of the BRD4-S/nucleosome complex (Figure S2, Figure S3c). This density extends towards the nucleosomal DNA backbone around SHL+0.5 before approaching the DNA minor groove at the nucleosome dyad. We observed that the central 5 residues of the BRD4 basic patch 1 sequence (KGRGRGRK) contains two overlapping RGR sequences found in the AT-hook DNA-binding motif where the two Arg side chains of the RGR sequence bind across and fill the DNA minor groove (Figure 4e)⁵⁶. For this reason, although we only observe cryo-EM density for the region just before BRD4 basic patch 1 and not for the basic patch 1 itself, we have modeled the BRD4 basic patch 1 in the minor groove (Figure 4c). Our biochemical studies show that the three Arg residues in the BRD4 basic patch 1, R179, R181 and R183, play important roles in BRD4-S binding to the nucleosome. Mutating just these three Arg residues to Glu was sufficient to reduce binding affinity to the H4 K12acK16ac nucleosomes 11-fold (Figure 4f, Figure S6a, Table S3). Furthermore, the basic charge of these three Arg residues is not sufficient for full binding activity since replacing these three Arg residues with Lys led to a 2.3-fold reduction in BRD4-S binding affinity to H4 K12acK16ac nucleosomes. This suggests that one or more of the three Arg side chains specifically interacts with the nucleosome, potentially in the DNA minor groove.

The conformation of the BRD4/nucleosome complex depends on histone acetylation In addition to using time-resolved FRET, we employed the native polyacrylamide gel electrophoresis (PAGE) gel mobility shift assay to analyze BRD4-S’s binding to the nucleosome. We find that the BRD4-S/nucleosome complex migrates slightly faster when H4 K12acK16ac nucleosomes instead of unmodified nucleosomes are used (Figure 4d: lanes 3 and 4). This faster mobility does not appear to be attributed to the acetylation charge neutralization of the H4 K12 and K16 residues since the mobility of the acetylated and unmodified nucleosomes were the same (Figure 4d, lanes 1 and 2). Given that the mobility of a complex in a native PAGE gel depends on both the charge and the shape of the complex, the simplest interpretation of our results is that the BRD4-S/acetylated nucleosome complex is more compact than the BRD4-S/unmodified nucleosome complex.

Our structural and TR-FRET studies have pointed to three regions within or close to BRD4 BD1 that interact with the nucleosome: the BRD4 BD1 peptide binding pocket which interacts with acetylated H4 tail, the R68, K72 and K76 basic residues in BRD4 BD1 helix αZ that interact with nucleosomal DNA, and the BRD4 basic patch 1 residues that appear to interact with the nucleosomal DNA minor groove near the nucleosomal dyad. We find that mutations in each of these three regions decreases the mobility of the BRD4-S/H4 K12acK16ac nucleosome complex on native PAGE. Both the Y97F,N140A mutations (m2 mutation) in the BRD4 BD1 peptide binding pocket and the R68E,K72E,K76E mutations in BRD4 BD1 αZ (m28 mutation) result in BRD4/H4 K12acK16ac nucleosome complexes that migrate slower than the wild-type complex (Figure 4d, lanes 6–8).

The effects of mutating BRD4 basic patch 1 residues were even more dramatic with retardation of the complex when the five basic patch 1 residues were mutated to Ala (m11 mutation) and even larger retardation when the five basic patch 1 residues were mutated to Glu (m12 mutation) (Figure 4d, lanes 8–10). This was true whether H4 K12acK16ac or unmodified nucleosomes were used (Figure 4d, lanes 11–13). The effects were specific since mutating the BRD4 BD1 K99 and K102 residues on the BD1 surface facing away from the nucleosome to Glu (m27 mutation) did not affect the mobility of the BRD4-S/nucleosome complex (Figure 4d, lanes 4 and 5). It is worth noting that the slower mobility of the Ala or Glu mutations is not simply the effect of changing the charge on BRD4-S. Since the Ala or Glu mutations make the BRD4-S protein more negatively charged, one might expect that the BRD4-S/nucleosome complex would migrate faster in a native DNA PAGE gel. However, the Ala or Glu mutations in BRD4 helix αZ or basic patch 1 resulted in slower mobility BRD4-S/nucleosome complexes. To confirm that mobility shift is not an artifact of native gel electrophoresis, we have analyzed the elution of BRD4/nucleosome complexes by Superdex 200 Increase size exclusion chromatography (Figure 4g, h). We find that the BRD4-S/H4 K12acK16ac nucleosome complex elutes slightly later than the BRD4-S/unmodified nucleosome complex, indicative of a smaller hydrodynamic radius and consistent with a more compact structure. The BRD4-S/H4 K12acK16ac nucleosome complex containing the BRD4 basic patch 1 to Glu mutations elutes earlier than the wild-type BRD4-S/nucleosome complex, indicating a larger hydrodynamic radius. The BRD4 protein on its own elutes at essentially the same time with or without the basic patch 1 to Glu mutations (Figure 4i). These results indicate that BRD4-S adopts a more compact structure on H4 K12acK16ac nucleosomes than on unmodified nucleosomes and that disrupting BRD4’s contacts with the nucleosome decreases the compactness of the BRD4/nucleosome complex, consistent with the results from the electrophoresis mobility shift assay.

To examine the possibility that BRD4-S binding might affect the conformation of the BRD4/nucleosome complex at the nucleosome DNA ends, we employed a restriction enzyme accessibility assay (Figure S7a, b, c, d). The Widom 601 nucleosome positioning sequence used in our studies contains a HinfI restriction site 5 bp from one end of the 145 bp nucleosome core sequence. If BRD4-S influences the flexibility of the nucleosomal DNA ends, we would expect to observe differences in HinfI digestion of the nucleosomal DNA in the presence of BRD4-S. Although we do observe more HinfI digestion using H4 K12acK16ac nucleosomes compared to unmodified nucleosomes, similar increases were observed whether wt BRD4-S, BRD4-S m12 basic patch 1 mutant, the SIRT6 histone deacetylase or even BSA were used. This suggests that BRD4-S binding does not increase the mobility of the nucleosomal DNA ends.

BRD4 mutations in basic patches 2, 3 and 4 also decreased the mobility of the BRD4/nucleosome complex in native gel electrophoresis, although generally to a lesser degree than for basic patch 1 (Figure 5a, lanes 3–8). Similar effects were observed with H4 K12acK16ac and unmodified nucleosomes, consistent with binding to the acetylated H4 tail not being the major determinant in nucleosome binding by BRD4 (Figure 5a vs 5b). In contrast to the basic patches 1–4, mutating BRD4 basic patch 5 did not affect the mobility of the BRD4-S/nucleosome complex despite our finding that the same mutations severely affected nucleosome binding in the TR-FRET assay (Figure 5a, lane 10). This suggests that the compactness of the BRD4-S/nucleosome is not necessarily directly related to how tightly BRD4 binds to the nucleosome. The effect of mutating combinations of BRD4 basic patches appears to be additive in the gel mobility assay, with mutating basic patches 1, 2 and 3 or patches 4 and 5 having a larger retardation effect than individual patches (Figure 5a, lanes 11–12).

Figure 5: Effect of BRD4 basic patch mutations on the compactness of BRD4-S/nucleosome complexes in gel mobility shift assay.

using (a) H4 K12acK16ac nucleosomes and (b) unmodified nucleosomes.

Discussion

We show that BRD4 interacts with its nucleosome target via multivalent interactions involving multiple regions of BRD4 and the nucleosome besides the well-characterized BRD4 bromodomain binding to acetylated histone peptides. Our structural studies reveal that BRD4 BD1 is positioned on the nucleosome face to bind to nucleosomal DNA in addition to binding the acetylated H4 tail through the bromodomain peptide binding pocket. This nucleosomal DNA binding is consistent with previous studies of the related BRDT protein which highlighted a positively charged patch on its BD1 and nonspecific binding of that bromodomain to DNA⁵⁷. BRD2, BRD3, BRD4 and BRDT all share a positively charged patch similarly localized on their BD1⁵⁷ and they all bind to acetylated H4 tails using BD1^30,49,58. Since the BRD4 residues R68, K72 and K76 are in position to bind nucleosomal DNA appear to be part of this conserved positively charged bromodomain patch (Figure S7e), it seems reasonable to anticipate that BD1 of BRD2, BRD3 and BRDT may bind in a similar orientation on the nucleosome as BRD4 does in our cryo-EM reconstruction. In contrast, since only the BRD2, BRD3, BRD4 and BRDT residues in BD2 corresponding to BRD4 K72 are positively charged (Figure S7e), BD2 of the other BET proteins might not bind to the nucleosome in this orientation.

Although most inhibitors of BRD4 target the peptide binding pocket, a new class of BRD4 inhibitors has been found to bind to a distinctly different site. The ZL0590 BRD4 inhibitor which possesses anti-inflammatory activities has been determined to bind to BRD4 BD1 helices αA and αB⁵⁹. Superposition of the BRD4 BD1/ZL0590 crystal structure with our BRD4/nucleosome cryo-EM structure shows that the ZL0590 compound is positioned away from the nucleosome (Figure 6). If ZL0590 acts in the context of chromatin and ZL0590 does not act as an allosteric effector of acetyl-lysine peptide binding, the BRD4 BD1 αA and αB surface it binds to may be targets of effector proteins associated with its anti-inflammatory properties.

Figure 6: Structural model for ZL0590 inhibitor bound to the BRD4/nucleosome complex.

Superposition of ZL0590/BRD4 BD1 crystal structure (PDB 6U0D) with BRD4/nucleosome structure (this work) via BRD4 BD1.

Based on the prevailing model that BRD4 is recruited to chromatin via bromodomain binding to acetylated histone tails, we had anticipated that BRD4 would require histone acetylation to bind to nucleosomes. Contrary to this expectation, we find that BRD4 binds tightly (10–40 nM dissociation constant) to both diacetylated and unmodified nucleosomes with only a 3 to 4-fold preference for diacetylated nucleosomes at 150 mM NaCl and even less preference at lower salt concentrations. During review of this manuscript, Lambrechts et al published their finding of a similar 2 to 4-fold preference of BRD4 for multiple permutations of H3 and/or H4 acetylated nucleosomes⁶⁰. Our mutational analyses indicate that critical interactions with the nucleosome are made by other BRD4 regions. We have identified five basic patches in BRD4 that appear to mediate redundant interactions with nucleosomal DNA. These basic patches likely bind nonspecifically to nucleosomal DNA in a conformationally heterogenous manner. This is consistent with the finding that residues in the BID region (which includes basic patches 4 and 5) help BRD4 bind DNA and native nucleosomes, and with NMR studies that a BRD4 BD1-BD2 construct binds DNA in a conformationally dynamic fashion^50,51. The lack of a unique binding mode likely accounts for why the BRD4 basic region interactions with DNA were not observed in our cryo-EM maps. This is analogous to the failure to observe the basic histone tails with nucleosomal DNA in cryo-EM structures of the nucleosome despite the essentially infinite local concentration of histone tails around the nucleosomal DNA. Consistent with the apparent nonspecific interaction of BRD4 basic patches with nucleosomal DNA, analysis of publicly available ChIP-seq data did not detect association of BRD4 binding in vivo with sequence-specific motifs except for motifs for known BRD4-interacting transcription factors as HSF1⁶¹ (personal communication, Shaun Mahony).

We provide evidence that although BRD4 binds with similar affinity to diacetylated and unmodified nucleosomes, the resulting complexes are conformationally different. Experiments using the gel mobility shift assay and size exclusion chromatography suggest that BRD4 adopts a more compact conformation on acetylated vs non-acetylated nucleosomes. Mutating the BD1 peptide binding pocket and the BD1 nucleosome DNA-binding surface had similar effects in the gel mobility shift assay as removing the histone H4 acetylation marks, consistent with these BRD4 BD1 regions interacting with the nucleosome. Mutating BRD4 basic patches 1–4 also appear to make the corresponding BRD4/nucleosome complexes less compact, with mutations in basic patch 1 immediately following BD1 having the largest effect.

A mechanism of the role of basic patch 1 in nucleosome binding is suggested by our cryo-EM maps which show that the BRD4 region immediately C-terminal to BD1 approaches the nucleosomal DNA near the dyad. This positions the basic patch 1 to enter the DNA minor groove. We propose that the two overlapping RGR sequences enable redundant AT-hook-like binding in the DNA minor groove. Doing so provides an additional tether of BRD4 BD1 to the nucleosome in addition to the bromodomain-acetylated H4 tail and the bromodomain αZ-nucleosome DNA interactions. Disrupting any of these tethers destabilizes the positioning of BRD4 BD1 on the nucleosome, resulting in a conformationally less compact and presumably more flexible complex. The finding that mutations of the basic residues in the basic region 1 have a stronger effect on the affinity of nucleosome binding and on the extent of the gel mobility shift underlines the importance of basic region 1’s interactions with the nucleosome, perhaps to help position BD1 appropriately on the nucleosome face. The interactions of BD1 with the nucleosome do not involve a large contact surface area to stabilize the binding (occluded solvent accessible surface area of 1450 Ų vs 2725 Ų for SIRT6/nucleosome or 3570 Ų for SIR3/nucleosome^39,62 for example). Instead, the three contact points (BD1 binding pocket with acetylated histone H4 tail, BD1 αZ interactions with nucleosomal DNA and basic patch 1 interactions with nucleosomal DNA at the dyad) tether BD1 to the nucleosome without a large contact surface. The potential resulting wobbliness of BD1 on the nucleosome may account for the lower resolution of cryo-EM reconstruction for the bromodomain component of the complex compared to the nucleosome.

The interactions of the BRD4 basic patch 1 with the nucleosome are particularly interesting because they suggest a mechanism for regulating the recruitment of BRD4 to chromatin. BRD4 has been found to be methylated at R179, R181 and R183, the very arginine residues in the RGRGR overlapping RGR sequence in basic region 1 that we have identified as potentially interacting with the DNA minor groove at the nucleosome dyad^63–65. In one study, methylation of these three BRD4 arginine residues by the arginine methyltransferases PRMT2/4 regulated transcription and DNA repair⁶³. In a second study, methylation of the same three BRD4 arginine residues promotes BRD4 phosphorylation and ovarian cancer metastasis⁶⁴. More recently, methylation of these three BRD4 arginine residues by PRMT1 was found to promote partial epithelial-mesenchymal transformation and renal fibrosis⁶⁵. Our findings which suggest that Arg179, Arg181 and Arg183 stabilize BRD4 interaction to the nucleosome by binding in the DNA minor groove could provide the molecular basis for how arginine methylation affects BRD4 function through its binding to the nucleosome. Further studies will be needed to examine how symmetric arginine methylation by PRMT2/4 and asymmetric arginine methylation by PRMT1 affect BRD4 binding to the nucleosome.

The binding of BRD4 bromodomain 1 to an acetylated H4 tail in our cryo-EM reconstruction is structurally distinct from other chromatin reader domains binding to modified nucleosomes (Figure 7). In the structures of the LEDGF PWWP domain binding to H3K36me3 nucleosome⁶⁶, the replication ORCA WD40 domain binding to H4K20me3 nucleosomes⁶⁷ and the 53BP1 Tudor domain binding to H4K20me2 nucleosomes⁶⁸, the chromatin reader domain largely occludes the modified histone tail, making it much less accessible. In contrast, BRD4 BD1 is positioned on the nucleosome such that the H4 tail in its binding pocket is accessible. We speculate that this arrangement allows other chromatin factors to detect the acetylated histone tail as it is bound by the BRD4 bromodomain on a nucleosome, thereby increasing the specificity of the chromatin factor interaction with the BRD4/acetylated nucleosome complex.

Figure 7: Modified histone peptide is exposed in BRD4/nucleosome complex compared to other chromatin reader/nucleosome complexes.

BRD4-S bromodomain 1/H4 K12acK16ac nucleosome (this work), LEDGF PWWP domain/H3 K36me3 nucleosome (PDB 6S01), ORCA/Orc2/H4 K20me3 nucleosome (PDB 8SIY), 53bp1Tudor domain/gH2AX K15ub-H4 K20me2 nucleosome (PDB 7YQK). The chromatin reader protein is show in pink and the modified lysine residue is shown in dark blue. Histones H2A, H2B, H3 and H4 are shown in yellow, red, blue and green and nucleosomal DNA is show in grey.

Limitations of the Study

Our finding that BRD4 binds tightly to unmodified nucleosomes and that H4 K12 and K16 acetylation improves binding affinity only 3–4 fold at 150 mM NaCl concentration is contrary to the expectation that histone acetylation is required for BRD4 to bind to chromatin based on observations that BRD4 binds specifically to acetylated chromatin in cells^26,69,70. It is, however, consistent with the finding that almost all of BRD4 is bound to chromatin requiring high salt to extract from nuclei^26,46,71, and with immunofluorescence results that essentially all of BRD4 is associated with mitotic chromatin⁷². It is also consistent with findings that the JQ1 compound, which competes for acetylated histone tail binding in the bromodomain peptide binding site, only partially removed BRD4 from chromatin in vivo²¹. It could also explain why 35% of BRD4 genomic binding sites do not contain H4 acetylation marks⁶⁹. Our results also provide more information to understand how targeted BRD4 degradation can inhibit cancer more effectively than BET inhibition^19,73. Nevertheless, our finding that BRD4-S binds tightly to nucleosomes with ~40 nM dissociation constant in the absence of acetylation begs the question of how BRD4 distinguishes between unmodified and acetylated chromatin in the cell. It is unlikely that nucleosome arrays in cells play a critical role since we find a similar preference of BRD4-S for acetylated vs unmodified nucleosomes whether mononucleosome or nucleosome array substrates are used. There are at least two possible resolutions to this apparent paradox. Firstly, BRD4 in a cell might contain post-translational modifications, including phosphorylation, which might modulate its binding to chromatin⁴⁶. Secondly, BRD4 has been shown to form biomolecular condensates providing different environments from our in-solution experiments^50,74,75. Our results that BRD4 binds tightly to unmodified nucleosomes does suggest that specific recruitment of BRD4 to acetylated nucleosome in cells may involve mechanisms that prevent BRD4 from binding to unmodified nucleosome as well as ones that promote binding to acetylated nucleosomes. It is possible that the tighter binding of BRD4-S versus BRD4-L to acetylated chromatin observed in vivo²⁷ may play a role.

Resource Availability

Lead Contact

Song Tan, sxt30@psu.edu

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement

Data and code availability

The atomic coordinates of the BRD4-S/acetylated nucleosome and BRD4-S with basic patch 1/acetylated nucleosome complexes have been deposited to the RCSB Protein Data Bank with PDB ID 36IV and PDB 36IU. The cryo-EM Coulomb potential maps were deposited in the Electron Microscopy Data Bank as EMD-77606 (BRD4-S/acetylated nucleosome) and EMD-77605 (BRD4-S with basic patch 1/acetylated nucleosome). The raw cryo-EM data will be shared by the lead contact upon request. Original imaging data have been deposited at Mendeley at DOI:10.17632/c5pwxv8c2s.1 and are publicly available as of the date of publication.

This paper does not report original code

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request

Experimental Model and Study Participant Details

E. coli strains TG1 and HB101 were used for gene cloning and expression of nucleosomal DNA, respectively. E. coli BL21(DE3)pLysS and Rosetta 2(DE3)pLysS was used for expression of BRD4 and its mutants. C321.ΔA.exp was used for expression of acetylated histone H4 by unnatural ammino acid incorporation. Plasmids used in this study are detailed in the key resources table.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit monoclonal anti-Histone H4K12ac	EpiCypher	Cat# 13-0037
rabbit polyclonal anti-Histone H4K16ac	Acitve Motif	Cat# 39068; RRID:AB_2636968
Histone H4 antibody	Abcam	Cat# ab17036; RRID:AB_1209245
m-IgGκ BP-HRP antibody	Santa Cruz Biotechnology	Cat# sc-516102; RRID:AB_2687626
Bacterial and virus strains
E. coli BL21(DE3)pLysS	Invitrogen	Cat# C602003
Rosetta 2(DE3)pLysS Singles	MilliporeSigma	Cat# 71401-3
E. coli TG1	Toby Gibson
E. coli HB101	ATCC	Cat# 67593
C321.ΔA.exp	Addgene	Cat# 49018
Chemicals, peptides, and recombinant proteins
Q5 High-Fidelity DNA polymerase	New England Biolabs	Cat# M0491L
Benzamidine-HCl	RPI	Cat# B12000100.0
2-mercaptoethanol	Acros	Cat# 12547-0010
TEV protease	This study	N/A
1,4-Dithio-DL-thretiol	Gold Bio	Cat# DTT100
Nε-Acetyl-L-lysine	Chem Impex	Cat# 05364
Nicotinamide	Millipore Sigma	Cat# N3376
UREA	Fisher Scientific	Cat# U15-50
Guanidine Hydrochloride	Fisher Scientific	Cat# BP178-1
(+)-JQ1	Selleck Chemical	Cat# S7110
Sodium phosphate, dibasic, anhydrous	Fisher Scientific	Cat# S374-1
Potassium phosphate, monobasic	Fisher Scientific	Cat# BP362-1
NP-40	CalBioChem	Cat# 492015
CHAPS	VWR	Cat# 97061-716
LANCE Ultra ULight-anti-6xHIS	Revvity, Inc.	Cat# TRF0134-M
LANCE Eu-8044 streptavidin	Revvity, Inc.	Cat# AD0060
Histone H3, H3 mutants	This study	N/A
Histone H4, H4 mutants	This study	N/A
Histone H2A	This study	N/A
Histone H2B	This study	N/A
Human BRD4, BRD4 mutants	This study	N/A
Critical commercial assays
West Pico PLUS Chemiluminescent Substrate	Thermo Scientific	Cat# 34580
Deposited data
BRD4 BD1-acetylated nucleosome complex	This study	PDB: 36IV
BRD4 BD1 with basic patch 1-acetylated nucleosome complex	This study	PDB: 36IU
BRD4 BD1-acetylated nucleosome complex Cryo-EM map	This study	EMDB: EMD-77606
BRD4 BD1 with basic patch 1-acetylated nucleosome complex Cryo-EM map	This study	EMDB: EMD-77605
TR-FRET raw data, blot images and uncropped gel	This study	DOI:10.17632/c5pwxv8c2s.1 https://data.mendeley.com/preview/c5pwxv8c2s?a=45bd27b8-8459-41cbbcf8-4ff7292a5990
Oligonucleotides
177 bp 601 Cy5 forward primer (ATCATGTGATGGACCCTATACGCGGCCGCCCTGGAG)	IDT	N/A
Recombinant DNA
pGEX-6P-1 BRD4 full-length	Addgene	Plasmid #14447
pST50Tr-HISNhBRD4t1	This study	expresses HISNhBRD4Δ1 and hBRD4Δ1=hBRD4 (2–722)
pST50Tr-HISNhBRD4t1x2	This study	expresses HISNhBRD4Δ1m2 (Y97F,N140A)
pST50Tr-HISNhBRD4t1x3	This study	expresses HISNhBRD4Δ1m3 (A80V)
pST50Tr-HISNhBRD4t1x8	This study	expresses HISNhBRD4Δ1m8 (Y390F,N433A)
pST50Tr-HISNhBRD4t1x9	This study	expresses HISNhBRD4Δ1m9 (K283A,K285A,K286A,K289A,R290A, K291A)
pST50Tr-HISNhBRD4t1x10	This study	expresses HISNhBRD4Δ1m10 (K283E,K285E,K286E,K289E,R290E, K291E)
pST50Tr-HISNhBRD4t1x11	This study	expresses HISNhBRD4Δ1m11 (177A,R179A,R181A,R183A,K184A)
pST50T r-HISNhBRD4t1 x12	This study	expresses HISNhBRD4Δ1m12 (K177E,R179E,R181E,R183E,K184E)
pST50Tr-HISNhBRD4t1x13	This study	expresses HISNhBRD4Δ1m13 (K314A,K317A,R321A,R322A,R326A, K329A,K332A,K333A)
pST50T r-HISNhBRD4t1 x14	This study	expresses HISNhBRD4Δ1m14 (K314E,K317E,R321E,R322E,R326E, K329E,K332E,K333E)
pST50Tr-HISNhBRD4t1x15	This study	expresses HISNhBRD4Δ1m15 (K535A,K537A,K538A,K539A,K541A, K543A,K544A,K546A,K547A,K548A, K550A,K552A,R553A,K554A)
pST50T r-HISNhBRD4t1x16	This study	expresses HISNhBRD4Δ1m16 (K561A,K562A,K564A,K566A,K571A, K572A,K574A,K575A)
pST50Tr-HISNhBRD4t1x19	This study	expresses HISNhBRD4Δ1m19 (K283A,K285A,K286A,K289A,R290A, K291A,K314A,K317A,R321A,R322A, R326A,K329A,K332A,K333A)
pST50Tr-HISNhBRD4t1x20	This study	expresses HISNhBRD4Δ1m20 (K283E,K285E,K286E,K289E,R290E, K291E,K314E,K317E,R321E,R322E, R326E,K329E,K332E,K333E)
pST50Tr-HISNhBRD4t1x21	This study	expresses HISNhBRD4Δ1m21 (K177A,R179A,R181A,R183A,K184A, K283A,K285A,K286A,K289A,R290A, K291A,K314A,K317A,R321A,R322A, R326A,K329A,K332A,K333A)
pST50Tr-HISNhBRD4t1x26	This study	expresses HISNhBRD4Δ1m26 (K177E,R179E,R181E,R183E,K184E, K283E,K285E,K286E,K289E,R290E, K291E,K314E,K317E,R321E,R322E, R326E,K329E,K332E,K333E,K535A, K537A,K538A,K539A,K541A,K543A, K544A,K546A,K547A,K548A,K550A, K552A,R553A,K554A,K561A,K562A, K564A,K566A,K571A,K572A,K574A, K575A)
pST50Tr-HISNhBRD4t1x27	This study	expresses HISNhBRD4Δ1m27 (K99E,K102E)
pST50Tr-HISNhBRD4t1x28	This study	expresses HISNhBRD4Δ1m28 (R68E,K72E,K76E)
pST50Tr-HISNhBRD4t1x37	This study	expresses HISNhBRD4Δ1m37 (W75A)
pST50Tr-HISNhBRD4t1x38	This study	expresses HISNhBRD4Δ1m37 (R179E,R181E,R183E)
pST50Tr-HISNhBRD4t1x42	This study	expresses HISNhBRD4Δ1m42 (R179A,R181 A,R183A)
pST50Tr-HISNhBRD4t1x43	This study	expresses HISNhBRD4Δ1m43 (R179K,R181 K,R183K)
pST50Tr-HISNhBRD4t1x45	This study	expresses HISNhBRD4Δ1m45 (R181E)
pST50Tr-HISNhBRD4t1x46	This study	expresses HISNhBRD4Δ1m46 (K561A,K562A,K564A,K566A,K571A, K572A,K574A,K575A)
pST50Tr-HISNhBRD4t6	This study	expresses HISNhBRD4Δ6=HISNhBRD4 (40– 170)
pST50Tr-HISNhBRD4t11	This study	expresses HISNhBRD4Δ11=HISNhBRD4 (349– 470)
pST101-15xNCP601 a165M	This study	to isolate 165 bp 601 sequence
pST50T rc2-xH4t13	This study	express Xenopus H4 (24–102)
pCDF_PylT-HISNxH3	gift from Jason Chin	N/A
pBK-AcKRS3	gift from Jason Chin	N/A
pCDF_PylT-HISNxH3x55	This study	expresses Xenopus H3 with acetylated K18
pBAD-CDF-xH3(Δ93-98)-TEV-xH4	gift from Heinz Neumann	N/A
pBAD-CDF-HISxH3t14NxH4x51	This study	expresses Xenopus H4 with acetylated K12 and K16.
Software and algorithms
CryoSparc V 4.6	Punjani et al.	https://cryosparc.com
UCSF Chimera X	Pettersen et al.	https://www.cgl.ucsf.edu/chimerax/
Phenix v1.20	Afonine et al.	https://phenix-online.org
ImageJ	Schneider et al.	https://imagej.nih.gov/ij/
Visual Studio Code	Microsoft Corporation	https://code.visualstudio.com
PyMol	Schrödinger	https://pymol.org
Other
Ni-NTA Superflow resin	QIAGEN	Cat# 30410
Talon Superflow metal affinity resin	Clontech	Cat# 635670
Nitrocellulose blotting membrane	GE Healthcare	Cat# 10600009
Source 15S cation-exchanger	GE Healthcare	Cat# 17-0944-01
Source 15Q anion-exchanger	GE Healthcare	Cat# 17-0947-01
Superdex 200 Increase 10/300 GL	GE Healthcare	Cat# 28-9909-44
Vivaspin 500 centrifugal concentrator MWCO:30kDa	Sartorius	Cat# VS0122
Vivaspin 20 centrifugal concentrator MWCO:30kDa	Sartorius	Cat# VS2022
Quantifoil R1.2/1.3 Cu 300 mesh grid	Quantifoil	Cat# N1-C14nCu30-01
Vitrobot Mark IV	Thermo Fisher Scientific	N/A
384-well non-binding microplate	Greiner Bio-One	Cat# GBO-784904
VICTOR Nivo Multimode Microplate Reader	Revvity, Inc.	N/A

Method details

Plasmid construction

The coding region of the short isoform of human BRD4 (residues 1–722) was amplified from the plasmid pGEX-6P-1 BRD4 full-length (a gift from Peter Howley, AddGene plasmid #14447⁷⁶) and cloned into the pST50Tr-HISN expression vector⁷⁷. Point mutants were generated using PCR-based site-directed mutagenesis. The plasmids pCDF_PylT-HISNxH3 and pBK-AcKRS3 were gifts from Jason Chin, and pBAD-CDF-xH3(Δ93–98)-TEV-xH4 was a gift from Heinz Neumann. These plasmids were used for the expression of acetylated histones H3 and H4 through unnatural amino acid incorporation. Amber codons were introduced in place of the target lysine codons by PCR-based site-directed mutagenesis.

Protein preparation

HIS-tagged BRD4 and its mutants were expressed in E. coli BL21(DE3)pLysS or Rosetta 2(DE3)pLysS cells. Cells were grown at 37 °C in 2xTY broth for IPTG-induced expression or in autoinduction medium. Protein expression was induced with 0.2 mM IPTG when the culture reached an OD600 of 0.6–0.8 or by autoinduction⁷⁷ and was carried out at 23°C for 16–18 h. Cell pellets were resuspended in lysis buffer (50 mM sodium phosphate, pH 7.0, 300 mM NaCl, 1 mM benzamidine, and 10 mM 2-mercaptoethanol), flash-frozen in liquid nitrogen, and lysed by sonication. Clarified lysates were loaded onto a metal affinity column (Talon resin, Clontech), and bound proteins were eluted with 100 mM imidazole. For structural studies, the HIS tag was removed by tobacco etch virus (TEV) protease, whereas it was retained for TR-FRET studies. Further purification was performed using Source S cation-exchange chromatography (Cytiva) followed by Source Q anion-exchange chromatography (Cytiva). Dynamic light scattering analysis confirmed that the purified BRD4 proteins were free of detectable aggregation.

Diacetylated histone H4 was generated through Amber suppression as previously described^78,79. Briefly, pBK-AcKRS3 was transformed into C321.ΔA.exp cells to generate competent cells, which were subsequently transformed with the expression vector pBAD-CDF-xH3(Δ93–98)-TEV-xH4. A 50 mL starter culture was grown overnight at 37 °C and used to inoculate 500 mL of 2xTY broth. When the culture reached an OD600 of approximately 1.0, N^ε-Acetyl-L-lysine and nicotinamide were added to final concentrations of 5 mM and 20 mM, respectively. After incubation for an additional 1 h, protein expression was induced with 0.2% arabinose and continued at 37 °C for 18–20 h. Cells were harvested and lysed following the same procedure used for BRD4 expression. Acetylated H4 was isolated from inclusion bodies as previously described⁸⁰ and resolubilized in 6 M guanidine-HCl, 20 mM Tris (pH 8.0). The protein was further purified by metal affinity chromatography (Ni-NTA Superflow resin, QIAGEN) and eluted with buffer containing 20 mM NaAc (pH 5.2), 50 mM NaCl, 50 mM NH₄Cl, 10 mM 2-mercaptoethanol, 20 mM nicotinamide, 8 M urea, and 300 mM imidazole. The protein tag was removed by TEV protease during dialysis, and the lyophilized protein was stored at −20°C.

Nucleosome preparation

Recombinant Xenopus laevis and human core histones, as well as nucleosome core particles, were prepared as previously described⁸⁰. Nucleosomes contained Xenopus histones H3 (98.5% identical to human H3) and H4 (identical to human H4), together with human H2A and H2B. Briefly, wild-type histones were expressed in E. coli BL21(DE3)pLysS cells, isolated from inclusion bodies, and refolded into H3/H4 tetramers and H2A/H2B dimers. Histone complexes were further purified by Source S cation-exchange chromatography (GE Healthcare, Cat#17–0944-01). Nucleosomes were reconstituted by salt dialysis of 601 DNA with H3/H4 tetramers and H2A/H2B dimers, transitioning from high-salt buffer (Tris-Cl, pH 7.5, 1 mM EDTA, 1 mM DTT, and 2 M KCl) to low-salt buffer (Tris-Cl, pH 7.5, 1 mM EDTA, 1 mM DTT, and 0.25 M KCl). Reconstituted nucleosomes were separated from free DNA and excess histone dimers and tetramers by Source Q anion-exchange chromatography (GE Healthcare, Cat#17–0947-01) and subsequently dialyzed into nucleosome storage buffer (10 mM potassium cacodylate, pH 6.5, and 0.1 mM EDTA). The tailless histone H4 construct corresponds to H4(24–102). Histone mutations were generated by PCR-based site-directed mutagenesis. H4 K12acK16ac modifications were validated by mass spectrometry and Western blotting (Figure S7f, g). Nucleosomes used for cryo-EM studies contained a 165 bp (10+145+10) Widom 601 nucleosome-positioning sequence.

Cryo-EM sample preparation and data collection

BRD4-S was reconstituted with nucleosomes in reconstitution buffer [10 mM HEPES pH 7.5, 20 mM KCl, and 1 mM dithiothreitol (DTT)] at a BRD4-S to nucleosome ratio of 2.1:1. The resulting complex was cross-linked using the GraFix method⁸¹. A glycerol gradient was prepared using a light buffer [10 mM HEPES pH 7.5, 20 mM KCl, 1 mM DTT, and 10% glycerol] and a heavy buffer [10 mM HEPES pH 7.5, 20 mM KCl, 1 mM DTT, 40% glycerol, and 0.15% glutaraldehyde], generating a 10–40% glycerol gradient with a corresponding 0–0.15% glutaraldehyde gradient. Gradient fractions were analyzed via native polyacrylamide gel electrophoresis (PAGE), and fractions containing the BRD4/nucleosome complex were concentrated to ~4 μM.

Cryo-EM grids of the BRD4/nucleosome complexes were prepared using standard procedures⁸². A 3 μl aliquot of the BRD4/nucleosome sample was applied to holey carbon 1.2/1.3 Cu300 mesh grids (Quantifoil) using a FEI Vitrobot Mark IV, maintained at 4°C with 100% humidity. The sample was blotted for 3.5 seconds with a blot force of −1 and subsequently plunge-frozen into liquid ethane.

The microscopy dataset was collected at the Pacific Northwest Cryo-EM Center using a FEI Titan Krios microscope operated at 300 keV and equipped with a Gatan K3 direct electron detector. Two datasets, consisting of 11,012 and 18,227 movies, respectively, were combined for a total of 29,239 movies. Data were collected in counting mode at a magnification of ×22,500, corresponding to a pixel size of 1.0125 Å. Images were recorded with a defocus range of −0.5 to −2.0 μm and an accumulated electron exposure of ~40 e⁻/Ų distributed across 40 frames (Table S1).

Cryo-EM data processing

The BRD4/nucleosome dataset was processed in cryoSPARC⁸³. Raw movies were motion-corrected using Patch Motion Correction (multi), and dose-weighted micrographs⁸⁴ were generated. Defocus values were estimated with Patch CTF Estimation (multi). Initial particle picking was performed on a subset of micrographs using Blob Picker, with the particle diameter set to 100–200 Å. Picked particles were extracted with 4× binning in a 256-pixel box Fourier-binned to 64 pixels, resulting in a pixel size of 4.236 Å. Two-dimensional (2D) classification was conducted to remove classes containing obvious junk (Figure S2).

Remaining particles were subjected to ab initio reconstruction to generate input classes for heterogeneous refinement. A cleaned subset of nucleosomal particles was then used to generate templates for particle picking on the complete dataset. Template Picker, with a particle diameter set to 200 Å, was employed to obtain particles across all micrographs. Subsequent rounds of 2D classification removed additional junk classes, and ab initio reconstruction was repeated to refine input classes for heterogeneous refinement.

Multiple rounds of 3D classification on a cleaned dataset of nucleosomal particles ultimately yielded a single class with clear density for the BRD4 bromodomain. To improve the quality of the complex, focused 3D classification was performed with a mask centered on the BRD4 bromodomain. From this process, a class of 80,704 particles, representing the most stable positioning of BRD4 bromodomain I on the nucleosome, was selected. These particles were further refined using non-uniform refinement in cryoSPARC⁸⁵.

Model building and refinement

The nucleosome component of the complex was modeled using Protein Data Bank (PDB) entry 3LZ0⁸⁶ as the starting model, while PDB entry 3UVX⁴⁹ served as the starting model for BRD4 BD1. Both models were rigid-body fitted into the 2.89-Å reconstruction using the “fit in map” function in UCSF ChimeraX⁸⁷ and subsequently optimized in Coot⁸⁸.

The highest-resolution 2.89-Å reconstruction was used in Coot to refine the histones and core DNA, and to build the histone H4 N-terminal tail. The final model was refined with phenix.real_space_refine⁸⁹, incorporating secondary structure, Ramachandran, and rotamer restraints. The model underwent manual validation in Coot and comprehensive validation (cryo-EM) in Phenix using MolProbity⁹⁰. Model statistics are provided in Table S2.

Time-resolved FRET nucleosome binding assays

TR-FRET assays were performed as described previously⁴³. Assay buffer (20 mM HEPES pH 7.5, 70–150 mM NaCl, 5 mM DTT, 5% glycerol, 0.01% NP-40, 0.01% CHAPS, 0.1 mg/mL BSA) was used to prepare 2x acceptor mixtures containing ULight anti-6xHIS acceptor antibody (Revvity) with 6xHIS-tagged BRD4 proteins at a ratio of 1:4 or 1:8 and serially diluting across 12 concentrations. 2x donor mixtures were prepared by mixing 4 nM streptavidin-Eu (Revvity) with or without 2 nM nucleosomes containing 177 bp of Widom 601 DNA (31+145+1) with a 5′ biotin group on the 31-bp extension. When used, the JQ1 compound was present at a concentration of 20 μM.

Samples were prepared by mixing 5 μL of donor mixtures with 5 μL of acceptor mixtures at each concentration in white, non-binding 384-well plates. Following a 30-minute incubation, fluorescence signals were acquired at room temperature in a Victor Nivo multimode fluorescent plate reader (PerkinElmer) using an excitation filter at 320 nm and emission filters at 615 and 665 nm. Emission signals were measured simultaneously following a 100-μs delay. Binding data were fit to the Hill equation:

$$Y=\frac{X^{nH}}{{K_D}^{nH}+X^{nH}}$$

where Y is the normalized fluorescence signal, X is the acceptor protein concentration, K_d is the apparent dissociation constant, and nH is the Hill coefficient⁹¹. Data were fit using the “curve_fit” function of the SciPy Python library⁹². Data are plotted as means ± standard deviations of triplicate measurements, and K_d and Hill coefficient values are reported as means ± standard error of the mean from independent fits. Statistical significance of K_d differences between unmodified and acetylated nucleosomes or between wild-type and mutant BRD4-S proteins were evaluated using a two-tailed unpaired t-test. We estimate we are able to detect nucleosome binding with dissociation constants less than 20 μM.

Time-resolved FRET competition assay

TR-FRET competition assays were performed as described previously⁴³. A 12× 172 bp Widom 601 array DNA was prepared from the pWM+12×601_172NRL plasmid, a generous gift from the Michael Rosen Lab, as described^93,94. Nucleosome arrays were prepared using this 12x array DNA and unmodified or H4 K12acK16ac histones⁹⁴. 10 nM unlabeled 12x nucleosome arrays in assay buffer (20 mM HEPES pH 7.5, 70 or 150 mM NaCl, 5 mM DTT, 5% glycerol, 0.01% NP-40, 0.01% CHAPS, 0.1 mg/mL BSA) was serially diluted across 12 concentrations. 5 μL of the competitor dilutions were added to 5 μL of a 2x master mix containing 2 nM nucleosome with 177 bp of Widom 601 DNA (31+145+1) 5’ biotin labeled on the 31-bp extension, 4 nM streptavidin-Eu (Revvity), 200 nM 6xHIS-tagged BRD4-S, and 10 nM ULight anti-6xHIS acceptor antibody (Revvity). The ratio of emission signals (665 to 615 nm) was used to fit IC₅₀ values for the competition data according to the following equation:

$$Y=\left(F_{min}-F_{max}\right)\left(\frac{X}{I{C}_{50}+X}\right)+F_{max}$$

where Y is the normalized fluorescence signal, X is the competitor nucleosome array concentration, F_min and F_max are the minimum and maximum signals, respectively, and IC₅₀ is the half-maximal inhibitory concentration. Data were fit using the “cure_fit” function of the SciPy Python library⁹². Data were plotted as means ± standard deviations of triplicate measurements, and IC₅₀ values reported as means ± standard error of the mean from independent fits.

Electrophoresis gel mobility shift assay

20 μl binding reactions were performed with 100 nM BRD4 variants incubated with 100 nM nucleosome core particles for 30 minutes at room temperature. The binding buffer consisted of 20 mM HEPES pH 7.5, 125 mM NaCl, 5 mM DTT, 5% glycerol, 0.01% NP-40, 0.01% CHAPS, and 100 μg/ml BSA. Following incubation, the reactions were resolved on 4.5% native PAGE gels at 120 V for 40 minutes at room temperature. Gels were stained with ethidium bromide and imaged using a GelDoc Go system (Bio-Rad). Representative results from at least two replicates are shown.

Size exclusion chromatography

BRD4 and its mutants were reconstituted with nucleosomes in a buffer containing 10 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM DTT, and 0.1 mM PMSF before injecting onto a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with the reconstitution buffer and eluted at 0.5 ml/min. A 300 μL sample of the BRD4-nucleosome complex or nucleosome alone at a concentration of ~1 μM was loaded onto the column and eluted using the same buffer.

Restriction enzyme accessibility assay

Restriction enzyme accessibility assays were performed in 10 μL reactions containing 1X NEB CutSmart buffer [50 mM potassium acetate, 20 mM Tris-acetate pH 7.9, 10 mM magnesium acetate, 100 μg/mL BSA], 1 μg of reconstituted nucleosomes containing the same 177 bp Widom 601 DNA (31+145+1) used in the TR-FRET binding assay except with a 5′ Cy5 group on the 31-bp extension, and recombinant BRD4 mutants or control proteins at a 2:1 molar ratio relative to nucleosomes. After incubation at room temperature for 10 minutes, 10 units of HinfI restriction enzyme (New England Biolabs) was added to each reaction, followed by additional 30 minutes incubation at room temperature. To terminate the reactions and degrade histone proteins, 10 μL of 2X quenching buffer [20 mM HEPES pH 7.5, 10% glycerol, 0.2% SDS, 10 mM EDTA, and 200 μg/mL proteinase K] was added, followed by incubation at 37 °C for 1–1.5 hours. Digestion products were resolved on a 10% native PAGE gel and the gel imaged using both a BioRad GelDoc Go system and a Typhoon laser scanner (Cytiva). Band intensities were quantified using ImageJ software⁹⁵. Representative results from at least two replicates are shown.

Quantification and statistical analysis

Quantification and statistical analyses are described in the figure legends and STAR Methods section. All TR-FRET experiments were performed with three independent titrations (n = 3). Data are presented as the mean ± standard deviation. All electrophoretic mobility shift assays and restriction enzyme accessibility assays were repeated at least once.

The Data S1.zip file contains the pdb coordinate files, wwPDB validation reports and cryo-EM maps for the BRD4/nucleosome cryo-EM models without and with basic patch 1.

Supplementary Material

Supplemental information

Figures S1-S7

Tables S1-S3

Highlights.

How the BRD4 first bromodomain binds to acetylated nucleosomes is revealed by cryo-EM

BRD4 BD1 presents the histone tail for potential interactions with other proteins

BRD4-S binds with high affinity to both acetylated and unacetylated nucleosomes

Histone H4 acetylation influences the conformation of the BRD4-S/nucleosome complex