Structural analysis of nucleosomal barrier to transcription

Our experiments were conducted using mononucleosomal templates that recapitulate many important aspects of the Pol II-type mechanism of chromatin transcription in vivo (21, 27, 28). The mechanism of formation of the nucleosomal barrier to transcription was studied using the 601 and 603 mononucleosomes (Fig. S1A) that contain a polar barrier DNA sequence (PBS) and form a very strong nucleosomal barrier in one of the two transcriptional orientations (nonpermissive orientation) (14, 16, 29). The templates were designed to allow stalling of Pol II at the positions −41 and −5 (41 or 5 bp upstream of the promoter–proximal nucleosomal boundary) in the presence of various partial combinations of NTPs (Fig. S1A). Pol II is spontaneously arrested during transcription of the nonpermissive 601 nucleosome at 150 mM KCl at the +46 and +48 positions (Fig. S1B). The arrest is reversible, as Pol II can transcribe the entire template upon histone octamer destabilization at 1 M KCl (Fig. S1B).

For initial analysis of the nucleosomal barrier, we have conducted detailed mutagenesis of the PBS +(87–102, regions R1–R3, Fig. 1A) region that dictates the height of the nucleosomal barrier to transcription (16). The 603 template was selected for this study because its PBS sequence was recently mapped (16).

The effect of mutations in R1, R2, and R3 subregions of the PBS on the affinity to core histones to DNA and the barrier to Pol II transcription were analyzed (Fig. 1 and Fig. S2). Only mutations in the 603-R2 sequence result in a strong decrease of the high affinity to histones (Fig. 1 A and B and Fig. S2A). Note that none of the individual sets of R mutations (R1, R2, or R3) alone affects the mobility of the nucleosomes in the native gel that indicates identical nucleosome positioning on the template (Fig. S2A) (16). Surprisingly, the R2 mutations only minimally affect the height of nucleosomal barrier to transcription, determined primarily by the promoter–distal R3 +(99–102) sequence (Fig. 1B and Fig. S2B).

Previously we have proposed that the nucleosomal barrier is relieved as a result of uncoiling of ∼100-bp DNA region in front of the enzyme after formation of a small, Pol II-containing DNA loop on the surface of histone octamer (Ø loop) (Fig. 1C) (16, 30). Our current data support this model, because otherwise it is very difficult to explain how the R3 sequence can affect the nucleosomal barrier to Pol II formed almost 60 bp upstream of R3. Because R3 is the most promoter–distal sequence, the data also suggest that uncoiling of the promoter–distal 100-bp region of nucleosomal DNA after Ø-loop formation is initiated at the promoter–distal end of nucleosomal DNA (Fig. 1C).

Taken together with our previous data (16), our current results suggest that all three R sequences cooperatively contribute to nucleosome positioning, R2 is primarily responsible for DNA–histone affinity, and R3 is largely responsible for the height of the nucleosomal barrier to transcription. Uncoiling of downstream DNA is likely initiated from the promoter–distal end of nucleosomal DNA.

The state of Pol II after encountering the high nucleosomal barrier was addressed using the 601 template that presents the strongest nucleosomal barrier to transcribing Pol II (14) (Fig. 2A). Pol II was spontaneously arrested at the positions +46 and +48 on the 601-nucleosomal template (Fig. 2B).

To determine the positions of the active center of Pol II, the +46 and +48 complexes were incubated in the presence of transcription elongation factor IIS (TFIIS). TFIIS is a transcription factor that strongly facilitates RNA cleavage at the active center of Pol II after backtracking (31). Incubation of the elongation complexes with TFIIS resulted in formation of homogeneous 42-mer RNA (Fig. 2B), indicating that both complexes were backtracked by 4 or 6 bp to the position +42. The nucleosomal +42, +46, and +48 complexes remain fully functional and can be reactivated upon nucleosomal removal in the presence of 1 M KCl (Fig. 2B). In summary, when Pol II encounters a strong nucleosomal barrier, it stops and backtracks by 4–6 bp. The backtracking could result in formation of a DNA gap between Pol II complex and the nucleosome (Fig. 2C), unless the nucleosomal DNA recoils back on the octamer surface.

The rotational orientation of Pol II at the +42 position after backtracking is incompatible with formation of the Ø loop (16). Therefore, we expected that nucleosomal DNA would be uncoiled from the octamer upstream of the +42 complex (Fig. 2C).

Accessibility of nucleosomal DNA in the +42 complex was analyzed using a restriction enzyme sensitivity assay (Fig. 3A). The nucleosome strongly protects DNA from digestion by restriction enzymes (32). In the elongation complex (EC) +42, only the AflIII site localized upstream of the active center (but not RsaI or MfeI sites localized downstream) is sensitive to cleavage (Fig. 3B). Thus, DNA in the +42 complex is uncoiled from the octamer upstream, but is tightly associated with histones downstream of the active center of the enzyme.

The structures of various elongation complexes invading the nucleosome were further analyzed by single-hit DNase I footprinting (16) (Fig. 4A). Such analysis requires high amounts of homogeneous complexes that can be obtained only with model E. coli RNA polymerase (RNAP) that uses the Pol II-specific mechanism of transcription through chromatin in vitro (14, 16, 33). Currently Pol II cannot be used for these experiments because even in the most advanced experimental systems in vitro the efficiency of escape from assembled or promoter-initiated complexes to stable elongation complexes is quite low (it occurs on less than 20% of templates) (27, 34), and purification of the productive elongation complexes is extremely difficult. In contrast, RNAP forms a nearly homogeneous +48 complex during transcription through the 601 nucleosome (Fig. S3). After formation of the complex, RNAP immediately backtracks, forming a stable and functional +42 complex (Fig. S4). Therefore, as expected, RNAP recapitulates all critical properties of Pol II transcription through the high nucleosomal barrier.

The −39, −5, and +42 complexes have distinct mobilities in the native gel, confirming that all complexes are nearly homogeneous and have unique conformations. As expected (16), each elongation complex protects the ∼30-bp region and the nucleosome protects the ∼150-bp region from DNase I digestion (Fig. 4 B and C). The EC −39 and EC −5 complexes have structures that are very similar to the structures of similar complexes formed on the permissive 603 template (16). The footprint of each complex is a sum of the footprint’s characteristic for the complex RNAP–DNA and nucleosome (Fig. 4 B and C).

RNAP completely uncoils upstream nucleosomal DNA in the +42 complex, whereas downstream DNA–histone interactions persist (Fig. 4 B and C). RNAP forms a tight complex with the nucleosome, with no DNase I-sensitive region between them. Two DNaseI-hypersensitive sites (hs-sites) are detected at the positions +96 and +102, surrounding the R3 DNA region that triggers arrest of Pol II at the +48 position (Fig. 1). The tight structure of the complex and the presence of the hs-sites suggest that after arrest, nucleosomal DNA is partially recoiled back on the histone octamer surface after backtracking of RNAP (Fig. 4B). RNAP collides with the promoter–distal end of the nucleosomal DNA, resulting in accumulation of DNA tension and formation of the hs-sites. The observation that the distorted region colocalizes with the PBS-R3 region is consistent with the key role of this sequence in formation of the high barrier to Pol II transcription (Fig. 1).

The structure of the arrested +42 complex is very similar to the structure of stalled EC +41 formed on the permissive 603 template (16). The striking similarities between the complexes formed at similar positions on the permissive 603 and nonpermissive 601 templates suggest that their structures are determined primarily by the position of the enzymes within the nucleosome. At the same time, EC formed on the permissive 603 nucleosomal template at similar positions (+49) has strikingly different properties: it is fully active, maintains its position on nucleosomal DNA, and partial DNA uncoiling from the octamer occurs downstream of the enzyme (16). The data suggest that RNAPs on the nonpermissive 601 nucleosome at position +48 are unable to form a stable Ø-loop intermediate, and instead backtrack by 4–6 bp to displace the upstream DNA from the octamer and to form the arrested EC +42. Thus, formation of the Ø loop and uncoiling of downstream DNA during Pol II transcription could occur during the single base pair step (+48 to +49).

Taken together, the data suggest that after arrest at +(46–48) positions in the nucleosome, the RNAPs backtrack to position +42. The enzymes remain functional, indicating that DNA–histone interactions themselves form the nucleosomal barrier to transcription. Backtracking of RNAPs is likely followed by recoiling of nucleosomal DNA on the surface of the histone octamer. Backtracking results in formation of a compact arrested +42 complex, with a close encounter between the RNAP and nucleosome. Formation of EC +42 is accompanied by formation of hypersensitive sites flanking the PBS-R3 sequence, most likely due to steric interference between the RNAP and promoter–distal end of nucleosomal DNA.

Next, the structure of the +42 complex of RNAP with nucleosome was determined using electron microscopy and single particle analysis (Fig. 5). The +42 EC has a combined molecular mass of about 545 kDa, making it difficult to observe them embedded in ice. Therefore, image contrast was enhanced using uranyl acetate negative staining (Fig. 5A). A total of 8,500 +42 EC particles were analyzed (Fig. 5 B and C). The preliminary 3D structure of the complex (Fig. 5D) was further refined. The final structure (Fig. 6) contains a larger (about 20 nm in diameter), and a smaller (about 14 nm in diameter) density. Fine structural features of the complex cannot be resolved at the obtained 22-Å resolution. Nevertheless, the obtained 3D structure allowed building of the model of the EC +42 elongation complex by docking of the crystal structures of RNAP (PDB code 4JKR) into the larger density, and nucleosome (PDB code 1KX5) into the smaller density (Fig. 6). Our EM data support the view that the EC +42 contains a single mononucleosome and one RNAP complex.

In the +42 complex, RNAP and the nucleosome are in close proximity to each other (Fig. 6). The spacer DNA connecting them is largely hidden by the proteins; thus the structure is consistent with DNase I footprinting data detecting no accessibility of the spacer DNA (Fig. 4). The tight structure of the complex is consistent with the proposal that in the arrested complex backtracking of RNAPs is followed by recoiling of nucleosomal DNA on the surface of histone octamer (Fig. 4B). Docking of a yeast Pol II elongation complex structure into the obtained model suggest that a similar complex can be formed by a eukaryotic Pol II (Fig. S5).

The most striking and surprising feature of the +42 complex is the opening of the DNA-interacting surface of the promoter–proximal histone dimer H2A/H2B. Indeed, the model of the +42 complex (Fig. 6 and Fig. S5) suggests that proximal H2A/H2B dimer is not stabilized by interactions with DNA or RNAP, but remains associated with the octamer. Furthermore, our previous data have shown that progression from the +42 to +49 complex is accompanied by full recoiling of nucleosomal DNA back on the surface of the dimer (16), suggesting that the dimer does not leave DNA before Pol II progresses more than 49 bp into the nucleosome. The stability of the dimer during transcription is remarkable, considering that the DNA-free octamer completely loses both H2A/H2B dimers in less than 1 s (35). Thus, the presence of nucleosomal DNA that remains partially coiled around the H3/H4 tetramer and promoter–distal H2A/H2B dimer results in a decrease of the rate of dissociation of the promoter–proximal dimer. The data suggest that binding of the promoter–proximal H2A/H2B dimer to H3/H4 tetramer in the +42 complex could be allosterically stabilized by the remaining DNA–histone interactions in the nucleosome. Because displacement of the dimer constitutes the first step to eviction of the entire histone octamer (36), this mechanism is likely important for nucleosome survival during various processes involving nucleosome remodeling.

Taken together with our previous footprinting studies of transcription-permissive nucleosomes, our data allow us to propose a detailed mechanism of Pol II transcription through the region of nucleosomal DNA [the +(40–50) region] that largely dictates the height of the nucleosomal barrier and nucleosome fate during transcription (14, 16) (Fig. 7 and Fig. S6).

As Pol II approaches a strong nucleosomal barrier [largely dictated by the R1, +(75–79) and by the R3, +(99–102) sequences, Fig. 1B and refs. 17, 28], it undergoes pausing or sequence-dependent arrest accompanied by Pol II backtracking along DNA (complex 2′, Figs. 2C and 7). The paused Pol II state (complex 2, Fig. 7) can be resolved by recoiling of DNA behind the enzyme back onto the surface of the histone octamer, forming a small loop on the nucleosome surface (complex 3, Ø loop) (16). In the complexes 2, 2′, and 3 Pol II induces partial tension in promoter–distal nucleosomal DNA (Figs. 4 and 7). However, only after formation of the complex 3 Pol II can drive partial unwrapping of promoter–distal end of nucleosomal DNA, clearing the way for further progression of the enzyme along DNA (complexes 4 and 5) (16). Unwrapping of the promoter–distal end of nucleosomal DNA and further Pol II progression can be blocked by certain DNA sequences; when present, they form strong DNA–histone interactions and block uncoiling of nucleosomal DNA from histones in front of the enzyme (complex 4, Figs. 1 and 7).

The plasmids containing 601 and 603 sequences were described previously (14). The 601 nucleosomal template for Pol II was prepared as described (14), where the 147-bp nucleosome positioning sequence was amplified by PCR and digested with TspRI (NEB) for further ligation to the assembled Pol II EC −39. To prepare the templates for E. coli RNAP transcription, the nucleosome positioning sequences digested by TspRI (NEB) was ligated through the TspRI site to the T7A1 promoter-bearing fragment (33). The ligated product was reamplified with 5′ end-labeled primers, gel purified, and assembled into nucleosomes. Details regarding the design of the templates and primer sequences will be provided upon request.

Hexahistidine-tagged E. coli RNAP was purified according to published protocols (50). GreB protein was purified as described (51). TFIIS was purified as described (52). Nucleosomes were reconstituted on end-labeled 226-bp DNA templates by histone octamer transfer from the chicken -H1 erythrocyte donor chromatin DNA at a 1:3 DNA-to-chromatin ratio (16). -H1 chromatin also provides unlabeled DNA that serves as a competitor for nucleosome formation. This approach allows semiquantitative comparison of relative free energies (affinities) of histone–DNA interactions (16, 53). The nucleosomes were analyzed in a native gel (16); the mobility of a nucleosome in the gel reflects nucleosome positioning.

E. coli RNAP elongation complexes containing 11-mer RNA (EC −39) were formed on preassembled nucleosomal templates as described (33). In the case of experiments with labeled RNA, EC −39 was pulse labeled in the presence of [α-³²P]-GTP (3,000 Ci/mmol, PerkinElmer Life Sciences). Then EC −39 was extended in the presence of a subset of NTPs to form EC −5 (Fig. S1). In the case of the footprinting experiments, all steps were performed in solution. In the case of experiments involving labeled RNA and GreB treatment, EC −5 was immobilized on Ni-NTA-agarose (33). After extensive washes, the complexes were eluted from Ni-NTA beads in the presence of 100 mM imidazole and transcription was continued in solution. To form EC +48 the EC −5 was extended in the presence of 10 μM NTPs at 25 °C for 15 min in TB300 (20 mM Tris⋅HCl pH 8.0, 5 mM MgCl₂, 2 mM β-ME, 300 mM KCl). Then the immobilized ECs were washed in TB40 (20 mM Tris⋅HCl pH 8.0, 5 mM MgCl₂, 2 mM β-ME, 40 mM KCl), eluted by 100 mM imidazole, and incubated with GreB. Labeled RNA was purified and separated by denaturing PAGE.

Pol II assembled elongation complex containing 9-mer RNA (EC −41), formed as described (14), was immobilized on Ni-NTA beads and extended in the presence of 20 μM ATP and GTP to form EC −39, washed, and eluted by 100 mM imidazole. EC −39 was ligated to the nucleosomal template through the TspRI site and transcribed to EC −5 and EC +48 as described for E. coli RNAP.

Because the complex of RNAP, stalled in position +42 of the 603 nucleosome, is more structurally homogeneous than the spontaneously arrested +42 complex on 601 nucleosome, the stalled +42 603 complex was used for the structural analysis. To obtain EC +42 stalled complex, the EC −5 was formed as described previously (16) and the sample was dialyzed against 2× TB300 in a cold room for 2 h to remove ATP in solution. The EC −5 was then extended in the presence of 20 μM CTP, UTP, and GTPs and 1 mM 3′ ATP (TriLink BioTechnologies) at 25 °C for 4 min to form EC +42.

DNaseI footprinting was carried out at a final concentration of 2.5 μg/mL of labeled templates in the presence of 10-fold weight excess of unlabeled -H1 chicken erythrocyte chromatin in TB100 (20 mM Tris⋅HCl pH 8.0, 5 mM MgCl₂, 2 mM β-ME, 100 mM KCl). DNaseI was added to the final concentration 20–50 units/mL for 30 s in 37 °C after formation of the desired ECs. The reactions were terminated by adding EDTA to 10 mM. The samples were resolved in a native gel (54). Gel fragments containing desired complexes were cut, DNA purified, and analyzed by denaturing PAGE. The gels were quantified using a PhosphorImager.

To reveal the EC +42 3D structure, negative staining of samples with 1% uranyl acetate was used. Freshly prepared EC +42 was applied to the carbon-coated glow-discharged grid, stained for 30 s with uranyl acetate, and air dried. Grids were studied in JEOL 2100 TEM, operated at 200 kV at low-dose conditions. Micrographs were captured by the Gatan CCD camera. The initial 3D reconstitution was obtained from random conical tilt (RCT) data with −45/0 degrees tilt. Using EMAN2 software 2,500 tilted pairs were picked up, using the e2RCTboxer.py command. The correction for contrast transfer function of the microscope was performed in EMAN2 using command e2ctf.py. Reference free class averages (Fig. 5A) were obtained using the e2refine2d.py command. The 3D structure of EC +42 (Fig. 5D) was revealed with the resolution of 30 Å. An additional 6,000 particles were added to the set and three structurally different sets of images were separated using IMAGIC5. One set was RNA polymerase alone (5% of particles). Two other sets differed by the position of the nucleosome: more distant from RNA polymerase (85% of all particles) and the ones where the nucleosome is closer to the RNAP (10% of all particles). The most representative model was further refined using 85% of particles to obtain a final 3D structure of EC +42 at a resolution of 22 Å, estimated by Fourier shell correlation method at 0.5 (Fig. 6). Multimeric statistical analysis (MSA) of aligned particles produced eigen images (Fig. S7) and demonstrated that there are two distinct parts of the complex. The contour level was chosen based on an average protein density of 810 Da/nm³ and volume corresponded to ∼545 kDa for EC +42. The docking of the molecular model of EC +42 complex (see below) into obtained 3D density resulted in cross-correlation coefficient 0.73.

To model the +42 RNAP–nucleosome complex, the X-ray structure of Thermus thermophilus RNAP elongation complex (55) was at first combined with the X-ray structure of a nucleosome core particle (56) by elongating its downstream DNA duplex with a segment of straight B-DNA and connecting it to various corresponding positions along the nucleosomal DNA while respecting the correct DNA geometry and chemical bonding. The T. thermophilus RNAP was then replaced by the E. coli RNAP structure (57) via structure superposition; the structures of the two enzymes are very similar (Fig. S8). The modeling pipeline was based on University of California San Francisco Chimera (58) and Nucleic Acid Builder (NAB) (59) programs. The obtained family of models was fitted automatically into an experimentally obtained electron density map using the Chimera EM fitting utility (the atomistic model was converted into a simulated electron density map with 30-Å resolution for fitting purposes and extensive search with 40 different trial starting placements was done). The best model was selected based on maximum cross-correlation of electron densities. In the obtained model, the active center of the RNAP is in position +42; the downstream duplex of DNA enters the nucleosome at position +71 (3 bp before the position of the dyad axis). Finally, the structure of yeast Pol II elongation complex (PDB ID 1Y1W) (60) was superimposed on the E. coli RNAP structure to obtain the Pol II +42 elongation complex (Fig. S5).